RCSED - A Value-Added Reference Catalog of Spectral Energy Distributions of 800,299 Galaxies in 11 Ultraviolet, Optical, and Near-Infrared Bands: Morphologies, Colors, Ionized Gas and Stellar Populations Properties
Igor Chilingarian, Ivan Zolotukhin, Ivan Katkov, Anne-Laure Melchior, Evgeniy Rubtsov, Kirill Grishin
DDraft version December 8, 2016
Preprint typeset using L A TEX style emulateapj v. 01/23/15
RCSED – A VALUE-ADDED REFERENCE CATALOG OF SPECTRAL ENERGY DISTRIBUTIONS OF 800,299GALAXIES IN 11 ULTRAVIOLET, OPTICAL, AND NEAR-INFRARED BANDS: MORPHOLOGIES, COLORS,IONIZED GAS AND STELLAR POPULATIONS PROPERTIES
Igor V. Chilingarian , Ivan Yu. Zolotukhin , Ivan Yu. Katkov , Anne-Laure Melchior , Evgeniy V.Rubtsov , Kirill A. Grishin Smithsonian Astrophysical Observatory, 60 Garden St. MS09, Cambridge, MA, 02138, USA Sternberg Astronomical Institute, M.V. Lomonosov Moscow State University, 13 Universitetsky prospect, Moscow, 119991, Russia Universit´e de Toulouse; UPS-OMP, IRAP, 9 avenue du Colonel Roche, BP 44346, F-31028 Toulouse Cedex 4, France Special Astrophysical Observatory of the Russian AS, Nizhnij Arkhyz 369167, Russia Sorbonne Universit´es, UPMC Univ. Paris 6, Observatoire de Paris, PSL Research University, CNRS,UMR 8112, LERMA, Paris,France and Department of Physics, M.V. Lomonosov Moscow State University, 1, Leninskie Gory, Moscow, Russia, 119991
Draft version December 8, 2016
ABSTRACTWe present RCSED a , the value-added Reference Catalog of Spectral Energy Distributions of galaxies,which contains homogenized spectrophotometric data for 800,299 low and intermediate redshift galax-ies (0 . < z < .
6) selected from the Sloan Digital Sky Survey spectroscopic sample. Accessiblefrom the Virtual Observatory (VO) and complemented with detailed information on galaxy propertiesobtained with the state-of-the-art data analysis, RCSED enables direct studies of galaxy formationand evolution during the last 5 Gyr. We provide tabulated color transformations for galaxies of dif-ferent morphologies and luminosities and analytic expressions for the red sequence shape in differentcolors. RCSED comprises integrated k -corrected photometry in up-to 11 ultraviolet, optical, andnear-infrared bands published by the GALEX, SDSS, and UKIDSS wide-field imaging surveys; resultsof the stellar population fitting of SDSS spectra including best-fitting templates, velocity dispersions,parameterized star formation histories, and stellar metallicities computed for instantaneous starburstand exponentially declining star formation models; parametric and non-parametric emission line fluxesand profiles; and gas phase metallicities. We link RCSED to the Galaxy Zoo morphological classifica-tion and galaxy bulge+disk decomposition results by Simard et al. We construct the color–magnitude,Faber–Jackson, mass–metallicity relations, compare them with the literature and discuss systematicerrors of galaxy properties presented in our catalog. RCSED is accessible from the project web-siteand via VO simple spectrum access and table access services using VO compliant applications. Wedescribe several SQL query examples against the database. Finally, we briefly discuss existing andfuture scientific applications of RCSED and prospectives for the catalog extension to higher redshiftsand different wavelengths. Keywords: galaxies: (classification, colors, luminosities, masses, radii, etc.) — galaxies: photom-etry — galaxies: stellar content — galaxies: fundamental parameters — AstronomicalDatabases: catalogs — Astronomical Databases: virtual observatory tools INTRODUCTION AND MOTIVATION
During the last decade we witnessed a breakthroughin wide field imaging surveys across the electromagneticspectrum. The new era started with the Sloan DigitalSky Survey (SDSS) that used a 2.5-m telescope and cov-ered over 11,600 sq. deg. of the sky in 5 optical photomet-ric bands ( ugriz ) down to the 22nd AB magnitude in itslatest 7th legacy data release (Abazajian et al. 2009). Ithad a spectroscopic follow-up survey that targeted over1 million galaxies and quasars and half a million starsdown to the magnitude limit of r = 17 .
77 AB mag. Eventhough by the end of 2015, the data from SDSS and itssuccessors, SDSS-II, and SDSS-III were used in about20,000 research papers , the SDSS potential for scien-tific exploration remains far from exhaustion. A E-mail: [email protected] B E-mail: [email protected] a The data tables and other supporting technical information areavailable at the project web-site: http://rcsed.sai.msu.ru/ According to NASA ADS, http://ads.harvard.edu/
In the late 2000s, deep wide field surveys went beyondthe optical spectral domain. The Galaxy Evolution Ex-plorer (GALEX) satellite (Martin et al. 2005) providednearly all-sky photometric coverage in two ultravioletbands centered at 154 and 228 nm down to the limitingmagnitudes AB = 20 . AB = 23 . Y JHK ). The Large AreaSurvey of the UKIRT Deep Sky Survey (UKIDSS LAS,Lawrence et al. 2007) provides a sub-arcsecond resolutionand the flux limit comparable to that of SDSS in the op-tical domain. It reaches AB ∼ . a r X i v : . [ a s t r o - ph . GA ] D ec Chilingarian et al.
IRAS point source catalog (Saunders et al. 2000), theFaint Images of the Radio sky 20 cm survey FIRST(Becker et al. 1995), and additional data on galaxiesfrom the 3rd Reference Catalogue of Bright Galaxies(de Vaucouleurs et al. 1991) and the Two-Degree FieldGalaxy Redshift Survey (Colless et al. 2001). Now, adecade after the NYU-VAGC has been published, thereis a sharp need to assemble a next generation of a value-added galaxy catalog based on modern survey data thatwere not available back then.Here we present a new generation and a different flavorof a value-added catalog of galaxies based on a combina-tion of data from SDSS, GALEX, and UKIDSS surveysthat also includes comprehensive analysis of absorptionand emission lines in galaxy spectra. Our main motiva-tion is to use the synergy provided by the joint panchro-matic dataset for extragalactic astrophysics: the opticaldomain is traditionally the best studied and there existwell calibrated stellar population models; the UV fluxesare sensitive to even small fractions of recently formedstars and therefore contain valuable information on starformation histories; the near-IR band is substantially lesssensitive to the internal dust reddening and stellar pop-ulation ages, and therefore can provide good stellar massestimates. Our mission is to build a reference multi-wavelength spectrophotometric dataset and complementit with additional detailed information on galaxy proper-ties so that it will allow astronomers to study galaxy for-mation and evolution at redshifts z = 0 . . k -corrections in opticaland NIR bands (Chilingarian et al. 2010). Then we ex-tended our algorithm to GALEX FUV and
NUV bandsand discovered a universal 3-dimensional relation of
NUV and optical galaxy colors and luminosities (Chilingarian& Zolotukhin 2012). Then, we fitted SDSS spectra us-ing state-of-the-art stellar population models, derived ve-locity dispersions and stellar ages and metallicities, andprovided our measurements to the project that calibratedthe fundamental plane of galaxies (Djorgovski & Davis1987) in SDSS by vigorous statistical analysis (Saulderet al. 2013). Our dataset also helped to find and charac-terize massive compact early-type galaxies at intermedi-ate redshifts (Damjanov et al. 2013, 2014). Finally, weused a complex set of selection criteria and discovereda large sample of previously considered extremely rarecompact elliptical galaxies (Chilingarian & Zolotukhin2015).The paper is organized as follows: in
Section 2 we de-scribe the construction of the catalog that includes cross-matching of the three surveys, adding third-party cata-logs, absorption and emission line analysis of SDSS spec-tra; in
Section 3 we discuss the photometric properties ofthe sample and derive mean colors of galaxies of differentmorphological types across the spectrum; in
Section 4 weexplore the information derived from our spectral anal-ysis;
Section 5 contains the description of the catalogaccess interfaces;
Section 6 provides the summary of ourproject; and
Appendices include some technical detailson the catalog construction, detailed description of ta-bles included in the database, and discussion of system-atic uncertainties of emission line measurements. CONSTRUCTION OF THE CATALOG
The input sample and data sources used.
We compiled the photometric catalog by re-processingseveral publicly available datasets. Our core object listis the SDSS Data Release 7 (Abazajian et al. 2009) spec-tral sample of non-active galaxies (marked as “GAL EM”or “GALAXY” specclass in the SDSS database) in theredshift range 0 . ≤ z < .
6. We provide the exactquery that we used to select this sample in the SDSSCasJobs Data System in Appendix B. The query ex-ecuted in the DR7 CasJobs context returned 800,299records. We deliberately excluded quasars and Seyfert-1 galaxies (specclass=“QSO”) because neither the k -correction technique, nor stellar population analysis al-gorithm supported that object type. We used the out-put table as an input list for positional cross-matchesagainst GALEX Data Release 6 (Martin et al. 2005) andUKIDSS Data Release 10 (Lawrence et al. 2007).For the UKIDSS cross-match we queried the UKIDSSLarge Area Survey catalog using the best match criterionwithin a 3 arcsec radius. In order to perform this query,we employed the WFCAM Science Archive for the pro-grammatic access to the International Virtual Observa-tory Alliance (IVOA) ConeSearch service with a mul- http://skyserver.sdss3.org/CasJobs/ http://surveys.roe.ac.uk/wsa/ CSED – Reference Catalog of Galaxy SEDs ◦ ◦ ◦ − ◦ +90 ◦ Figure 1.
A full sky aitoff projection in equatorial coordinatesdemonstrating the footprint of our catalog. Green areas denotethe availability of all three input photometric datasets, SDSS,UKIDSS, and GALEX; red areas are for SDSS and GALEX; andblue areas are for SDSS only. Note that we include all objectsfrom the input datasets that have at least one flux measurementin them.
Table 1
Number of objects in the combined samplewith photometric measurements available fromthree input photometric catalogs.Photometric bands Number of galaxiesSDSS ugriz
F UV + ugriz NUV + ugriz F UV + NUV + ugriz ugriz + UKIDSS Y ugriz + UKIDSS J ugriz + UKIDSS H ugriz + UKIDSS K ugriz + Y JHK
NUV + ugriz + Y JHK tiple cone search (“multi-cone”) capability. The queryreturned 280,870 UKIDSS objects matching the galax-ies from our input sample. We used the stilts softwarepackage (Taylor 2006) in order to access the UKIDSSdata and merge the tables.Then we uploaded the input SDSS galaxy list tothe GALEX CasJobs web interface and searched bestmatches within 3 arcsec similarly to the UKIDSS cross-match. The query returned 485,996 GALEX objects.As a result of this selection procedure we compiledan input catalog of 800,299 spectroscopically confirmedSDSS galaxies, out of which 90,717 have 11 band photom-etry (two GALEX FUV and
NUV , 5
SDSS ugriz bands,4 UKIDSS
Y JHK bands), 163,709 have all UKIDSSbands and at least one UV band, 582,534 have at leastone additional photometric band to SDSS bands. InFig. 1 we present the footprint of our catalog on theall-sky aitoff projection marking the regions covered byall three wide field imaging surveys using different colors.The statistics of galaxies measured in different photomet-ric bands is given in Table 1.Then we linked the following published datasets to ourcatalog in order to contribute the spectrophotometricinformation with some of the most widely used galaxyproperties: (i) the results of the two-dimensional light http://galex.stsci.edu/casjobs/ profile decomposition of SDSS galaxies by Simard et al.(2011) that include structural properties of all objectsin our catalog; (ii) the morphological classification tablefrom the citizen science “Galaxy Zoo” project (Lintottet al. 2008, 2011) that provides a human eye classifica-tion of well spatially resolved SDSS galaxies made bycitizen scientists. 661,319 objects in our sample have 10or more morphological classifications in the Galaxy Zoocatalog ( nvote ≥ The photometric catalog
Petrosian and aperture magnitudes
All three photometric surveys used in our study pro-vide extended source photometry along with aperturemeasurements made in several different aperture sizes(GALEX and UKIDSS).For the SED photometric analysis and construction ofscaling relations involving galaxy luminosity, we needtotal magnitudes. For this purpose we adopt Pet-rosian (1976) magnitudes available in SDSS and UKIDSSas measurements which do not significantly depend ongalaxy light profile shapes conversely to SDSS model-mags (see discussion in Chilingarian & Zolotukhin 2012).The GALEX catalog provides “total” magnitudes thatare close to Petrosian magnitudes for exponential surfacebrightness profiles (i.e. disc galaxies) and up-to 0.2 magbrighter for elliptical galaxies (Yasuda et al. 2001). How-ever, given the average photometric uncertainty in theGALEX
N U V fluxes of red galaxies of 0 . fibermags for the optical SED part. To be noticed,that the spatial resolution of the GALEX survey in the N U V band is about 5 arcsec, therefore 3-arcsec aperturemagnitudes will be slightly underestimated for small ob-jects. For compact (point-like) sources a 3-arcsec aper-ture
N U V magnitude can be underestimated by as muchas 0.3 mag, however, such objects are very rare in theSDSS DR7 galaxy sample. Damjanov et al. (2013); Zahidet al. (2015, 2016) found a couple of thousands compactsources in SDSS and SDSS- iii
BOSS, only a few hundredsof which were in SDSS DR7. We estimated a number ofcompact galaxies in our sample by selecting the sourceswhere the average difference of aperture and Petrosianmagnitudes in ugriz bands was < . < .
1% of the sample.We corrected the obtained sets of Petrosian and 3-arcsec aperture magnitudes for the Galactic foregroundextinction by using the E ( B − V ) values computed fromthe Schlegel et al. (1998) extinction maps. Then, we com-puted k -corrections for both sets of photometric pointsusing the analytic approximations presented in Chilin-garian et al. (2010) and updated for GALEX bands inChilingarian & Zolotukhin (2012).In Fig. 2 we provide an example of a fully corrected
Chilingarian et al.
250 500 1000 2000
Wavelength, nm -28 -27 -26 F ν , e r g s − c m − H z − A B m a g n i t u d e fuv nuv u g r i z y j h k Figure 2.
Example of fully corrected SED in 11 bands for a latetype spiral galaxy at redshift 0.035. Blue and red symbols representtotal (Petrosian) and 3-arcsec fiber magnitudes correspondingly.The rest framed SDSS spectrum is overplotted and demonstratesa typical excellent agreement with the corrected fiber magnitudesfor that galaxy. The inset shows an 36 ×
36 arcsec optical SDSSfalse color image.
SED for a late type spiral galaxy ( z = 0 . AB magnitudes. One can see aremarkable agreement between the corrected photomet-ric points and the observed spectral flux density, typicalfor our catalog. Correcting the SDSS–UKIDSS photometric offset
An important problem of the UKIDSS photometriccatalog of extended sources is the observed spread ofcolors including optical SDSS and NIR UKIDSS photo-metric measurements (e.g. g − J for red sequence galax-ies). We detected this inconsistency in Chilingarian et al.(2010) and applied an empirical correction to UKIDSSmagnitudes based on the assumption of continuous SEDsof galaxies. We computed z − Y colors by interpolatingover all other available colors approximating the SEDwith a low order polynomial function. This approach,however, required the availability of the Y band pho-tometry in the UKIDSS catalog. We have analyzed theSDSS–UKIDSS Petrosian magnitude offset amplitude fordifferent galaxies and concluded, that it originates fromthe surface brightness limitation imposed by relativelyshort integration time in UKIDSS and by high and vari-able sky background level in the NIR. Therefore, Pet-rosian radii and magnitudes become underestimated, andcomparison of original UKIDSS extended source magni-tudes with SDSS and GALEX integrated photometry be-comes impossible, because any color including data fromUKIDSS and another data source depends on the galaxysurface brightness and size.Here we propose a general and simple empirical solu-tion. We exploit the UKIDSS Galactic Cluster Surveyphotometric catalog that includes the Z band photome-try, convert it into SDSS z with the available color trans-formation (Hewett et al. 2006) for both Petrosian and3-arcsec aperture magnitudes, and compare it to actualSDSS z band measurements from the SDSS DR7 catalogfor exactly the same objects. It turns out, that (i) thePetrosian magnitude difference z SDSS , Petro − z UKIDSS , Petro F l u x , E - e r g / c m / s / A n g Figure 3.
Example of the nbursts full spectrum fitting for anSDSS spectrum of an early type galaxy. An observed galaxy spec-trum is shown in black, the best fitting template is in red, residualsare in blue. Regions of emission lines excluded from the fitting areshown red in the residuals. The observed and rest-frame wave-length are shown in the bottom and top of the plot respectively. correlates with the galaxy mean surface brightness; (ii)the fiber magnitude difference z SDSS , fib − z UKIDSS , (cid:48)(cid:48) isclose to zero within 0.02 mag; (iii) differences betweenPetrosian and fiber magnitudes in all UKIDSS photomet-ric bands ( ZY JHK ) are almost identical that indicatesvirtually flat NIR color profiles in most galaxies. Thissuggests that the correction for UKIDSS Petrosian mag-nitudes should be calculated as: ∆( mag
UKIDSS , Petro ) =( z SDSS , fib − z SDSS , Petro ) − ( Y UKIDSS , (cid:48)(cid:48) − Y UKIDSS , Petro ).This transformation adjusts the UKIDSS integrated pho-tometry in a way that the differences between the 3 arc-sec and Petrosian magnitudes of a galaxy in z and Y bands become equal. For objects, where Y magnitudesare not available in the UKIDSS survey, we use the nextavailable photometric band ( J , H , or K ).In this fashion, we obtained fully corrected FUV-to-NIR spectral energy distributions converted into rest-frame magnitudes for a large sample of galaxies in 3-arcsec apertures and integrated over entire galaxies.
The spectral catalog: absorption lines
We fitted all SDSS spectra using the nbursts full spec-trum fitting technique (Chilingarian et al. 2007a,b) anddetermined their radial velocities v , stellar velocity dis-persions σ , and parameterized star formation historiesrepresented by an instantaneous star burst (simple stel-lar populations, SSP) or an exponentially declining starformation history (exp-SFH) assuming that it startedshortly after the Big Bang. We chose these two fami-lies of stellar population models because: (i) SSP modelsare widely used in extragalactic studies by different au-thors and we wanted our data to be directly comparableto other sources; (ii) exponentially declining SFHs weredemonstrated to be a better representation of broadbandSEDs of non-active galaxies (Chilingarian & Zolotukhin2012) than SSPs. We should, however, notice, that exp-SFH models cannot adequately describe young stellarpopulations with mean ages t < . CSED – Reference Catalog of Galaxy SEDs t –[Fe/H]) space, thenconvolving it with a Gaussian-Hermite representation ofthe line-of-sight velocity distribution (LOSVD) of starsin a galaxy described by v, σ, h , h
4, and finally multiply-ing it by a low order Legendre polynomial continuum (itsparameters are determined linearly in a separate loop) inorder to absorb flux calibration imperfections and possi-ble internal extinction in a galaxy. Hence, the procedurereturns values of v, σ, h , h , t, [Fe/H], and coefficients ofthe multiplicative polynomial continuum. Here we use apure Gaussian LOSVD shape with h h nbursts algorithm is similar to the penalizedpixel fitting approach by Cappellari & Emsellem (2004).It, however, has some important differences. (i) We usea linear fit of the low order multiplicative polynomialcontinuum because its parameters are decoupled fromgalaxy kinematics and stellar populations. (ii) Insteadof using a fixed grid of template spectra and interpret-ing stellar populations using their relative weights in alinear combination, we interpolate in a grid of modelsinside the minimization loop in order to obtain the best-fitting stellar population parameters of each starburst (oran exponentially declining model). As a result, for thesimplest case of a single component SSP model, we ob-tain the best-fitting SSP-equivalent age and metallicity.These values are usually close to the luminosity weightedones, however, in cases of complex SFHs approximatedby an SSP there might be biases similar to those affect-ing Lick indices (Serra & Trager 2007). Chilingarianet al. (2007b, 2008) demonstrated that SSP equivalentages and metallicity remain unbiased for galaxies withsuper-solar α -element abundances ([Mg/Fe] > i , NaD, OH, etc.) and by the A and B telluric absorption bands from the fitting procedure.We also re-ran the fitting code excluding 8–14 ˚A-wideregions around locations of bright emission lines for ob-jects, where the reduced χ value of the fit exceeded thethreshold χ /DOF =0.8, that was selected empiricallyfrom a sample of galaxies without and with emission linesof different intensity levels. We used three grids of stellar population models allcomputed with the pegase.hr evolutionary synthesiscode (Le Borgne et al. 2004):1. SSP models based on the high resolution(R=10000) ELODIE.3.1 empirical stellar library(Prugniel & Soubiran 2004; Prugniel et al. 2007)covering the wavelength range 3900 < λ < − . < [Fe/H] < . < t < Published SDSS spectra are slightly oversampled in wave-length, therefore, flux uncertainties in neighboring pixels are cor-related and, hence, the reduced χ for a spectrum well representedby its model is less than 1 (around 0.6). ity and wavelength ranges as those for SSP models,covering the range of exponential decay timescales10 < τ < . . < z < .
5. Weused the exponential decay timescale τ in the samefashion as the SSP age in the minimization pro-cedure. For every galaxy, we first computed agrid of τ − [Fe/H] models with the star formationepoch corresponding to its redshift assuming thata galaxy was formed at very high redshift, e.g. for z = 0 . R = 2300)based on the MILES empirical stellar library(S´anchez-Bl´azquez et al. 2006) covering the wave-length range 3600 < λ < − . < [Fe/H] < . < t < t > τ < < λ < pegase.hr mod-els with the 5-th degree of the multiplicative polynomialcontinuum; (2) pegase.hr based exponentially decliningSFR models in the wavelength range 3915 < λ < χ value reaches a “plateau” as ex-plained in Chilingarian et al. (2008). We performed theSSP fitting with the MILES- pegase.hr models in thetruncated wavelength range with a very low order poly-nomial continuum in order to minimize the artifacts orig-inating from imperfections in the SSP model grid (seeSection 4.2). In the publicly available Simple SpectrumAccess Service we provide the results of the MILES- pegase.hr based SSP fitting in the wavelength range3600 < λ < ii ] 3727 ˚A doublet. The spectral catalog: emission lines
Our full spectral fitting procedure precisely matchesthe stellar continuum of each galaxy by the best-fittingstellar population model (see example in Fig. 3). Al-though the regions of all Balmer absorption lines are agesensitive, they contain at most 20% of the age sensi-tive information from the entire optical spectral range(Chilingarian 2009). Chilingarian et al. (2007b) havedemonstrated that masking the H β and H γ regions bi-ases neither age nor metallicity determinations by the Chilingarian et al.nbursts procedure. Hence, we do not expect to intro-duce significant template mismatch by masking the re-gions of emission lines when fitting SDSS spectra. Havingsubtracted the best-fitting model we obtain clean emis-sion line spectra unaffected by stellar absorptions that isespecially important for the Balmer lines. The precisionof our stellar continuum fitting allows us to recover faintemission lines at a few per cent level of the continuum in-tensity, whereas very often such lines are not detected inthe SDSS spectral pipeline results. In Table 2 we providethe statistics of the emission line detection and strengthin our sample.In order to measure fluxes and equivalent widths (EW)of emission lines we applied two different approaches,namely Gaussian and non-parametric fitting of emissionlines profiles.In some galaxies, emission lines profiles cannot bedescribed by a Gaussian. This often becomes a casein galaxies with peculiar gas kinematics, e.g. multi-component bulk gas motions and outflows can producecomplex asymmetric lines. Also, this is crucially impor-tant for active galactic nuclei (AGN) with broad compo-nents in Balmer lines. Approximation of such emissionlines by a Gaussian profile results in biased estimates offlux and kinematic parameters. We address this problemby employing a non-parametric fitting approach whichallows us to recover arbitrary line profiles and measuretheir fluxes with higher precision. At the same time,this method requires several lines with sufficiently highsignal-to-noise ratio to be present in a spectrum, and mayproduce biased results when dealing with noisy data. We,therefore, perform a “classical” Gaussian profile fittingtoo in order to allow for cross-comparison and validationof our line fitting results.Both non-parametric and Gaussian fitting techniquestake into account the SDSS line-spread function com-puted individually for each spectrum by the standardSDSS pipeline and provided in FITS (Flexible ImageTransport System) tables in the RCSED distribution.
Gaussian fitting
This approach consists of simultaneously fitting the en-tire set of emission lines (see the line list in Table 2) withGaussians pre-convolved with the SDSS line-spread func-tion. We allow two different sets of redshifts and intrinsicwidths for recombination and forbidden lines. We esti-mate the kinematic parameters with the non-linear least-square minimization that is implemented by the MPFITpackage (Markwardt 2009). The emission line fluxes arecomputed linearly for each minimization iteration. Whensolving the linear problem, we constrain the line fluxesto be non-negative. For this purpose we use the BVLS(bounded-variables least-squares) algorithm (Lawson &Hanson 1995) and its implementation by M. Cappellari . Non-parametric emission line fitting
Our non-parametric emission line fitting method in-cludes two main steps which we repeat several times untilthe convergence is achieved. First, we derive discretelysampled emission line profiles, i.e. line-of-sight velocity distributions (LOSVDs) of ionized gas. During the sec-ond step, we estimate emission lines fluxes. Because al-lowed and forbidden transitions often originate from dif-ferent regions of a galaxy having very different physicalproperties (i.e. density, temperature, mechanism of exci-tation), however, all emission lines of each type (allowedand forbidden) have similar shapes, our procedure re-covers two different non-parametric profiles, one for eachtype.The LOSVD derivation is organized as follows. Wenote that convolution of any logarithmically rebinnedobserved spectrum S obs of m elements with LOSVD L having n elements can be expressed as a linear matrixequation A ∗ L = S obs , where A is a m × n matrix oftemplate spectra having lengths of m pixels each. Everytemplate spectrum from the i -th row in the matrix A isshifted by the velocity which represents the i -th positionwithin the LOSVD vector. Here a template spectrum isa synthetic spectrum made of a set of flux normalizedGaussians with LSF widths representing emission linesdetected in the observed spectrum. Such approach allowsus to take into account the SDSS instrumental resolutioninstead of a set of Dirac δ -functions. The continuum levelof a template spectrum is set to zero.Thus, we end up with a linear inverse problem whosesolution L can be derived by a least square technique.The emission line profiles obviously cannot be negativeand, therefore, we use the BVLS algorithm mentionedabove.Once the LOSVD has been derived, we compute emis-sion line fluxes by solving a similar linear problem tothat described in Section 2.4.1. This finishes the firstiteration.At the same time, the LOSVD derivation step requiresthe knowledge of emission line fluxes in order to constructbetter template spectra. During the first iteration whenthey are unknown, we set all fluxes to unity and the alltemplate spectra hence are made of equally normalizedGaussians. Typically, 3 iterations of this procedure isenough to reach the convergence.A linear inversion is an ill-conditioned problem and issensitive to noise in the data. In order to improve theprofile reconstruction quality, we exploit a regularizationtechnique, which minimizes the squared third derivativeof the recovered line profile. This approach, however,causes artefacts in sharp narrow line profiles. Therefore,we apply the regularization only in the wings of emis-sion lines where flux levels are generally low and, conse-quently, noise is higher. The regularization techniqueyields the dramatic improvement of recovered Balmerline profiles for faint AGNs. In the catalog we providemeasurements of non-parametric emission lines with andwithout regularization.The comparison between the parametric (Gaussian)and non-parametric fitting results for a complex emis-sion line profile in a Seyfert galaxy is presented at Fig. 5.A Gaussian approximation for such lines is often inad-equate and causes serious biases in the kinematics thatcan reach few hundred km s − .We ran Monte-Carlo simulations for a random sampleof 2,000 objects with emission lines of different intensitylevels in order to estimate realistic flux uncertainties ob-tained with the non-parametric fitting technique. Theyturned out to be consistent with statistical uncertain- CSED – Reference Catalog of Galaxy SEDs Wavelength, A F l u x , E - e r g / c m / s / A n g [OII] λ H δ λ H γ λ +[OIII] λ H β λ [OIII] λ [NI] λ HeI λ [OI] λ H α +[NII] [SII] λ RCSED mjd=53050 plate=1362 fiberid=209
Figure 4.
Example of the nbursts full spectrum fitting for an SDSS spectrum of a late type galaxy together with the emission line fitting.Central panel is similar to Fig. 3, panels on the sides demonstrate recovered profiles after the continuum subtraction of individual emissionlines (black) and the best-fitting models (red). Blue lines show emission line flux uncertainties. Vertical red dashed lines represent thegalaxy redshift in the SDSS database. ties of Gaussian emission line fluxes for most objects andup-to a factor of 2 lower for AGNs with broad line com-ponents. The RCSED database will be updated withMonte-Carlo based uncertainties as we compute them:this procedure is very computationally intensive and willtake a couple of months to complete.
Gas phase metallicities
We used our emission line flux measurements in orderto estimate the gas phase metallicities for galaxies whereemissions originate from the star formation induced exci-tation. We exploited two different techniques to measurethe metallicity, (i) a new calibration by Dopita et al.(2016) and (ii) the IZI Bayesian technique (Blanc et al.2015) using a grid of models ( κ = ∞ ) with κ -distributedelectron energies (Dopita et al. 2013). We selected starformation dominated and “transition type” galaxies us-ing the standard BPT (Baldwin et al. 1981) diagram thatexploits hydrogen, nitrogen, and oxygen emission lineswith the criteria defined in Kauffmann et al. (2003).The Dopita et al. (2016) calibration uses only the H α ,[N ii ], and [S ii ] emission lines, all located in a very nar-row spectral interval and is, therefore, virtually insensi-tive to the internal extinction within an observed galaxy. This calibration is presented in a form of a simple for-mula which makes it very easy to use. However, a dis-advantage of this approach in application to our datasetis that at redshifts z > . λ < z = 0 .
36 when the forbidden sul-fur line [S ii ] 6730 ˚A shifts out of the wavelength range.Besides, the calibration critically depends on the [N/O]relation and is therefore sensitive to possible galaxy togalaxy [N/H] abundance variations. In the catalog weincluded the metallicity estimates obtained using the Do-pita et al. (2016) calibration for Gaussian emission lineanalysis ( rcsed gasmet table). We computed the uncer-tainties of the gas phase metallicities by propagating thestatistical flux errors through the calculations accordingto the formula in Dopita et al. (2016).The IZI technique (Blanc et al. 2015) takes advantageof all available emission line measurements and, hence,is more robust and can in principle be used for the en- Chilingarian et al.
Table 2
Emission line detection statistics (the parametric Gaussian fit) at different signal-to-noise ratio (SNR). The “Covered” columnreflects the number of objects with a corresponding line in the wavelength coverage. The “Wavelength” column provides airwavelengths. The “Prefix” column gives the prefix of column names for the corresponding spectral line in the emission line FITStables from the RCSED distribution.Line Wavelength Prefix Covered
SNR > SNR > SNR > SNR > ii ] 3726.03 f3727 oii 780665 543374 69.6% 354307 45.4% 225669 28.9% 92161 11.81%[O ii ] 3728.82 f3730 oii 782417 562707 71.9% 387264 49.5% 257713 32.9% 110287 14.10%H κ ι θ η iii ] 3868.76 f3870 neiii 798635 246202 30.8% 45733 5.7% 19500 2.4% 7579 0.95%He i ζ (cid:15) ii ] 4068.60 f4070 sii 799003 202235 25.3% 14766 1.8% 2352 0.3% 149 0.02%[S ii ] 4076.35 f4078 sii 799010 149342 18.7% 4755 0.6% 517 0.1% 55 0.01%H δ γ iii ] 4363.21 f4364 oiii 799293 118667 14.8% 8001 1.0% 2569 0.3% 787 0.10%He ii iv ] 4711.37 f4713 ariv 799381 79310 9.9% 3477 0.4% 530 0.1% 110 0.01%[Ar iv ] 4740.17 f4742 ariv 799380 118077 14.8% 7031 0.9% 1108 0.1% 85 0.01%H β iii ] 4958.91 f4960 oiii 799372 449021 56.2% 164285 20.6% 92442 11.6% 47287 5.92%[O iii ] 5006.84 f5008 oiii 799371 638852 79.9% 404135 50.6% 244845 30.6% 119215 14.91%[N i ] 5197.90 f5199 ni 799360 144430 18.1% 9742 1.2% 1345 0.2% 178 0.02%[N i ] 5200.25 f5202 ni 799360 184255 23.1% 16318 2.0% 2676 0.3% 226 0.03%[N ii ] 5754.59 f5756 nii 799131 196670 24.6% 12800 1.6% 2763 0.3% 966 0.12%He i i ] 6300.30 f6302 oi 784763 439640 56.0% 177144 22.6% 77850 9.9% 16856 2.15%[O i ] 6363.78 f6366 oi 780219 285886 36.6% 40626 5.2% 9143 1.2% 1395 0.18%[N ii ] 6548.05 f6550 nii 764832 596254 78.0% 422810 55.3% 289961 37.9% 133553 17.46%H α ii ] 6583.45 f6585 nii 761376 641883 84.3% 553212 72.7% 479386 63.0% 334901 43.99%He i ii ] 6716.43 f6718 sii 745687 571758 76.7% 423064 56.7% 320126 42.9% 186973 25.07%[S ii ] 6730.81 f6733 sii 743742 554071 74.5% 374143 50.3% 263155 35.4% 135572 18.23% tire sample of SDSS galaxies. The algorithm is imple-mented in an idl software package distributed by theauthors along with 17 grids of photoionization models.However, this technique relies on the external dust at-tenuation correction which must be applied to emissionlines fluxes prior to fitting. It also requires (similar tothe Dopita et al. (2016) approach) a pre-selection of starforming galaxies. We estimated the internal dust atten-uation using the typical value of the Balmer decrementH α /H β = 2 .
83 (Groves et al. 2012) and corrected allemission line fluxes accordingly. In galaxies where theobserved H α /H β ratio fell below that value, we assumedthe extinction to be zero. Finally, the fluxes were sup-plied to the IZI software package with the Dopita et al.(2013) model grid and the resulting [O/H] and ionizingparameter values for Gaussian emission line fluxes wereincluded in the gas phase metallicity table rcsed gasmet of the catalog. PHOTOMETRIC PROPERTIES OF THE SAMPLE
Completeness at different redshifts
Because our catalog uses the SDSS DR7 spectroscopicgalaxy sample as its master list, and the legacy SDSSspectroscopic survey was magnitude limited with the r = 17 .
77 mag limit in a 3 arcsec aperture, we sam-ple different parts of the galaxy luminosity function withthe redshift dependent completeness. Also, there is an important fiber collision effect, that is when two fibers inthe SDSS multi-object spectrograph cannot be put tooclose to each other: because of this, there is a systematicundersampling of dense clusters and groups of galaxies.In Fig. 6 (top panel) we present a two-dimensionaldistribution of our galaxies in the ( M z , g − r ) color–magnitude space. We identify the regions traditionallyreferred to as “the red sequence” and “the blue cloud”as well as the locus of typical post-starburst (E+A)galaxies. The density in the plot corresponds to theobject number density in our catalog at a given posi-tion of the parameter space. We also show by smallcrosses the tidally stripped systems, compact ellipticalgalaxies, from the sample of Chilingarian & Zolotukhin(2015) which reside systematically above the red se-quence. One can see the bimodality of the galaxy dis-tribution by color for intermediate luminosity and dwarfgalaxies ( M z > − . CSED – Reference Catalog of Galaxy SEDs Figure 5.
An example of the complex emission line profile of aSeyfert galaxy, and results of its fitting with two different tech-niques. Black stepped line in the upper panel shows the observedspectrum of H α and N ii lines in relative flux units, green dot-ted line is a Gaussian fit result, red solid line is a non-parametricfitting result. Individual H α and N ii profiles recovered by the non-parametric fitting are shown in orange and blue respectively. Lowerpanel shows fitting residuals. In the case of complex asymmetricemission lines profiles non-parametric fitting method is clearly pre-ferred over Gaussian one. end decline is due to the intrinsic shape of the luminos-ity function, while the low luminosity tail drops becauseof the SDSS completeness and target selection biasedagainst very extended (and therefore nearby) galaxies.We clearly see how the magnitude limit constraint ofSDSS causes the drop in the number of galaxies furtherand further up the luminosity function as we move tohigher redshifts. Fig. 6 confirms that we start probingthe dwarf galaxy regime ( M z > − . z < . Red sequence in different bands
For practical reasons such as selection of candidateearly-type members in galaxy clusters using photometricdata, it is important to know the shape of the red se-quence in different photometric bands. Here we providethe best fitting second degree polynomial approximationsof the red sequence shape for a set of galaxy colors span-ning optical and NIR bands.First we created a sample of red sequence galaxiesby the following criteria: (1) We selected all objectsat redshifts z < .
27; (2) we applied a color cut on
N U V − r colors by selecting all objects on ( M r , N U V − r )plane that resided above the straight line passing through p = ( − . , .
5) mag and p = ( − . , .
0) mag; (3)we applied a color cut on g − r colors by selecting allobjects on ( M r , g − r ) plane that resided above thestraight line passing through q = ( − . , .
5) mag and q = ( − . , .
75) mag and also satisfying the criterion( g − r ) < .
95 mag.Then, in order to account for two orders of mag-
Figure 6. (Top) Optical color–magnitude diagram for extinction-and k -corrected Petrosian magnitudes of all galaxies in our sample.(Bottom) Redshift distributions of galaxies in corresponding binson absolute magnitude. nitude variations of galaxy density along the red se-quence, for every combination of colors and magnitudes(e.g. g − r , M r ): (1) We selected measurements hav-ing statistical uncertainties < . AB magnitudesof galaxies in our sample stay mostly within the range − < M < −
15 mag in all optical red ( riz ) and NIR fil-ters, we added 20.0 mag to all absolute magnitudes priorto fitting.0
Chilingarian et al. ( u − r ) = +2 . − . · M r − . · M r ; σ = 0 . u − i ) = +2 . − . · M i − . · M i ; σ = 0 . u − z ) = +3 . − . · M z − . · M z ; σ = 0 . g − r ) = +0 . − . · M r − . · M r ; σ = 0 . g − i ) = +1 . − . · M i − . · M i ; σ = 0 . g − z ) = +1 . − . · M z − . · M z ; σ = 0 . g − Y ) = +1 . − . · M Y − . · M Y ; σ = 0 . g − J ) = +2 . − . · M J − . · M J ; σ = 0 . g − H ) = +2 . − . · M H − . · M H ; σ = 0 . g − K ) = +2 . − . · M K − . · M K ; σ = 0 . M = M col + 20 . σ in Eqs. 1) and stress that the actual fitting residuals formedian values are usually an order of magnitude smaller.We notice that in the most widely used parameterspaces, ( M r , g − r ), ( M r , u − r ), ( M i , g − i ), and ( M z , g − z ),the red sequence does not show any substantial curva-ture which is indicated by negligible 2nd order polyno-mial terms. This suggests that there is no “red sequencesaturation” at the bright end. Color transformations for galaxies of differentmorphologies and luminosities
Chilingarian & Zolotukhin (2012) demonstrated thatthe Hubble morphological classification derived by a hu-man eye (Fukugita et al. 2007) correlates very well withthe total
N U V − r color of a galaxy. With a computeddispersion of 0 . t , where t is the Hubble type, it corre-sponds to the subjective precision of such a classification.Here we use this relation in order to derive median val-ues of galaxy colors across the Hubble sequence for threegalaxy luminosity classes defined on a basis of their r -band luminosities.We dissect the ( M r , N U V − r ) color–magnitude planeinto 18 quadrangular regions by assuming that the mor-phological type for giant galaxies ( M r = −
24 mag)can be estimated by linearly varying the (
N U V − r )color from +0 . . Sd to E . Atthe same time, we assume that in the dwarf regime( M r = −
16 mag) the step reduces to 0.75 mag per Hub-ble type that corresponds to the observed reduction ofthe (
N U V − r ) color range. We choose 3 luminosity bins, − . ≤ M r < − . − . ≤ M r < − . − . ≤ M r < − . k -corrections. We extend their resultsat z = 0 (see table 3 in Fukugita et al. 1995) to near-UV and NIR colors and also towards intermediate and lowluminosity galaxies. The direct comparison of our valueswith those of Fukugita et al. (1995) reveals a good agree-ment of optical colors except (a) S0 galaxies which aresystematically redder in our case and stay really close tothe ellipticals; (b) the u − g color of ellipticals that is some0.25 mag bluer in our case. We assign the latter system-atics to our improved k -correction prescriptions for the u band photometry and generally higher quality of the u band SDSS photometric data compared to the datasetused in Fukugita et al. (1995). On the other hand, weattribute redder colors of lenticular galaxies in our datato the specificities of the synthetic color estimation tech-nique used in Fukugita et al. (1995) that underestimatedcolors of 2 of 4 their lenticular galaxies by 0.1–0.15 mag(see their table 1). SPECTROSCOPIC PROPERTIES OF THE SAMPLE
Stellar kinematics of galaxies
In comparison to original SDSS measurements of stel-lar kinematics based on cross-correlation with a limitedset of template spectra, our approach yields significantlysmaller template mismatch between models and observedspectra for non-active galaxies. We, therefore, achieveon average 30% lower statistical uncertainties of radialvelocity and velocity dispersion measurements. More-over, there is a known degeneracy between stellar metal-licity and velocity dispersion estimates when using pixelspace fitting techniques (Chilingarian et al. 2007b), be-cause an underestimated metallicity (i.e. using a metalpoor template for a metal rich galaxy) can be compen-sated by a lower velocity dispersion that would smearthat template spectrum to a lesser degree. Therefore, byusing a grid of stellar population models ranging from low([Fe/H]= − . . − , thus goingfar into the dwarf galaxy regime (Chilingarian 2009).As it was already pointed out by Fabricant et al.(2013), stellar velocity dispersion measurements in theSDSS DR7 catalog are systematically underestimated forluminous elliptical galaxies compared to the values ob-tained by the full spectrum fitting, that is likely causedby the template mismatch and degeneracy with metal-licity mentioned above. Here we observe a very simi-lar trend: our SSP velocity dispersion measurements formassive ellipticals ( σ ∼ >
250 km s − ) are up-to 30 km s − higher than those reported in the SDSS DR7 catalog, andthis difference goes down to 7–10 km s − for low lumi-nosity galaxies ( σ ∼
100 km s − ). In Fig. 7 (top panel)we present the comparison made for our entire samplefor 361,421 galaxies with velocity dispersion uncertain-ties better than 7% of the value (i.e. ∆ σ =7 km s − for σ = 100 km s − ). The velocity dispersions estimatedfrom the fitting of exponentially declining SFH modelscomputed with pegase.hr , are a little bit closer to thevalues in SDSS DR7, however, the general trend lookssimilar (Fig. 7, bottom panel). CSED – Reference Catalog of Galaxy SEDs Table 3
Median rest-frame colors of galaxies of different morphological types and luminosities in AB magnitudes. For every color (left column)there are three groups corresponding to giant (1st group), intermediate luminosity (2nd group), and dwarf (3rd group) galaxies of 6 valuesfor 6 Hubble types. Standard deviation values for each median color are presented in the adjacent table rows. − . ≤ M r < − . − . ≤ M r < − . − . ≤ M r < − . F UV - r NUV - r u - r g - r g - i g - z g - Y g - J g - H g - K σ SDSS , km/s4020020406080 EXP-SFH σ R C S E D - σ S D SS , k m / s Figure 7.
Comparison of RCSED stellar velocity dispersion mea-surements with those published by the SDSS DR7. Top and bottompanel correspond to the two sets of stellar population models, SSPand exponentially declining SFHs correspondingly.
In Fig. 8 we present the relation between galaxy lu-minosities and velocity dispersions or the Faber–Jackson(1976) relation constructed for 52,506 elliptical galaxieswhich were morphologically selected by the Galaxy Zoo(Lintott et al. 2011) citizen science project and had sta-tistical uncertainties of their velocity dispersion measure-ments better than 10% of the value. We have correctedvelocity dispersion measurements to their global valuesaccording to Cappellari et al. (2006) using half-light radiifrom Simard et al. (2011) included in our catalog. Weused the criterion formulated in Saulder et al. (2013): in order to be included in our early type galaxy sample, anobject has to be classified by at least 10 Galaxy Zoo usersof whom at least 70% classify it as an elliptical galaxy.Six panels present measurements in six different redshiftintervals shown as dots while the contours display the cu-mulative distribution at all redshifts. The lowest redshiftpanel contains the measurements for a sample of galaxiesin the Abell 496 cluster ( z = 0 . L g ∝ σ . at z = 0 presented in Bernardiet al. (2003). We see that the slope changes to L g ∝ σ . at fainter luminosities M g > − . σ <
100 km s − ) at lowredshifts (0 . < z < .
06) that includes hundreds ofmassive galaxy clusters and groups.RCSED velocity dispersion measurements were usedprior to publication by Saulder et al. (2013) for the cal-ibration of the Fundamental Plane (Djorgovski & Davis1987). We refer to that work for an intensive discussionregarding the FP of elliptical galaxies observed by theSDSS.2
Chilingarian et al. -16. -18 -20 -22. -24.50100200500 σ V ( k m / s ) . Faber–Jackson relation for 52,506 morphologically classified elliptical galaxies in our sample (Galaxy Zoo classification). Inorder to remove outliers, a 10% cut has been applied on the relative errors on g magnitude and on the velocity dispersion and a goodadjustment has been required χ < . 8. The contours correspond to the whole sample (smoothed with a 3 × Z/H ] displayed in Figure 10.The dashed red line corresponds to the maximum likelihood estimate of the slope L g ∝ σ . at z = 0 computed by Bernardi et al. (2003).The blue points in the low redshift sub-sample (in the bottom left panel) correspond to the dwarf galaxies sample of Chilingarian et al.(2008). Stellar populations from absorption line analysis In our catalog we include stellar population parametersobtained by the fitting of galaxy spectra using two stellarpopulation model grids computed with the pegase.hr evolutionary synthesis code: (i) SSP models based onthe intermediate resolution MILES stellar library char-acterized by ages ( t ) and metallicities ([Fe/H]); and (ii)models with exponentially declining SFHs based on thehigh resolution ELODIE-3.1 stellar library characterizedby exponential timescales ( τ ) and metallicities ([Fe/H]).In Fig 9 we present distributions of galaxies in the twoparameter spaces.One can clearly see a spotty structure in the SSP bestfitting results and the lack of such a structure for theexponentially decaying models. We also performed simi-lar tests for original MILES stellar population models byVazdekis et al. (2010) and models by Bruzual & Charlot(2003) for a sub-sample of SDSS DR7 spectra. We willprovide a complete description and detailed discussion inthe forthcoming paper (Katkov et al. in prep.), here we present a brief summary and conclusions of our study.The observed spotty structure represents artifactscaused by the improper implementation of the interpola-tion algorithm in the stellar population code most likelyon the stellar library interpolation step propagating intostellar population models and not by the nbursts pop-ulation fitting procedure. The nbursts code uses a non-linear minimization technique that requires the secondpartial derivatives on all parameters to be continuous.Discontinuities will cause the solution to be either at-tracted to some region of the parameter space, or pushedaway from it.Our conclusion is supported by the following obser-vations: (i) the morphology of the spotty structure re-mains very similar when using two different sets of SSPmodels computed with the same pegase.hr code butwith different stellar libraries, MILES and ELODIE; (ii)switching to original MILES models (Vazdekis et al.2010) where the interpolation procedure is much sim-pler than in pegase.hr (linear interpolation between 5 CSED – Reference Catalog of Galaxy SEDs τ ∼ > . τ < . pegase.hr MILES based SSP models usingbasic splines ( b -splines) on age removes most of the pat-tern.We also notice, that extending the working wavelengthrange to shorter wavelengths ( < λ > τ . Despitethe artifact structure in ages that extends into horizontalstripes on this plot, there is a 1-to-1 correspondence be-tween t and τ in a wide range of ages. Short timescales τ correspond to old stellar populations while τ = 20 Gyris equivalent to t ≈ . ∼ − . < [Fe/H] < − . ∼ pegase.hr SSP mod-els in the wavelength range 3910–6790 ˚A and comparedthe metallicity and velocity dispersion measurements tothose obtained from the fitting of MILES– pegase mod-els against the same spectra. In Fig. 11 we present the re-lation between the differences of velocity dispersions andSSP metallicities obtained using the two sets of mod-els at different signal-to-noise ratios. We can clearlysee the degeneracy manifested by the elongated shapeof the cloud that decreases with the increasing signal-to-noise ratio up-to the signal-to-noise of 30. Above 30 theimprovement becomes insignificant. This result suggestthat published velocity dispersion values obtained withthe full spectral fitting of intermediate resolution spectra R = 1500 − [ Z / H ] , d e x SSP . % . % . % . % . % . % . % τ , Gyr1.51.00.50.00.5 [ Z / H ] , d e x exp-SFH . % . % . % Figure 9. Distributions of galaxies in the age–metallicity spacefrom the fitting of SSP (top panel) and exponentially decaying SFH(bottom panel) models. Emission line properties Comparison of line fluxes with the MPA–JHU andOSSY catalogs and between the two techniques We compare a subset of our catalog containing mea-surements of emission line fluxes obtained from the para-metric Gaussian fitting to the results from the MPA–JHUcatalog distributed by the SDSS project (Brinchmannet al. 2004; Tremonti et al. 2004) and with the OSSYcatalog (Oh et al. 2011). We used OSSY emission linemeasurements prior to the internal extinction correction.Given that the fluxes were computed using very similarmethodologies with the main different corresponding tothe subtraction of the underlying stellar population andthe Galactic extinction correction techniques used, weexpect a very good agreement for well detected emissionlines. We directly compare fluxes of the [O ii ] (3727 ˚A),[O iii ] (5007 ˚A), H α , and [N ii ] (6584 ˚A) emission lines fora sample of galaxies where they were detected at a levelexceeding 10 σ (i.e. Flux/ σ (Flux) > α flux distribution. At the faint end(10 σ detection), the systematic difference stays within2% while the standard deviation grows to 3%. Hence,4 Chilingarian et al. τ , Gyr124691318 A g e , G y r . % . % . % . % . % . % SS P [ Z / H ] , d e x . % . % Figure 10. Comparison of SSP ages to timescales τ for exponen-tially decaying models (top) and metallicity measurements (bot-tom). we conclude that our emission line fitting code works asexpected and does not introduce any substantial system-atic errors to flux measurements.Compared to H α , the H β line is much more sensitiveto the age of the stellar population being subtracted. InAppendix A we discuss the systematic errors of the H β measurements as a function of the age mismatch. Incase of faint emission lines, the systematics dominatesthe measurements if the age was determined incorrectly,and makes them useless for the emission line diagnostics.The principal difference of our results with those pub-lished earlier is the non-parametric approach to the emis-sion line fitting. For galaxies which exhibit some signsof nuclear activity, the Balmer lines fluxes derived non-parametrically significantly exceed the values obtainedwith the Gaussian fitting. In Fig. 13 we present the H α flux ratio between the two approaches. The inset con-tains the same Seyfert galaxy which we presented earlierin Fig. 5 and the arrow indicates its position in the di-agram that suggests that its non-parametric H α flux es-timate is about 20% higher than that obtained with theGaussian profile fitting. BPT diagrams and gas phase metallicities In Fig. 14 we present three flavors of the BPT dia-gram which use different combinations of emission lines σ km/s ( PEGASE.HR / MILES )0.20.00.20.4 [ Z / H ] d e x ( P E G A S E . H R − M I L E S ) SNR Figure 11. Degeneracy between metallicity and velocity disper-sion estimates shown as the ratio between velocity dispersions vsSSP metallicities obtained from the fitting of some 420,000 SDSSspectra using pegase.hr and MILES–PEGASE SSP models. Thecontours display the 1 σ values which correspond to the areas con-taining 68% galaxies with the spectra having signal-to-noise ratioswithin 20% of a displayed value. The number of galaxies for eachcontour ranges from ∼ ∼ computed using the non-parametric fitting. The pointsare color-coded corresponding to the H α emission lineEW. Cid Fernandes et al. (2010) proposed to use theH α EW to discriminate between Seyfert and LINER ac-tivity (instead of the traditionally used [O iii ]/H β ratio),because H β is often too weak to be detected and mea-sured. We clearly see the bimodal distribution of non-starforming galaxies in the bottom two panels whichcorresponds to Seyfert galaxies (cloud to the top) andLINER/shockwave/post-AGB ionization (cloud to thebottom). The top panel displays the original BPT re-lation. The region between the red solid and the bluedashed lines defines “transitional” galaxies (Kewley et al.2006) which we included in the calculation of metallic-ities in addition to the star forming galaxies located tothe bottom left of the blue dashed line.As we described above, RCSED includes gas phasemetallicity measurements calculated with the Bayesianmethod implemented in the IZI software package withthe Dopita et al. (2013) model grid, which uses all avail-able emission lines in a spectrum; and a recent techniqueby Dopita et al. (2016) that relies on the [N/O] calibra-tion and uses only 5 emission lines around H α .Kewley & Ellison (2008) demonstrated that differentemission line calibrations yield largely inconsistent gasphase metallicity estimates when applied to the sameinput dataset with the differences reaching 0.7 dex (5times). There is currently no consensus in the astronom-ical community about which calibrations produce morereliable metallicity estimates with arguments for bothdirect (Andrews & Martini 2013) and indirect (L´opez-S´anchez et al. 2012) methods. Gas and stellar metallic-ities also seem to strongly disagree (Yates et al. 2012).We notice, however, that all emission line calibrationsresult in the gas phase [O/H] mass–metallicity relations CSED – Reference Catalog of Galaxy SEDs M P A - J H U / R C S E D [OII] λλ [OIII] λ H α [NII] λ 10 20 30 50 80SNR in line0.920.940.960.981.001.021.041.06 O SS Y / R C S E D 10 20 30 50 80SNR in line 50 70 100 140SNR in line 30 50 70 100SNR in line Figure 12. Comparison of RCSED Gaussian emission line fluxes for 4 emission lines with the MPA–JHU (top row of plots) and OSSY(bottom row of plots) catalogs. The median and standard deviation of the distribution are shown by brown symbols with the error bars. χ G / χ NP F ( H α ) N P / F ( H α ) G α mjd=51811 plate=422 fiberid=4557700 7720 7740 7760 7780Wavelength, 505 Figure 13. Comparison of the H α fluxes obtained for the para-metric (Gaussian) and non-parametric emission line profile fittingas a function of the χ ratio. An example profile decomposition isshown in the inset for an object with highly discrepant flux esti-mates. spanning much lower range of metallicities than that ofstellar metallicities for a given galaxy stellar mass range.All gas metallicity relations saturate at high metallic-ities and the saturation occurs at different values (seee.g. fig. 10 in Andrews & Martini 2013). The highestrange of metallicities is covered by the calibration usedby Tremonti et al. (2004) and provided in the MPA–JHUcatalog.In Fig. 15 we present the comparison of the Dopitaet al. (2016) calibration with the MPA–JHU metallicities(blue shaded area) for 231,107 galaxies with the signal-to-noise ratios of H α , [N ii ], [O ii ], and [O iii ] lines exceeding10. The agreement is very good at 12+[O/H] < . 08 dex.At higher abundances Dopita et al. (2016) metallicitiesbecome slightly higher than those from the MPA–JHUdataset.We ran the IZI metallicity determination code fora small sub-sample of 20,000 randomly selected star- forming galaxies with high signal-to-noise emission lines(S/N > 10) using all available grids of models and com-pared the derived metallicities with those obtained withthe Dopita et al. (2016) calibration for the same galaxysample. The only model grid that demonstrated a satis-factory agreement was that from Dopita et al. (2013). Asexpected, it also provides a satisfactory agreement withthe MPA–JHU catalog (see Fig. 15, orange shaded areas)with the standard deviation of the difference 0 . 10 dex.In Fig. 16 we show the luminosity–metallicity relation(left panel) and the comparison of gas phase and SSPstellar metallicities (right panel) for the IZI–based deter-mination using the Dopita et al. (2013) models (orangecontours) and the Dopita et al. (2016) calibration (bluepoints). The mass–metallicity relation is well defined andwe clearly see that the Dopita et al. (2013) model gridused in IZI yields a flatter shape than the more recentcalibration (Dopita et al. 2016).The comparison of gas phase and stellar metallicitiesreveal a substantial offset ranging from about 0.3 dexat solar stellar metallicities to 0.8 dex at the low end([Fe/H] star = − . M/L ratios and the star formation rate while the gas phasemetallicity reflects the current chemical abundance pat-tern in the ISM enriched with metals, therefore, we ex-pect to see the offset in metallicities. On the other hand,for the constant metal production rate per solar mass, thedifference at low metallicities will be higher because the6 Chilingarian et al. [ NII ] /H α (6585 / [ O III ] / H β ( / ) H α e q u i v a l e n t w i d t h () [ SII ] /H α (6718 +6733 / [ O III ] / H β ( / ) H α e q u i v a l e n t w i d t h () [ OI ] /H α (6302 / [ O III ] / H β ( / ) H α e q u i v a l e n t w i d t h () Figure 14. Three flavors of a BPT diagram with the color coded H α equivalent width. In each panel we display only those galaxieswhere all emission lines used in the corresponding plot have S/N > 3. The contours correspond to the galaxy density smoothed witha moving average based on a 4 × + l o g O / H , d e x ( R C S E D ) Dopita et al. (2016)IZI Figure 15. Comparison of gas phase metallicities published in theMPA–JHU catalog (horizontal axis) to our measurements (verticalaxis). The results of the IZI Bayesian technique are shown in brownand the measurements obtained with the Dopita et al. (2016) cal-ibration are shown in blue. metallicity scale is logarithmic, therefore stellar metal-licities should span a larger range of value compared togas phase metallicities and the mass–metallicity relationslopes for gas will be shallower than that for stars. CATALOG ACCESS: WEB-SITE AND VIRTUALOBSERVATORY ACCESS INTERFACES Efficient, convenient, and intuitive data access mecha-nisms and interfaces are essential for a complex projectlike RCSED. Therefore, we decided to build access inter-faces for both interactive and batch access to the data.RCSED includes several different data types (e.g. spec-tra and tabular data) and our access infrastructure (seeFig. 17) is organized to simplify their usage through dif-ferent interfaces. The most natural way to access the cat-alog is by using the web application at http://rcsed.sai.msu.ru/ . It provides a single-field google -stylesearch interface where one can query the catalog by anobject identifier, coordinates or object properties, e.g. select all galaxies with redshifts z < . having red col-ors g − r > . 5. Every object in the sample has its ownweb page with the summary of all its properties, SED,spectral data available in the catalog, and image cutoutsdisplaying the object at different wavelength provided byGALEX, SDSS, and UKIDSS surveys. An example of aspectrum summary plot presented on such web pages forevery object is given in Fig. 4.We developed an Application Programming Interface(API) to UKIDSS data, which allow us to extract im-age cutouts around an arbitrary position with a givenbox size in every filter. From cutout images in the JHK bands we generate a color composite image and displayit in the object web-page. The API implemented in python is available for download from the project web-site. Another service we present is an interactive spec-trum plotter implemented in JavaScript, our alternative CSED – Reference Catalog of Galaxy SEDs M r , mag7.88.18.48.79.09.39.6 + l o g O / H , d e x Dopita et al. (2016)IZI 1.5 1.2 0.9 0.6 0.3 0.0 Z SSP , dex Figure 16. Relation between a gas phase metallicity and a galaxy luminosity (left) and an SSP stellar metallicity (right). Blue dots andorange density contours display Dopita et al. (2016) and IZI metallicities correspondingly. The median and standard deviation of Dopitaet al. (2016) measurements are shown by brown symbols with the error bars. to the SDSS spectrum plotter. It contains a number ofvalue-added features, such as the display of best-fittingtemplates and identification of emission lines.In addition to the custom web application, our datadistribution infrastructure has the open source GAVODaCHS data center suite in its core (see Fig. 17) whichprovides a set of VO data access mechanisms.The data for SDSS spectra and their best-fitting SSPmodels are provided as FITS files that can be fetchedby direct unique URLs. One can find a URL for everyparticular object spectrum file either in the object’s webpage or by querying the provided IVOA Simple SpectralAccess Protocol (SSAP) web service using object coor-dinates. The SSAP web service answers essentially witha list of spectra URLs and it is convenient to access pro-grammatically or by using VO compatible client applica-tions such as TOPCAT (Taylor 2005), SPLAT-VO orVO-Spec which can directly load spectral data for fur-ther analysis by analyzing the SSAP web service queryresult.For the ultimate flexibility of querying tabular cata-log data, we provide a Table Access Protocol (TAP) webservice. IVOA TAP is an access interface, which allowsa user to query the entire relational database schema(see Fig. 18) using a powerful SQL-like language. It canbe considered as an open source equivalent of the SDSSCasJobs service. Again, TAP web service can be used forscript-based access as well as by using desktop VO appli-cations. In particular, TOPCAT has a very useful TAPquery dialogue with built-in help, query examples, syntaxhighlighting and given database schema assistance tools.We encourage our users to access the RCSED TAP webservice through TOPCAT. We also note that our TAPservice has a table upload capability, so that the usercan upload his/her own tables and use it in subsequentSQL queries i.e. in join clauses, that is convenient for http://soft.g-vo.org/dachs cross-identification of user provided object samples withthe RCSED objects without the need of downloading ourfull catalog.Below we give several query examples that are helpfulto start using the RCSED database. More query exam-ples and science case tutorials are provided on the projectwebsite http://rcsed.sai.msu.ru . Our TAP web ser-vice can be used for joining tables from the databaseschema specphot presented in Fig. 18, so that it is easyto retrieve a single table with the GalaxyZoo morphol-ogy, the photometric bulge+disk decomposition and theRCSED basic parameters combined for any galaxy of in-terest. An example of such a query to select all thosedata for a particular object would be: SELECTr.*, g.*, s2.*FROMspecphot.rcsed AS rJOIN specphot.galaxyzoo AS gON r.objid = g.objidJOIN specphot.simard_table2 AS s2ON r.objid = s2.objidWHEREr.objid = 587731891649052703 Note that specphot prefix for table names correspondsto the name of the database schema where RCSED tablesare stored. A query to retreive all data from the RCSEDon galaxies, classified as ellipticals in GalaxyZoo is: SELECTr.*FROMspecphot.rcsed AS rJOIN specphot.galaxyzoo AS gON r.objid = g.objidWHEREg.elliptical = 1 A query to select the data for a BPT (Baldwin et al.1981) diagram for 10,000 galaxies with S / N > 10 in thecorresponding line fluxes obtained with the Gaussian fit-ting looks like this:8 Chilingarian et al. Figure 17. Block diagram of the catalog data access infrastruc-ture. Data are stored in the relational database (catalog tables andspectra metadata) and on the disk (FITS files with spectra andcontinuum models). They are accessed by applications (a customweb application and the GAVO DaCHS suite) which in turn ex-pose several public access interfaces suitable for convenient queriesand data retrieval by the multitude of user client programs, bothVO-compatible and generic. SELECTTOP 10000f6550_nii_flx / f6565_h_alpha_flx AS BPT_x,f5008_oiii_flx / f4863_h_beta_flx AS BPT_yFROMspecphot.rcsed_lines_gaussWHEREf6565_h_alpha_flx / f6565_h_alpha_flx_err > 10AND f5008_oiii_flx / f5008_oiii_flx_err > 10AND f4863_h_beta_flx / f4863_h_beta_flx_err > 10AND f6550_nii_flx / f6550_nii_flx_err > 10 Finally, all the catalog tables (see Fig. 18) are availablefor download as FITS tables from the project’s websitefor the offline use. SUMMARY We presented a reference catalog of homogeneousmulti-wavelength spectrophotometric information forsome 800,000 low to intermediate redshift galaxies(0 . < z < . 6) from the SDSS DR7 spectroscopicgalaxy sample with value-added data. For every galaxywe provide: • a k -corrected and Galactic extinction corrected far-UV to NIR broad band SED for integrated fluxes compiled from the SDSS (optical), GALEX (UV),and UKIDSS (NIR) surveys • a k -corrected and Galactic extinction corrected far-UV to NIR broad band SED for fluxes in circular 3-arcsec apertures that correspond to SDSS spectralapertures • results of the full spectrum fitting of an SDSS spec-trum using the nbursts technique that includes:(a) an original SDSS spectrum; (b) the best-fittingsimple stellar population template in the wave-length range 3700 < λ < < λ < • results of the emission line analysis using paramet-ric (Gaussian) and non-parametric line profiles thatinclude: (a) emission line fluxes corrected for theGalactic extinction; (b) estimates of the reddeninginside a galaxy for star formation dominated sys-tems derived from the observed Balmer decrement;(c) radial velocity offsets with respect to stars; (d)intrinsic emission line widths for the parametric fit-ting • cross-match of a galaxy with third-party catalogsproviding its structural parameters from the two-dimensional light profile fitting (Simard et al. 2011)and galaxy morphology by the Galaxy Zoo project(Lintott et al. 2008)The catalog is fully integrated into the internationalVirtual Observatory infrastructure and available via aweb application and as a Virtual Observatory resourceproviding IVOA TAP and IVOA SSAP interfaces in or-der to programmatically access tabular data and spectracorrespondingly.In addition to that, we presented best-fitting polyno-mial approximations for the red sequence shape in color–magnitude diagrams that include different colors, andmean colors for galaxies of 6 morphological types, fromelliptical to late type spirals and irregulars and 3 lumi-nosity classes (giants, intermediate luminosity, dwarfs).Our catalog has already been used in several researchprojects that can be categorized into two groups: (i) sta-tistical studies of galaxy properties; (ii) search and dis-covery of rare galaxies.The first interesting result obtained with RCSED wasthe discovery of a universal 3-dimensional relation of NUV and optical galaxy colors and luminosities (Chilin-garian & Zolotukhin 2012). It also demonstrated thatthe integrated NUV − r color is a good proxy for a mor-phological type. The spectum fitting results for ellipti-cal galaxies were later used in the re-calibration of thefundamental plane in SDSS (Saulder et al. 2013) whichallowed us to compute redshift independent distances toearly-type galaxies. Finally, we performed a calibrationof near-infrared stellar M/L ratios using optical colorsand computed stellar masses for a new catalog of groups CSED – Reference Catalog of Galaxy SEDs Figure 18. The entity–relationship diagram for the tables in the catalog database. Blue color denotes original tables computed inRCSED, green color is for external datasets added to the database for convenience. Main table is rcsed which has a 1-to-1 relation tothe rcsed fibermags table by the primary key column objid ,a 1-to-many (optional) relation to the rcsed gasmet table with gas phasemetallicity measurements, 1-to-many (optional) relations to rcsed lines gauss , rcsed lines nonpar tables with parametric (gauss) andnon-parametric emission lines measurements. galaxyzoo , simard table2 and simard table3 datasets are all linked to the main table by the objid column and provide morphological classification and structural properties of galaxies from our sample using the data form Lintottet al. (2011) and Simard et al. (2011) respectively. All these tables are stored in specphot database schema, so the properly qualified tablename is, for example, specphot.rcsed . We also note that all column names in the database are lowercase for homogeneity. Same schemaand relationships applies to the distribution of RCSED in the form of FITS tables available for download from project’s website. and clusters combining SDSS and 2MASS redshift surveydata (Saulder et al. 2016). Our non-parametric emissionline fitting results will be used to perform massive deter-minations of virial black hole masses in AGNs (Katkovet al. in prep). Other potential applications of statisticalstudies based on RCSED include but not limited to: en-vironmental dependence of galaxy scaling relations andstellar population properties; connecting AGNs to stellarpopulations in galaxy centers; comparing different starformation rate indicators (e.g. emission line fluxes, UVand MIR photometry).Thanks to the unique combination of photometric andspectral data as well as physical properties of galaxiesderived from them, RCSED becomes an efficient searchtool for rare or unique galaxies. Our dataset was usedto discover and characterize massive compact galaxies atintermediate redshifts 0 . < z < . z > . 5) and measure their volume density (Dam-janov et al. 2014). Then, using their fundamental planepositions, intermediate redshift compact galaxies wereshown to be an extension of normal ellipticals to the com-pact regime (Zahid et al. 2015). Finally, it was demon-strated that some massive compact early-type galaxiesactually stopped forming stars very recently (Zahid et al.2016). We also used the universal UV–optical color–color–magnitude relation to define complex selection cri-teria and discover 195 previously considered extremelyrare compact elliptical galaxies (Chilingarian & Zolo-tukhin 2015). One can identify other obvious extragalac-tic rarities easily searchable with RCSED: post-starburstgalaxies, candidate double-peaked AGNs, dwarf AGNhosts, “normal” galaxies with peculiarities detectable inmulti-wavelength data such as ellipticals with NUV ex-cess.0 Chilingarian et al. In the future, we anticipate to release intermediateand high redshift extensions of our catalog that willinclude the analysis of publicly available spectra fromthe Smithsonian Astrophysical Observatory Hectospecarchive collected with the Hectospec multi-fiber spec-trograph (Fabricant et al. 2005) the 6.5-m MMT andthe DEEP2 galaxy redshift survey (Newman et al. 2013)made with the DEIMOS spectrograph at the 10-m Kecktelescope. We also plan to expand the wavelength cover-age by adding the all-sky infrared data from the Wide-field Infrared Survey Explorer (WISE) satellite (Wrightet al. 2010). A major update to our catalog will be madewith the full spectrophotometric fitting the entire sam-ple using the nbursts+phot algorithm (Chilingarian &Katkov 2012) and resolving star formation histories forabout 10 galaxies with high quality UV and NIR data. ACKNOWLEDGMENTS We acknowledge our anonymous referee whose com-ments helped us to improve this manuscript. IC’s re-seach is supported by the Smithsonian Astrophysical Ob-servatory Telescope Data Center. IZ acknowledges thesupport by the Russian Scientific Foundation grant 14-50-00043 for the catalog assembly tasks and grant 14-12-00146 for the data publication and deployment sys-tem. The authors acknowledge partial support fromthe M.V.Lomonosov Moscow State University Programof Development, and a Russian–French PICS Interna-tional Laboratory program (no. 6590) co-funded by theRFBR (project 15-52-15050), entitled “Galaxy evolu-tion mechanisms in the Local Universe and at interme-diate redshifts”. The statistical studies of galaxy pop-ulations by IC, IZ, IK, and ER are supported by theRFBR grant 15-32-21062 and the presidential grant MD-7355.2015.2. The authors are grateful to citizen sci-entists M. Chernyshov, A. Kilchik, A. Sergeev, R. Ti-hanovich, and A. Timirgazin for their valuable help withthe development of the project website. In 2009–2011the project was supported by the VO-Paris Data Centreand by the Action Specifique de l’Observatoire Virtuel(VO-France). A substantial progress in our project wasachieved during our 2013, 2014, and 2015 annual Cha-monix workshops and we are grateful to our host O. Be-van at Chˆalet des Sapins. This research has made use ofTOPCAT, developed by Mark Taylor at the Universityof Bristol; Aladin developed by the Centre de Donn´eesAstronomiques de Strasbourg (CDS); the “exploresdss”script by G. Mamon (IAP); the VizieR catalogue accesstool (CDS). Funding for the SDSS and SDSS -II has beenprovided by the Alfred P. Sloan Foundation, the Par-ticipating Institutions, the National Science Foundation,the U.S. Department of Energy, the National Aeronau-tics and Space Administration, the Japanese Monbuka-gakusho, the Max Planck Society, and the Higher Ed-ucation Funding Council for England. The SDSS WebSite is . GALEX (Galaxy Evo-lution Explorer) is a NASA Small Explorer, launched inApril 2003. We gratefully acknowledge NASA’s supportfor construction, operation, and science analysis for the GALEX mission, developed in cooperation with the Cen-tre National d’Etudes Spatiales of France and the Korean http://oirsa.cfa.harvard.edu/ Ministry of Science and Technology.REFERENCES Abazajian, K. N., Adelman-McCarthy, J. K., Ag¨ueros, M. A.,et al. 2009, ApJS, 182, 543Andrews, B. H., & Martini, P. 2013, ApJ, 765, 140Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5Becker, R. H., White, R. L., & Helfand, D. J. 1995, ApJ, 450, 559Bernardi, M., Sheth, R. K., Annis, J., et al. 2003, AJ, 125, 1849Blanc, G. A., Kewley, L., Vogt, F. P. A., & Dopita, M. A. 2015,ApJ, 798, 99Blanton, M. R., Hogg, D. W., Bahcall, N. A., et al. 2003, ApJ,592, 819Blanton, M. R., Schlegel, D. J., Strauss, M. A., et al. 2005, AJ,129, 2562Brinchmann, J., Charlot, S., White, S. D. M., et al. 2004,MNRAS, 351, 1151Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Cappellari, M., & Emsellem, E. 2004, PASP, 116, 138Cappellari, M., Bacon, R., Bureau, M., et al. 2006, MNRAS, 366,1126Chilingarian, I., Prugniel, P., Sil’chenko, O., & Koleva, M. 2007a,in IAU Symposium, Vol. 241, Stellar Populations as BuildingBlocks of Galaxies, ed. A. Vazdekis & R. R. Peletier(Cambridge, UK: Cambridge University Press), 175–176,arXiv:0709.3047Chilingarian, I., & Zolotukhin, I. 2015, Science, 348, 418Chilingarian, I. V. 2009, MNRAS, 394, 1229Chilingarian, I. V., Cayatte, V., Durret, F., et al. 2008, A&A,486, 85Chilingarian, I. V., & Katkov, I. Y. 2012, in IAU Symposium,Vol. 284, The Spectral Energy Distribution of Galaxies - SED2011, ed. R. J. Tuffs & C. C. Popescu, 26–28Chilingarian, I. V., Melchior, A.-L., & Zolotukhin, I. Y. 2010,MNRAS, 405, 1409Chilingarian, I. V., Prugniel, P., Sil’chenko, O. K., & Afanasiev,V. L. 2007b, MNRAS, 376, 1033Chilingarian, I. V., & Zolotukhin, I. Y. 2012, MNRAS, 419, 1727Cid Fernandes, R., Stasi´nska, G., Schlickmann, M. S., et al. 2010,MNRAS, 403, 1036Colless, M., Dalton, G., Maddox, S., et al. 2001, MNRAS, 328,1039Damjanov, I., Chilingarian, I., Hwang, H. S., & Geller, M. J.2013, ApJ, 775, L48Damjanov, I., Hwang, H. S., Geller, M. J., & Chilingarian, I.2014, ApJ, 793, 39de Vaucouleurs, G., de Vaucouleurs, A., Corwin, Jr., H. G., et al.1991, Third Reference Catalogue of Bright Galaxies. Volume I:Explanations and references. Volume II: Data for galaxiesbetween 0 h and 12 h . Volume III: Data for galaxies between 12 h and 24 h .Djorgovski, S., & Davis, M. 1987, ApJ, 313, 59Dopita, M. A., Kewley, L. J., Sutherland, R. S., & Nicholls, D. C.2016, Ap&SS, 361, 61Dopita, M. A., Sutherland, R. S., Nicholls, D. C., Kewley, L. J.,& Vogt, F. P. A. 2013, ApJS, 208, 10Faber, S. M., & Jackson, R. E. 1976, ApJ, 204, 668Fabricant, D., Chilingarian, I., Hwang, H. S., et al. 2013, PASP,125, 1362Fabricant, D., Fata, R., Roll, J., et al. 2005, PASP, 117, 1411Fukugita, M., Shimasaku, K., & Ichikawa, T. 1995, PASP, 107,945Fukugita, M., Nakamura, O., Okamura, S., et al. 2007, AJ, 134,579Gallazzi, A., Charlot, S., Brinchmann, J., & White, S. D. M.2006, MNRAS, 370, 1106Groves, B., Brinchmann, J., & Walcher, C. J. 2012, MNRAS, 419,1402Hewett, P. C., Warren, S. J., Leggett, S. K., & Hodgkin, S. T.2006, MNRAS, 367, 454Hinshaw, G., Larson, D., Komatsu, E., et al. 2013, ApJS, 208, 19Kauffmann, G., Heckman, T. M., White, S. D. M., et al. 2003,MNRAS, 341, 33Kewley, L. J., & Ellison, S. L. 2008, ApJ, 681, 1183Kewley, L. J., Groves, B., Kauffmann, G., & Heckman, T. 2006,MNRAS, 372, 961 CSED – Reference Catalog of Galaxy SEDs Lawrence, A., Warren, S. J., Almaini, O., et al. 2007, MNRAS,379, 1599Lawson, C. L., & Hanson, R. J. 1995, Classics in AppliedMathematics, Vol. 15, Solving Least Squares Problems(Philadelphia, PA: Society for Industrial and AppliedMathematics (SIAM))Le Borgne, D., Rocca-Volmerange, B., Prugniel, P., et al. 2004,A&A, 425, 881Lintott, C., Schawinski, K., Bamford, S., et al. 2011, MNRAS,410, 166Lintott, C. J., Schawinski, K., Slosar, A., et al. 2008, MNRAS,389, 1179L´opez-S´anchez, ´A. R., Dopita, M. A., Kewley, L. J., et al. 2012,MNRAS, 426, 2630Markwardt, C. B. 2009, in Astronomical Society of the PacificConference Series, Vol. 411, Astronomical Data AnalysisSoftware and Systems XVIII, ed. D. A. Bohlender, D. Durand,& P. Dowler, 251Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJ,619, L1Matkovi´c, A., & Guzm´an, R. 2005, MNRAS, 362, 289Matteucci, F. 1994, A&A, 288, 57Newman, J. A., Cooper, M. C., Davis, M., et al. 2013, ApJS, 208,5Oh, K., Sarzi, M., Schawinski, K., & Yi, S. K. 2011, ApJS, 195, 13Oh, K., Yi, S. K., Schawinski, K., et al. 2015, ApJS, 219, 1Petrosian, V. 1976, ApJ, 209, L1Prugniel, P., & Soubiran, C. 2004, ArXiv Astrophysics e-prints,astro-ph/0409214Prugniel, P., Soubiran, C., Koleva, M., & Le Borgne, D. 2007,ArXiv Astrophysics e-prints, astro-ph/0703658S´anchez-Bl´azquez, P., Peletier, R. F., Jim´enez-Vicente, J., et al.2006, MNRAS, 371, 703Saulder, C., Mieske, S., Zeilinger, W. W., & Chilingarian, I. 2013,A&A, 557, A21 Saulder, C., van Kampen, E., Chilingarian, I., Mieske, S., &Zeilinger, W. W. 2016, A&A, 596, A14Saunders, W., Sutherland, W. J., Maddox, S. J., et al. 2000,MNRAS, 317, 55Schechter, P. 1976, ApJ, 203, 297Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500,525Serra, P., & Trager, S. C. 2007, MNRAS, 374, 769Simard, L., Mendel, J. T., Patton, D. R., Ellison, S. L., &McConnachie, A. W. 2011, ApJS, 196, 11Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131,1163Taylor, M. B. 2005, in Astronomical Society of the PacificConference Series, Vol. 347, Astronomical Data AnalysisSoftware and Systems XIV, ed. P. Shopbell, M. Britton, &R. Ebert, 29–+Taylor, M. B. 2006, in Astronomical Society of the PacificConference Series, Vol. 351, Astronomical Data AnalysisSoftware and Systems XV, ed. C. Gabriel, C. Arviset, D. Ponz,& S. Enrique, 666–+Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004,ApJ, 613, 898Vazdekis, A., S´anchez-Bl´azquez, P., Falc´on-Barroso, J., et al.2010, MNRAS, 404, 1639Worthey, G. 1994, ApJS, 95, 107Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010,AJ, 140, 1868Yasuda, N., Fukugita, M., Narayanan, V. K., et al. 2001, AJ, 122,1104Yates, R. M., Kauffmann, G., & Guo, Q. 2012, MNRAS, 422, 215Zahid, H. J., Baeza Hochmuth, N., Geller, M. J., et al. 2016, ApJ,831, 146Zahid, H. J., Damjanov, I., Geller, M. J., & Chilingarian, I. 2015,ApJ, 806, 122APPENDIX SYSTEMATICS IN EMISSION LINE MEASUREMENTS DUE TO STELLAR POPULATION TEMPLATE MISMATCH Absorption lines of the hydrogen Balmer series contain important information about stellar population ages (Worthey1994), they become weaker when stars get older. At the same time, emission Balmer lines are used for the ISMdiagnostic and star formation studies (Baldwin et al. 1981). For the vast majority of galaxies in our sample, wesee relatively weak emission lines on top of a stellar continuum. Therefore, in order to accurately measure emissionline fluxes, we need to precisely model stellar populations. Hence, when gas emission lines reside on top of a stellarcontinuum, any systematic uncertainty in the modelling of absorption lines will affect emission line measurements.Specifically, the age mismatch in the stellar population fitting will substantially bias Balmer line fluxes.In order to quantify this effect, we performed the following procedure: (i) We selected 2,000 spectra from our samplewith Balmer emission line intensities ranging from weak to strong based on their equivalent widths; (ii) we fitted thosespectra using stellar population model grids fixing the SSP age to 2, 4, 8, and 16 Gyr; (iii) we measured emission linefluxes in the fitting residuals in these four sets of spectra; (iv) we compared them to emission line fluxes obtained forbest-fitting stellar populations presented in our catalog.In Fig. 19 we present our results. It is clear, that the age mismatch affects emission line fluxes for weak lines:The systematic errors grow when lines become weaker, and the difference between the best fitting and the fixed agestemplates gets higher. When ages are underestimated by the fitting procedure (i.e. a galaxy is older than the ageof a template), Balmer emission line fluxes are underestimated too. Because forbidden lines often used in the gasstate diagnostics (e.g. [N ii ] or [O iii ]) do not lie on top of strong age sensitive absorption features, their fluxes remainvirtually unaffected, hence, moving a galaxy over the diagnostic plots (e.g. BPT) and potentially leading to theionization mechanism misclassification. CATALOG COMPILATION: SQL QUERY When selecting the core sample of galaxies we performed the following SQL query in the SDSS CasJobs service inthe DR7 context (see details in Section 2.1): SELECTp.objID, p.ra, p.dec,p.modelMag_u, p.modelMagErr_u, p.modelMag_g, p.modelMagErr_g,p.modelMag_r, p.modelMagErr_r, p.modelMag_i, p.modelMagErr_i,p.modelMag_z, p.modelMagErr_z, Chilingarian et al. β )-2-1012 E W ( H β ) - E W ( H β ) G y r SSP - 2Gyr Figure 19. The stellar population age mismatch effect on H β flux measurements. The difference of the H β EW computed using the bestfitting SSP template and a template with the age fixed to 2 Gyr is plotted against the measured H β EW for the best fitting SSP template.The age difference between the best fitting SSP age and 2 Gyr is color coded. petroMag_u, petroMagErr_u, petroMag_g, petroMagErr_g,petroMag_r, petroMagErr_r, petroMag_i, petroMagErr_i,petroMag_z, petroMagErr_z,p.fiberMag_u, p.fiberMagErr_u, p.fiberMag_g, p.fiberMagErr_g,p.fiberMag_r, p.fiberMagErr_r, p.fiberMag_i, p.fiberMagErr_i,p.fiberMag_z, p.fiberMagErr_z,p.petroR50_u, p.petroR50Err_u, p.petroR50_g, p.petroR50Err_g,p.petroR50_r, p.petroR50Err_r, p.petroR50_i, p.petroR50Err_i,p.petroR50_z, p.petroR50Err_z,p.extinction_u, p.extinction_g, p.extinction_r, p.extinction_i, p.extinction_z,s.specObjID, s.mjd, s.plate, s.fiberID,s.z, s.zerr, s.zconf, s.objType, s.sn_0, s.sn_1, s.sn_2,(SELECT stripe FROM dbo.fCoordsFromEq(p.ra,p.dec)) AS stripe,s.specClassINTO mydb.RCSED_SDSSFROM PhotoObj AS p, SpecObj as sWHEREs.bestObjid = p.objIDAND s.z >= 0.007AND s.z < 0.6AND s.specClass IN (dbo.fSpecClass(’GAL_EM’), dbo.fSpecClass(’GALAXY’)) This query returned 800,311 rows with 12 duplicate objects for which SDSS SpecObj table contains 2 records despiteit is documented to be clean from duplicates. We discard these duplicate spectra by keeping the record with higherS/N out of each pair of duplicates (and hence having e.g. better redshift estimate). From now we continue with thesample of 800,299 galaxies.The coordinates of obtained galaxies were then uploaded to the GALEX CasJobs service and the following querywas performed there in GALEXGR6Plus7 context: SELECTsdss.objid,galex_objid,nuv_mag, nuv_magerr, fuv_mag, fuv_magerr, CSED – Reference Catalog of Galaxy SEDs nuv_mag_aper_1, nuv_magerr_aper_1, nuv_mag_auto,fuv_mag_aper_1, fuv_magerr_aper_1, fuv_mag_auto,e_bvINTO mydb.RCSED_SDSS_GALEXFROM ( SELECTs.objid,(SELECT objid FROM dbo.fGetNearestObjEq(s.ra, s.dec, 0.05)) AS galex_objidFROM mydb.RCSED_SDSS_coords AS s) AS sdssJOIN photoObjAll AS pON sdss.galex_objid = p.objid This query returned 485,996 rows. CATALOG COLUMN DESCRIPTIONS In Tables 4–8 we provide descriptions and metadata for columns of the original tables of RCSED, which are shownin blue in Fig. 18. The external datasets available in the RCSED database are described in the corresponding originalpapers (see the text for references).This column information is identical for FITS tables distribution of the catalog, as well as when accessing theRCSED database through the Table Access Protocol, or using the catalog website http://rcsed.sai.msu.ru . Foreach column name in every table we give: (i) units (dash sign indicates that a column is dimensionless or units arenot applicable to it); (ii) data type in the database convention in order to guide a user on the precisionm and puropseof a column; (iii) IVOA Unified Content Descriptor (UCD) that helps one to identify equivalent physical quantitiesavailable for comparison in the VO or to associate a column and its uncertainty; and (iv) human readable descriptionof the column contents. When a table includes many similar columns as in the case of spectral lines properties in the rcsed lines gauss and rcsed lines nonpar database tables, we only give metadata for first group of columns in itand abridge the rest (Table 7 and Table 8). The complete list of emission lines included in our catalog and the columnname prefixes in rcsed lines gauss and rcsed lines nonpar are given in Table 2.4 Chilingarian et al. Table 4 Main catalog table ( rcsed ) columns metadata and descriptions. Column Units Datatype UCD Descriptionobjid - bigint meta.id;meta.main SDSS ObjID (unique identifier)specobjid - bigint meta.id SDSS SpecObjID (unique identifier within spectral galax-ies sample)mjd - integer time.epoch MJD of observationplate - smallint meta.id SDSS plate IDfiberid - smallint meta.id SDSS fiber IDra deg double pos.eq.ra;meta.main RA (J2000) of galaxydec deg double pos.eq.dec;meta.main Dec (J2000) of galaxyz - real src.redshift Galaxy redshiftzerr - real stat.error;src.redshift Uncertainty of galaxy redshiftzconf - real stat.fit.param;src.redshift SDSS r edshift confidencepetror50 r arcsec real phys.angSize SDSS r adius containing 50% of Petrosian fluxe bv mag real phot.color.excess E(B-V) at this (l,b) from SFD98specclass - smallint src.spType SDSS spectral classificationcorrmag fuv mag real phot.mag;em.UV.FUV Galactic extinction corrected total (Kron-like ellipticalaperture) magnitude in GALEX F UV filtercorrmag nuv mag real phot.mag;em.UV.NUV Same as above for GALEX NUV filtercorrmag u mag real phot.mag;em.opt.U Galactic extinction corrected total (Petrosian) magnitudein SDSS u filtercorrmag g mag real phot.mag;em.opt.B Same as above for SDSS g filtercorrmag r mag real phot.mag;em.opt.R Same as above for SDSS r filtercorrmag i mag real phot.mag;em.opt.I Same as above for SDSS i filtercorrmag z mag real phot.mag;em.opt.I Same as above for SDSS z filtercorrmag y mag real phot.mag;em.IR.J Same as above for UKIDSS Y filtercorrmag j mag real phot.mag;em.IR.J Same as above for UKIDSS J filtercorrmag h mag real phot.mag;em.IR.H Same as above for UKIDSS H filtercorrmag k mag real phot.mag;em.IR.K Same as above for UKIDSS K filtercorrmag fuv err mag real stat.error;phot.mag;em.UV.FUV Uncertainty of corrmag fuv columncorrmag nuv err mag real stat.error;phot.mag;em.UV.NUV Uncertainty of corrmag nuv columncorrmag u err mag real stat.error;phot.mag;em.opt.U Uncertainty of corrmag u columncorrmag g err mag real stat.error;phot.mag;em.opt.B Uncertainty of corrmag g columncorrmag r err mag real stat.error;phot.mag;em.opt.R Uncertainty of corrmag r columncorrmag i err mag real stat.error;phot.mag;em.opt.I Uncertainty of corrmag i columncorrmag z err mag real stat.error;phot.mag;em.opt.I Uncertainty of corrmag z columncorrmag y err mag real stat.error;phot.mag;em.IR.J Uncertainty of corrmag y columncorrmag j err mag real stat.error;phot.mag;em.IR.J Uncertainty of corrmag j columncorrmag h err mag real stat.error;phot.mag;em.IR.H Uncertainty of corrmag h columncorrmag k err mag real stat.error;phot.mag;em.IR.K Uncertainty of corrmag k columnkcorr fuv mag real arith.factor;em.UV.FUV K-correction for GALEX F UV magnitudekcorr nuv mag real arith.factor;em.UV.NUV Same as above for GALEX NUV magnitudekcorr u mag real arith.factor;em.opt.U K-correction for (Petrosian) SDSS u magnitudekcorr g mag real arith.factor;em.opt.B Same as above for SDSS g magnitudekcorr r mag real arith.factor;em.opt.R Same as above for SDSS r magnitudekcorr i mag real arith.factor;em.opt.I Same as above for SDSS i magnitudekcorr z mag real arith.factor;em.opt.I Same as above for SDSS z magnitudekcorr y mag real arith.factor;em.IR.J Same as above for UKIDSS Y magnitudekcorr j mag real arith.factor;em.IR.J Same as above for UKIDSS J magnitudekcorr h mag real arith.factor;em.IR.H Same as above for UKIDSS H magnitudekcorr k mag real arith.factor;em.IR.K Same as above for UKIDSS K magnitudeexp radvel km/s real spect.dopplerVeloc.opt Radial velocity (exp SFH)exp radvel err km/s real stat.error;spect.dopplerVeloc.opt Radial velocity error (exp SFH)exp veldisp km/s real phys.veloc.dispersion Velocity dispersion (exp SFH)exp veldisp err km/s real stat.error;phys.veloc.dispersion Velocity dispersion error (exp SFH)exp tau Myr real time.age Age (exp SFH)exp tau err Myr real stat.error;time.age Age error (exp SFH)exp met - real phys.abund.Z Metallicity (exp SFH)exp met err - real stat.error;phys.abund.Z Metallicity error (exp SFH)exp chi2 - real stat.fit.chi2 Goodness of fit (exp SFH)ssp radvel km/s real spect.dopplerVeloc.opt Radial velocity (SSP)ssp radvel err km/s real stat.error;spect.dopplerVeloc.opt Radial velocity error (SSP)ssp veldisp km/s real phys.veloc.dispersion Velocity dispersion (SSP)ssp veldisp err km/s real stat.error;phys.veloc.dispersion Velocity dispersion error (SSP)ssp age Myr real time.age Age (SSP)ssp age err Myr real stat.error;time.age Age error (SSP)ssp met - real phys.abund.Z Metallicity (SSP)ssp met err - real stat.error;phys.abund.Z Metallicity error (SSP)ssp chi2 - real stat.fit.chi2 Goodness of fit (SSP)zy offset mag real phot.mag;arith.diff Offset applied to UKIDSS magnitudes to correct for mis-match with SDSS onesspectrum snr - real stat.snr Signal-to-noise ratio of SDSS spectrum at 5500A (rest-frame) in the 20A box CSED – Reference Catalog of Galaxy SEDs Table 5 Fiber magnitudes table ( rcsed fibermags ) columns metadata and descriptions.Column Units Datatype UCD Descriptionobjid - bigint meta.id;meta.main SDSS ObjID (unique identifier)corrfibmag fuv mag real phot.mag;em.UV.FUV Galactic extinction corrected 3” aperture magnitudein GALEX F UV filtercorrfibmag nuv mag real phot.mag;em.UV.NUV Same as above for GALEX NUV filtercorrfibmag u mag real phot.mag;em.opt.U Galactic extinction corrected fiber (3” aperture) mag-nitude in SDSS u filtercorrfibmag g mag real phot.mag;em.opt.B Same as above for SDSS g filtercorrfibmag r mag real phot.mag;em.opt.R Same as above for SDSS r filtercorrfibmag i mag real phot.mag;em.opt.I Same as above for SDSS i filtercorrfibmag z mag real phot.mag;em.opt.I Same as above for SDSS z filtercorrfibmag y mag real phot.mag;em.IR.J Galactic extinction corrected 3” aperture magnitudein UKIDSS Y filtercorrfibmag j mag real phot.mag;em.IR.J Same as above for UKIDSS J filtercorrfibmag h mag real phot.mag;em.IR.H Same as above for UKIDSS H filtercorrfibmag k mag real phot.mag;em.IR.K Same as above for UKIDSS K filtercorrfibmag fuv err mag real stat.error;phot.mag;em.UV.FUV Uncertainty of corrfibmag fuv columncorrfibmag nuv err mag real stat.error;phot.mag;em.UV.NUV Uncertainty of corrfibmag nuv columncorrfibmag u err mag real stat.error;phot.mag;em.opt.U Uncertainty of corrfibmag ucorrfibmag g err mag real stat.error;phot.mag;em.opt.B Uncertainty of corrfibmag gcorrfibmag r err mag real stat.error;phot.mag;em.opt.R Uncertainty of corrfibmag rcorrfibmag i err mag real stat.error;phot.mag;em.opt.I Uncertainty of corrfibmag icorrfibmag z err mag real stat.error;phot.mag;em.opt.I Uncertainty of corrfibmag zcorrfibmag y err mag real stat.error;phot.mag;em.IR.J Uncertainty of corrfibmag ycorrfibmag j err mag real stat.error;phot.mag;em.IR.J Uncertainty of corrfibmag jcorrfibmag h err mag real stat.error;phot.mag;em.IR.H Uncertainty of corrfibmag hcorrfibmag k err mag real stat.error;phot.mag;em.IR.K Uncertainty of corrfibmag kkcorrfib fuv mag real arith.factor;em.UV.FUV K-correction for 3” aperture GALEX F UV magnitudekcorrfib nuv mag real arith.factor;em.UV.NUV Same as above for GALEX NUV magnitudekcorrfib u mag real arith.factor;em.opt.U K-correction for fiber (3” aperture) SDSS u magnitudekcorrfib g mag real arith.factor;em.opt.B Same as above for SDSS g magnitudekcorrfib r mag real arith.factor;em.opt.R Same as above for SDSS r magnitudekcorrfib i mag real arith.factor;em.opt.I Same as above for SDSS i magnitudekcorrfib z mag real arith.factor;em.opt.I Same as above for SDSS z magnitudekcorrfib y mag real arith.factor;em.IR.J K-correction for 3” aperture UKIDSS Y magnitudekcorrfib j mag real arith.factor;em.IR.J Same as above for UKIDSS J magnitudekcorrfib h mag real arith.factor;em.IR.H Same as above for UKIDSS H magnitudekcorrfib k mag real arith.factor;em.IR.K Same as above for UKIDSS K magnitude Chilingarian et al. Table 6 Gas phase metallicity table ( rcsed gasmet ) columns metadata and descriptions.Column Units Datatype UCD Descriptionid bigint meta.id;meta.main Primary keyobjid bigint SDSS ObjIDmjd d integer time.epoch MJD of observationplate smallint meta.id SDSS plate IDfiberid smallint meta.id SDSS fiber IDe bv mag real phot.color.excess Intrinsic E(B-V)gas oh d16 real phys.abund.Z Oxygen abundance of ionized gas (12 + log O/H) calculatedusing Dopita+16 calibration from Gaussian fit to emissionlinesgas oh d16 err real phys.abund.Z Error of oxygen abundance of ionized gas (12 + log O/H)calculated using Dopita+16 calibration from Gaussian fit toemission linesgas oh izi real phys.abund.Z Oxygen abundance of ionized gas (12 + log O/H) calculatedusing IZI calibration from Gaussian fit to emission linesgas oh izi errlo real stat.error;phys.abund.Z Lower error of oxygen abundance of ionized gas (12 + logO/H) calculated using IZI calibration from Gaussian fit toemission linesgas oh izi errhi real stat.error;phys.abund.Z Upper error of oxygen abundance of ionized gas (12 + logO/H) calculated using IZI calibration from Gaussian fit toemission linesq izi real phys.ionizParam.rad Ionization parameter calculated using IZI calibration fromGaussian fit to emission linesq izi errlo real stat.error;phys.ionizParam.rad Lower error of ionization parameter calculated using IZI cali-bration from Gaussian fit to emission linesq izi errhi real stat.error;phys.ionizParam.rad Upper error of ionization parameter calculated using IZI cali-bration from Gaussian fit to emission lines Table 7 Gaussian fit to emission lines table ( rcsed lines gauss ) columns metadata and descriptions.Column Units Datatype UCD Descriptionid bigint meta.id;meta.main Primary keyobjid bigint meta.id SDSS ObjIDmjd d integer time.epoch MJD of observationplate smallint meta.id SDSS plate IDfiberid smallint meta.id SDSS fiber IDforbid v km/s real phys.veloc Velocity measured simultaneously in all forbidden linesforbid v err km/s real stat.error;phys.veloc Uncertainty in the velocity measured simultaneouslyin all forbidden linesforbid sig km/s real phys.veloc.dispersion Velocity dispersion measured simultaneously in all for-bidden linesforbid sig err km/s real stat.error;phys.veloc.dispersion Uncertainty in the velocity dispersion measured simul-taneously in all forbidden linesallowed v km/s real phys.veloc Velocity measured simultaneously in all allowed linesallowed v err km/s real stat.error;phys.veloc Uncertainty in the velocity measured simultaneouslyin all allowed linesallowed sig km/s real phys.veloc.dispersion Velocity dispersion measured simultaneously in all al-lowed linesallowed sig err km/s real stat.error;phys.veloc.dispersion Uncertainty in the velocity dispersion measured simul-taneously in all allowed lineschi2 real stat.fit.chi2 Reduced goodness of fitf3727 oii flx 10 − erg/s/cm real phot.flux;spect.line Flux from Gaussian fit to continuum subtracted dataof [O ii ] (3727 ˚A) linef3727 oii flx err 10 − erg/s/cm real stat.error;phot.flux;spect.line Uncertainty in the flux from Gaussian fit to continuumsubtracted data of [O ii ] (3727 ˚A) linef3727 oii cnt 10 − erg/s/cm /˚A real phot.flux.density;spect.continuum Continuum level at [O ii ] (3727 ˚A) line centerf3727 oii cnt err 10 − erg/s/cm /˚A real stat.error;phot.flux.density;spect.continuum Uncertainty in the continuum level at [O ii ] (3727 ˚A)line centerf3727 oii ew ˚A real spect.line.eqWidth Equivalent width from Gaussian fit to continuum sub-tracted data of [O ii ] (3727 ˚A) linef3727 oii ew err ˚A real stat.error;spect.line.eqWidth Uncertainty in the equivalent width from Gaussian fitto continuum subtracted data of [O ii ] (3727 ˚A) line . . . . . . . . . . . . . . . CSED – Reference Catalog of Galaxy SEDs Table 8 Non-parametric fit to emission lines table ( rcsed lines nonpar ) columns metadata and descriptions.Column Units Datatype UCD Descriptionid bigint meta.id;meta.main Primary keyobjid bigint meta.id SDSS ObjIDmjd d integer time.epoch MJD of observationplate smallint meta.id SDSS plate IDfiberid smallint meta.id SDSS fiber IDforbid v km/s real phys.veloc Velocity measured simultaneously in all forbidden linesforbid sig km/s real phys.veloc.dispersion Velocity dispersion measured simultaneously in all for-bidden linesallowed v km/s real phys.veloc Velocity measured simultaneously in all allowed linesallowed sig km/s real phys.veloc.dispersion Velocity dispersion measured simultaneously in all al-lowed lineschi2 real stat.fit.chi2 Reduced goodness of fitf3727 oii flx 10 − erg/s/cm real phot.flux;spect.line Flux from non-parametric fit to continuum subtracteddata of [O ii ] (3727 ˚A) linef3727 oii flx err 10 − erg/s/cm real stat.error;phot.flux;spect.line Uncertainty in the flux from non-parametric fit to con-tinuum subtracted data of [O ii ] (3727 ˚A) linef3727 oii cnt 10 − erg/s/cm /˚A real phot.flux.density;spect.continuum Continuum level at [O ii ] (3727 ˚A) line centerf3727 oii cnt err 10 − erg/s/cm /˚A real stat.error;phot.flux.density;spect.continuum Uncertainty in the continuum level at [O ii ] (3727 ˚A)line centerf3727 oii ew ˚A real spect.line.eqWidth Equivalent width from non-parametric fit to contin-uum subtracted data of [O ii ] (3727 ˚A) linef3727 oii ew err ˚A real stat.error;spect.line.eqWidth Uncertainty in the equivalent width from non-parametric fit to continuum subtracted data of [O ii ](3727 ˚A) line](3727 ˚A) line