Mem.S.A.It.Vol.77,0 (cid:13)c SAIt 2004 Memoriedella SDSS spe tros opi survey of stars Zˇ.Ivezic´1,D.Schlegel2,A.Uomoto3,N.Bond4,T.Beers5,C.AllendePrieto6,R. Wilhelm7,Y.SunLee5,T.Sivarani5,M.Juric´4,R.Lupton4,C.Rockosi8,G.Knapp4,J. Gunn4,B.Yanny9,S.Jester9,S.Kent9,J.Pier10,J.Munn10,G.Richards11,H. 7 Newberg12,M.Blanton13,D.Eisenstein14,S.Hawley1,S.Anderson1,H.Harris10,F. 0 0 Kiuchi1,A.Chen1,J.Bushong1,H.Sohi1,D.Haggard1,A.Kimball1,J.Barentine15,H. 2 Brewington15,M.Harvanek15,S.Kleinman15,J.Krzesinski15,D.Long15,A.Nitta15,S. n Snedden15,fortheSDSSCollaboration a J 7 1 University of Washington, 2 Lawrence Berkeley National Laboratory, 3 Johns Hopkins 1 University, 4 Princeton University, 5 Michigan State University (JILA), 6 University of Texas,7 TexasTechUniversity8 UniversityofCaliforniaatSantaCruz,9 FermiNational 1 Accelerator Laboratory, 10 U.S. Naval Observatory, 11 Drexel University, 12 Rensselaer v PolytechnicInstitute,13NewYorkUniversity,14UniversityofArizona,15SpaceTelescope 9 ScienceInstitute,16ApachePointObservatory;e-mail:[email protected] 0 5 1 Abstract. Inadditiontoopticalphotometryofunprecedentedquality,theSloanDigitalSky 0 Survey(SDSS)isalsoproducingamassivespectroscopicdatabase.Wediscussdetermina- 7 0 tionofstellarparameters,suchaseffectivetemperature,gravityandmetallicityfromSDSS / spectra,describecorrelationsbetweenkinematicsandmetallicity,andstudytheirvariation h as a function of the position in the Galaxy. We show that stellar parameter estimates by p Beersetal.showagoodcorrelationwiththepositionofastarintheg−rvs.u−gcolor- - colordiagram,therebydemonstratingtheirrobustnessaswellasapotentialforphotometric o parameter estimation methods. Using Beers et al. parameters, we find that the metallic- r t ity distribution of the Milky Way stars at a few kpc from the galactic plane is bimodal s with a local minimum at [Z/Z ] ∼ −1.3. The median metallicity for the low-metallicity a ⊙ : [Z/Z⊙]<−1.3subsampleisnearlyindependentofGalacticcylindricalcoordinatesRandz, v whileitdecreaseswithzforthehigh-metallicity[Z/Z ] > −1.3sample.Wealsofindthat ⊙ i the low-metallicity sample has ∼2.5 times larger velocity dispersion and that it does not X rotate(atthe∼10km/slevel),whiletherotationalvelocityofthehigh-metallicitysample r decreasessmoothlywiththeheightabovethegalacticplane. a 1. Introduction cess that created a smooth distributions of stars, with the standard view exemplified by The formation of galaxies like the Milky the models of Bahcall&Soneira (1980) and Way was long thought to be a steady pro- Gilmore,Wyse,&Kuijken (1989), and con- strainedindetailbyMajewski (1993).Inthese Sendoffprintrequeststo:Zˇ.Ivezic´ Ivezic´etal.:SDSSspectroscopicsurveyofstars 1 models,theMilkyWayisusuallymodeledby encestherein).To brieflysummarizehere,the three discrete components: the thin disk, the fluxdensitiesofdetectedobjectsaremeasured thick disk, and the halo. The thin disk has a almost simultaneously in five bands (u, g, r, cold(σ ∼ 20kms−1)stellarcomponentanda i, and z) with effective wavelengths of 3540 z scale height of ∼300 pc, while the thick disk Å, 4760 Å, 6280 Å, 7690 Å, and 9250 Å. is somewhatwarmer(σ ∼ 40kms−1), with a The completeness of SDSS catalogs for point z largerscaleheight(∼1kpc)andloweraverage sourcesis∼99.3%atthebrightendanddrops metallicity ([Z/Z ] ∼ −0.6). In contrast, the to95%atmagnitudesof22.1,22.4,22.1,21.2, ⊙ halo component is composed almost entirely and 20.3 in u, g, r, i and z, respectively. The of low metallicity ([Z/Z ] < −1.5) stars and finalsurveyskycoverageofabout10,000deg2 ⊙ has little or no net rotation. Hence, the main will result in photometric measurements to differences between these components are in the above detection limits for about 100 mil- their rotational velocity, velocity dispersions, lion stars and a similar number of galaxies. andmetallicitydistributions. Astrometricpositionsareaccuratetoabout0.1 As this summary implies, most studies of arcsecpercoordinateforsourcesbrighterthan the Milky Way can be described as investiga- r ∼20.5m, and the morphologicalinformation tions of the stellar distribution in the seven- from the images allows robust point source- dimensionalspacespannedbythethreespatial galaxy separation to r ∼ 21.5m. The SDSS coordinates, three velocity components, and photometricaccuracyis0.02mag(root-mean- metallicity. Depending on the quality, diver- square,atthebrightend),withwellcontrolled sityandquantityofdata,suchstudiestypically tailsoftheerrordistribution.Theabsolutezero concentrate on only a limited region of this pointcalibrationoftheSDSSphotometryisac- space (e.g. the solar neighborhood), or con- curatetowithin∼0.02mag.Acompendiumof sider onlymarginaldistributions(e.g. number technical details about SDSS can be found in densityofstarsirrespectiveoftheirmetallicity Stoughtonetal. (2002)andontheSDSSweb orkinematics). site(http://www.sdss.org),whichalsoprovides Toenablefurtherprogress,adatasetneeds interfaceforthepublicdataaccess. to be both voluminous (to enable sufficient spatial, kinematic and metallicity resolution) 2.1. SDSSspectroscopicsurveyof and diverse (i.e. accurate distance and metal- stars licity estimates, as well as radial velocity and propermotionmeasurementsareneeded),and Targetsforthespectroscopicsurveyarechosen thesamplesneedtoprobeasignificantfraction fromtheSDSSimagingdata,describedabove, of the Galaxy. The Sloan Digital Sky Survey basedontheircolorsandmorphologicalprop- (hereafter SDSS, York et al. 2000), with its erties.Thetargetsinclude imaging and spectroscopic surveys, has re- cently provided such a data set. In this con- – Galaxies: simple flux limit for “main” tribution,wefocusontheSDSSspectroscopic galaxies, flux-color cut for luminous red surveyof stars andsome recentresults on the galaxies(cD) MilkyWaystructurethatitenabled. – Quasars:flux-colorcut,matchestoFIRST survey – Non-tiled objects (color-selected): cali- 2. SloanDigitalSkySurvey brationstars(16/640),interestingstars(hot white dwarfs, brown dwarfs, red dwarfs, The SDSS is a digital photometric and spec- carbonstars,CVs,BHBstars,centralstars troscopic survey which will cover up to ofPNe),sky one quarter of the Celestial Sphere in the North Galactic cap, and produce a smaller Here, (non)-tiledobjects refers to objects that area (∼225 deg2) but much deeper sur- are (not) guaranteed a fiber assignment. As vey in the Southern Galactic hemisphere an illustration of the fiber assignments, SDSS (Adelman-McCarthyetal. (2006) and refer- Data Release 5 contains spectra of 675,000 2 Ivezic´etal.:SDSSspectroscopicsurveyofstars galaxies, 90,000 quasars, and 155,000 stars. A pair of dual multi-object fiber-fed spectro- graphson the same telescope are used to take 640simultaneousspectra(spectroscopicplates havearadiusof1.49degrees),eachwithwave- lengthcoverage3800–9200Åandspectralres- olutionof∼2000,andwithasignal-to-noisera- tioof>4perpixelatg=20.2. Thespectraaretargetedandautomatically processedbythreepipelines: – target:Targetselectionandtiling – spectro2d:Extractionofspectra,skysub- Fig.1. Atestofthequalityofspectrophotometric traction, wavelength and flux calibration, calibration.Eachthincurveshowsaspectrumofa combinationofmultipleexposures hotwhitedwarf(whichwerenotusedincalibration) – spectro1d: Object classification, red- dividedbyitsbest-fitmodel.Thethickredcurveis shifts determination, measurement of line themedianofthesecurves. strengthsandlineindices Foreachobjectinthespectroscopicsurvey, similar in spectral type to the SDSS primary a spectral type, redshift (or radial velocity), standardBD+17 4708(an F8 star). The spec- and redshift error is determined by matching trum of each standard star is spectrally typed the measured spectrum to a set of templates. by comparing with a grid of theoretical spec- The stellar templates are calibrated using the trageneratedfromKuruczmodelatmospheres ELODIEstellarlibrary.Randomerrorsforthe usingthespectralsynthesiscodeSPECTRUM radialvelocitymeasurementsareastrongfunc- (Grayetal. 2001). The flux calibration vec- tionofspectraltype,butareusually<5kms−1 tor is derived from the average ratio of each forstarsbrighterthang∼18,risingsharplyto star and its best-fit model, separately for each ∼25kms−1forstarswithg=20.Usingasam- of the 2 spectrographs, and after correcting ple of multiply-observedstars, Pourbaixetal. forGalacticreddening.Sincetheredandblue (2005) estimate that these errors may be un- halvesofthespectraareimagedontoseparate derestimatedbyafactorof∼1.5. CCDs, separate red and blue flux calibration vectorsare produced.Thespectrafrommulti- pleexposuresarethencombinedwithbadpixel 3. TheutilityandanalysisofSDSS rejectionandrebinnedtoaconstantdispersion. stellarspectra The absolute calibration is obtained by tying TheSDSSstellarspectraareusedfor: ther-bandfluxesofthestandardstarspectrato the fiber magnitudes output by the photomet- 1. Calibrationofobservations ric pipeline (fibermagnitudesare correctedto 2. More accurate and robust source identifi- aconstantseeingof2arcsec,withaccounting cationthanthatbasedonphotometricdata for the contribution of flux from overlapping alone objectsinthefiberaperture). 3. Accuratestellarparametersestimation To evaluate the quality of spectrophoto- 4. Radialvelocityforkinematicstudies metric calibration on scales of order 100Å, the calibrated spectra of a sample of 166 hot DA white dwarfs drawn from the SDSS DR1 3.1. CalibrationofSDSSspectra White Dwarf Catalog (Kleinmanetal. 2004) Stellar spectra are used for the calibration of arecomparedtotheoreticalmodels(DAwhite allSDSSspectra.Oneachspectroscopicplate, dwarfsare usefulforthiscomparisonbecause 16 objects are targeted as spectroscopic stan- they have simple hydrogen atmospheres that dards. These objects are color-selected to be canbeaccuratelymodeled).Figure1showsthe Ivezic´etal.:SDSSspectroscopicsurveyofstars 3 results ofdividingeach white dwarfspectrum thebluetipofthestellarlocus(u−g<1),the byitsbestfitmodel.Themedianofthecurves expression showsa netresidualof order2%atthebluest [Z/Z ]=5.11(u−g)−6.33 (2) ⊙ wavelengths. Another test of the quality of spectropho- reproduces the spectroscopic metallicity with anrmsofonly0.3dex. tometric calibration is provided by the com- parison of imaging magnitudes and those These encouraging results are important synthesized from spectra, for details see for studies based on photometric data alone, andalsodemonstratetherobustnessofparam- VandenBerketal. (2004) and Smolcˇic´etal. (2006). With the latest reductions1 the two etersestimatedfromspectroscopicdata. types of magnitudes agree with an rms of ∼0.05mag. 3.4. MetallicityDistributionand Kinematics 3.2. SourceIdentification Due to large sample size and faint limiting magnitude (g ∼ 20), the SDSS stellar spec- SDSS stellar spectra have been success- tra are an excellent resource for studying the fully used for confirmation of unresolved bi- MilkyWaymetallicitydistribution,kinematics nary stars, low-metallicity stars, cold white andtheircorrelationallthewaytothebound- dwarfs, L and T dwarfs, carbon stars, etc. ary between the disk and halo at several kpc For more details, we refer the reader to above the Galactic plane (Juric´etal. 2006). Adelman-McCarthyetal. (2006) and refer- Herewepresentsomepreliminaryresultsthat encestherein. illustratetheongoingstudies. 3.3. StellarParametersEstimation 3.4.1. TheBimodalMetallicity SDSS stellar spectra are of sufficient qual- Distribution ity to provide robust and accurate stellar pa- rameters such as effective temperature, grav- Inordertominimizevariousselectioneffects, we study a restricted sample of ∼10,000 blue ity, metallicity, and detailed chemical compo- main-sequence stars defined by 14.5 < g < sition. Here we study a correlation between 19.5,0.7<u−g<2.0and0.25<g−r<0.35. thestellarparametersestimatedbyBeersetal. The last condition selects stars with the ef- group(AllendePrietoetal. 2006)andthepo- sitionofastarintheg−rvs.u−gcolor-color fectivetemperatureinthe narrowrange6000- 6500K.Thesestarsarefurtherconfinedtothe digram. main stellar locus by |s| < 0.04, where the s Figure 2 shows thatthe effective tempera- color,describedbyIvezic´etal. (2004),isper- turedeterminestheg−rcolor,buthasnegligi- pendicular to the locus in the g− r vs. u− g bleimpactontheu−gcolor.Theexpression color-color diagram (c.f. Fig. 2). We estimate log(T /K)=3.877−0.26(g−r) (1) distancesusingaphotometricparallaxrelation eff derivedbyJuric´etal.(2006). provides correct spectroscopic temperature The metallicity distribution for stars from withanrmsofonly2%(i.e.about100-200K) this sample that are at a few kpc from the forthe−0.3 < g−r < 1.0colorrange.While galacticplaneisclearlybimodal(seethemid- the medianmetallicityshowsa morecomplex dle panel in Fig. 3), with a local minimum at behaviorasfunctionoftheu−gandg−rcol- [Z/Z⊙] ∼ −1.3. Motivated by this bimodality, ors,itcanstillbeutilizedtoderivephotometric we split the sample into low- (L) and high- metallicity estimate. For example, for stars at metallicity (H) subsamples and analyze the spatialvariationoftheirmedianmetallicity.As 1 DR5/products/spectra/spectrophotometry.html, shown in the bottom panel in Fig. 3, the me- whereDR5=http://www.sdss.org/dr5 dian metallicity of the H sample has a much 4 Ivezic´etal.:SDSSspectroscopicsurveyofstars 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 0.8 0.9 1 1.1 1.2 1.3 0.8 0.9 1 1.1 1.2 1.3 Fig.2. Thetopleftpanelshowsthemedianeffectivetemperatureestimatedfromspectraof∼40,000stars as a function of the position in the g−r vs. u−g diagram based on imaging data. The temperature in eachcolor-colorbinislinearlycolor-codedfrom4000K(red)to10,000K(blue).Thebottomleftpanelis analogousexceptthatitshowsthebluetipofthestellarlocuswiththeeffectivetemperatureintherange 5300Kto6700K.Thetworightpanelsareanalogoustotheleftpanels,exceptthattheyshowthemedian metallicity,linearlycolor-codedfrom-0.5(red)to-2.5(blue). largergradientinthezdirection(distancefrom distribution appears Gaussian, with the width the plane),than in the R direction(cylindrical of0.35dexandcenteredon[Z/Z ]=−1.75. ⊙ galactocentricradius). In contrast, the median The decrease of the median metallicity metallicity of the L sample shows negligible with z for the H sample is well described by variationwiththepositionintheGalaxy(<0.1 [Z/Z ] = −0.65−0.15Z/kpcforZ < 1.5kpc ⊙ dexwithin4kpcfromtheSun)andthewhole and[Z/Z ]=−0.80−0.05Z/kpcfor1.5<Z < ⊙ Ivezic´etal.:SDSSspectroscopicsurveyofstars 5 Fig.4. The dots in the top panel show the radial velocity as a function of metallicity for stars with 0.25 < g−r < 0.35 and 160 < l < 200 (towards anticenter,wheretheradialvelocitycorrespondsto thev velocitycomponent). Thelargesymbolsare R the medians evaluated in narrow metallicity bins, andthedashedlinesshowthe2σenvelopesaround themedian.Theradialvelocitydistributionsforthe low-andhigh-metallicitysubsamples(separatedby 5 6 7 8 R (kp9c) 10 11 12 13 [Z/Z⊙] = −1.3)areshowninthebottompanel,to- Fig.3. Thedotsinthetoppanelshowthemetallic- getherwiththebest-fitGaussians(dashedlines)and ity of stars with 0.25 < g−r < 0.35 as a func- theirparameters.Notethatthevelocitydispersionis tion of the height above the Galactic plane. The ∼2.5timeslargerforthelow-metallicitysubsample. largesymbolsarethemediansevaluatedseparately for the low-metallicity ([Z/Z ] < −1.3) and high- ⊙ metallicity ([Z/Z ] > −1.3) subsamples, and the ⊙ dashedlinesshowthe2σenvelopesaroundtheme- 4 kpc (see the top panelin Fig. 3). For Z < 1 dian. Thehistogram inthe middlepanel illustrates kpc, most stars have [Z/Z⊙] > −1.3 and pre- the bimodality of metallicity distribution for stars sumably belong to thin and thick disks (for a with heights above the galactic plane between 1 recentdeterminationofthestellarnumberden- kpc and 2 kpc. The two solid lines are the best-fit sity based on SDSS data that finds two expo- Gaussians, and the dashed line is their sum. The nentialdisks,seeJuric´ 2006).Thedecreaseof dependence ofthemedianmetallicityforthehigh- themedianmetallicitywithzfortheHsample metallicitysubsampleonthecylindricalgalacticco- couldthusbeinterpretedasduetotheincreas- ordinatesRandzisshowninthebottompanel(lin- ingfractionofthelower-metallicitythickdisk early color-coded from −1.2 to −0.5, blue to red). stars. However, it is puzzling that we are un- Note that the z gradient is much larger than the R gradient. For the low-metallicity subsample, these abletodetectanyhintofthetwopopulations. gradientsarenegligible. Ananalogousabsenceofacleardistinctionbe- tween the thin and thick disks is also found whenanalyzingtheradialvelocitydistribution. 6 Ivezic´etal.:SDSSspectroscopicsurveyofstars Low-metallicity, velocity dispersion, 40-160 km/s Fig.6. AnalogoustoFig.5,exceptthatthevelocity dispersion is shown (color-coded from 40 km/s to 160km/s). Fig.7. A comparison of the radial velocity dis- tributionforlow-metallicitystarsobservedtowards l ∼ 90 and l ∼ 180. Note that the subsample ob- served towards the anti-center has a large velocity dispersion,inagreementwithFigure6. Fig.5. Thetoppanelshowsthemedianradialve- locityinLambertprojectionforstarsfromthelow- metallicity [Z/Z ] < −1.3 subsample, which have ⊙ component has about 2.5 times larger veloc- b>0andareobservedatdistancesbetween2.5kpc ity dispersion than the high-metallicity com- and 3.5 kpc. The radial velocity is linearly color- ponent.Ofcourse,thismetallicity–kinematics coded from -220 km/s to 220 km/s (blue to red, correlationwasknownsincetheseminalpaper green corresponds to0km/s). Thebottompanel is analogous, except thattheradialvelocitymeasure- by Eggen,Lynden-Bell&Sandage (1962), ments are corrected for thecanonical solar motion but here it is reproduced using a ∼100 times of220km/stowards(l=90,b=0). largersamplethatprobesasignificantlylarger Galaxyvolume. 3.4.2. TheMetallicity–Kinematics Correlation 3.4.3. TheGlobalBehaviorof Kinematics Inadditiontothebimodalmetallicitydistribu- tion, the existence of two populations is also The large sample size enables a robustsearch supported by the radial velocity distribution. for anomalous features in the global behavior As illustrated in Fig. 4, the low-metallicity ofkinematics,e.g.Sirkoetal. (2004).Forex- Ivezic´etal.:SDSSspectroscopicsurveyofstars 7 ample,whilethevariationofthemedianradial large velocity dispersion and a non-zero rota- velocity for the low-metallicity subsample is tionalcomponentinawell-defined∼1000deg2 well described by the canonical solar motion largeregion,perhapsduetostellarstreams. (Fig. 5), wefind anisolated∼1000deg2 large Thesepreliminaryresultsareonlybriefil- region on the sky where the velocity disper- lustrations of the great potential of the SDSS sion is larger (130 km/s) than for the rest of stellar spectroscopic database. This dataset the sky (100 km/s), see Figs. 6 and 7. This is willremaina cuttingedgeresourcefora long probablynotadataartefactbecausethedisper- time because other major ongoing and up- sion for the high-metallicity subsample does comingstellarspectroscopicsurveysareeither not show this effect. Furthermore, an analy- shallower(e.g.RAVE),orhaveasignificantly sis of the proper motion database constructed narrowerwavelengthcoverage(GAIA). by Munnetal. (2004) finds that the same starsalsohaveanomalous(non-zero)rotational Acknowledgements. Funding for the SDSS and velocity in the same sky region (Bondetal. SDSS-II has been provided by the Alfred P. SloanFoundation,theParticipatingInstitutions,the 2006). This kinematic behavior could be due National ScienceFoundation, theU.S.Department to the preponderanceof stellar streams in this of Energy, NASA, the Japanese Monbukagakusho, region (towards the anti-center, at high galac- theMaxPlanckSociety,andtheHigherEducation tic latitudes, and at distances of several kpc). FundingCouncilforEngland. Bondetal. (2006) also find, using a sample of SDSS stars for which all three velocity components are known, that the halo (low- References metallicitysample)doesnotrotate(atthe∼10 Adelman-McCarthy, J.K., et al. 2006, ApJS, km/s level), while the rotational velocity of 162,38 thehigh-metallicitysampledecreaseswiththe AllendePrieto,C.,etal.2006,ApJ,636,804 heightabovethegalacticplane. Bahcall, J.N. & Soneira, R.M. 1980, ApJSS, 44,73 4. Conclusions Bond,N.,etal.2006,inpreparation Eggen,O.J.,Lynden-Bell,D.&Sandage,A.R. We show that stellar parameter estimates by 1962,ApJ,136,748 Beers et al. show a good correlation with the Gilmore, G., Wyse, R.F.G. & Kuijken, K. position of a star in the g−r vs. u−g color- 1989,ARA&A,Volume27,pp.555-627 colordiagram,therebydemonstratingtheirro- Gray,R.O.,Graham,P.W. &Hoyt,S.R. 2001, bustnessaswellasapotentialforphotometric AJ,121,2159 stellar parameterestimationmethods.We find Ivezic´,Zˇ.,etal.2004,AN,325,583 that the metallicity distribution of the Milky Juric´,M.,etal.2006,submittedtoAJ Waystarsatafewkpcfromthegalacticplane Kleinman,S.J.,etal.2004,ApJ,607,426 is clearly bimodal with a local minimum at Majewski,S.R.1993,ARA&A,31,575 [Z/Z⊙] ∼ −1.3.Themedianmetallicityforthe Munn,J.A.,etal.2004,AJ,127,3034 low-metallicity [Z/Z⊙] < −1.3 subsample is Pourbaix,D.,etal.2005,A&A,444,643 nearly independentof Galactic cylindricalco- Sirko,E.,etal.2004,AJ,127,914 ordinatesRandz,whileitdecreaseswithzfor Smolcˇic´,V.,etal.2006,acceptedtoMNRAS thehigh-metallicity[Z/Z⊙]>−1.3sample.We Stoughton,C.,etal.2002,AJ,123,485 also find that the low-metallicity sample has VandenBerk,D.E.,etal.2004,ApJ,601,692 ∼2.5timeslargervelocitydispersion. York,D.G.,etal.2000,AJ,120,1579 Thesamplesdiscussedherearesufficiently largeto constrainthe globalkinematicbehav- iorandsearchforanomalies.Forexample,we findthatlow-metallicitystarsobservedathigh galactic latitudesat distancesof a few kpcto- wards Galactic anticenter have anomalously