Mon.Not.R.Astron.Soc.000,1–15(2002) PrintedAugust17,2015 (MNLATEXstylefilev2.2) Constraint on the time variation of the fine-structure constant / with the SDSS-III BOSS DR12 quasar sample 5 Franco D. Albareti,1(cid:63)† Johan Comparat,1 Carlos M. Gutie´rrez,2,3 Francisco Prada,1,4,5 1 Isabelle Paˆris,6 David Schlegel,7 Mart´ın Lo´pez-Corredoira,2,3 Donald P. Schneider,8,9 0 2 Arturo Manchado,2,3,10 D.A. Garc´ıa-Herna´ndez,2,3 Patrick Petitjean11 and Jian Ge12 g u 1InstitutodeF´ısicaTeo´rica(UAM/CSIC),UniversidadAuto´nomadeMadrid,Cantoblanco,E-28049Madrid,Spain A 2InstitutodeAstrof´ısicadeCanarias(IAC),LaLaguna,E-38205Tenerife,Spain 3DepartamentodeAstrof´ısica,UniversidaddeLaLaguna,LaLaguna,E-38206Tenerife,Spain 3 4CampusofInternationalExcellenceUAM+CSIC,Cantoblanco,E-28049Madrid,Spain 1 5InstitutodeAstrof´ısicadeAndaluc´ıa(CSIC),GlorietadelaAstronom´ıa,E-18080Granada,Spain 6INAF,OsservatorioAstronomicodiTrieste,ViaG.B.Tiepolo11,34131Trieste,Italy ] 7LawrenceBerkeleyNationalLaboratory,1CyclotronRoad,Berkeley,CA,94720,USA O 8DepartmentofAstronomyandAstrophysics,ThePennsylvaniaStateUniversity,UniversityPark,PA16802,USA C 9InstituteforGravitationandtheCosmos,ThePennsylvaniaStateUniversity,UniversityPark,PA16802,USA 10CSIC,Spain . h 11Institutd’AstrophysiquedeParis,CNRS-UPMC,UMR7095,98bisbdArago,75014Paris,France p 12DepartmentofAstronomy,UniversityofFlorida,Gainesville,FL32611-2055,USA - o r t AcceptedforpublicationinMNRAS. s a [ ABSTRACT 2 v From the Sloan Digital Sky Survey (SDSS) Data Release 12, which covers the full 0 Baryonic Oscillation Spectroscopic Survey (BOSS) footprint, we investigate the possi- 6 ble variation of the fine-structure constant over cosmological time-scales. We analyse the 5 0 largest quasar sample considered so far in the literature, which contains 13175 spectra 0 (10363fromSDSS-III/BOSS DR12 + 2812 from SDSS-II DR7) with redshift z < 1. We . applytheemission-linemethodonthe[Oiii]doublet(λλ4960,5008Å)andobtain∆α/α = 1 (0.9±1.8) × 10−5 for the relative variation of the fine-structure constant. We also investi- 0 gatethepossiblesourcesofsystematics:misidentificationofthelines,skyOHlines,Hβand 5 broadlinecontamination,GaussianandVoigtfittingprofiles,optimalwavelengthrangeforthe 1 : Gaussianfits,chosenpolynomialorderforthecontinuumspectrum,signal-to-noiseratioand v goodqualityofthefits.Theuncertaintyofthemeasurementisdominatedbytheskysubtrac- Xi tion.Theresultspresentedinthiswork,beingsystematicslimited,havesufficientstatisticsto constrainrobustlythevariationofthefine-structureconstantinredshiftbins(∆z≈0.06)over r a the last 7.9 Gyr. In addition, we study the [Neiii] doublet (λλ 3869,3968Å) present in 462 quasarspectraanddiscussthesystematiceffectsonusingtheseemissionlinestoconstrainthe fine-structureconstantvariation.Betterconstraintson∆α/α(<10−6)usingtheemission-line methodwouldbepossiblewithhigh-resolutionspectroscopyandlargegalaxy/qsosurveys. Key words: line: profiles – quasars: emission lines – cosmology: observations – surveys – large-scalestructureofUniverse. 1 INTRODUCTION mensionlessconstantsofphysicaltheories.Fundamentalconstants of physics could be thought of as parameters which enter in our Since Dirac’s philosophical argument (Dirac 1937) against the description of Nature but they cannot be predicted with our cur- fixed value of fundamental constants of Nature, several experi- renttheoriesandshouldbemeasured.Dirac’sideaisbasedonthe mentshave beenperformedto constrainpossiblevariationon di- unlikelyfactthatthemostfundamentalconstantsoftheUniverse haveacertainfixedvalue(atagivenenergy)withnoapparentre- lationwiththerealworld.Itismorelikelythattheirpresentval- (cid:63) ‘laCaixa’-SeveroOchoaScholar. ues are the result of a dynamical process, which had yielded the † E-mail:[email protected] (cid:13)c 2002RAS 2 F.D.Albaretietal. Table1.Summaryoftheresultsobtainedbyrecentworksbasedonthe[Oiii]emissionlinemethodforthepossiblevariationsofthefine-structureconstant. Reference Quasarspectra SDSSrelease zmin zmax Timeago(Gyr)(a) ∆α/α(×10−5) Bahcalletal.(2004) 42 EDR(Stoughtonetal.2002) 0.16 0.80 7.0 7±14 Gutie´rrez&Lo´pez-Corredoira(2010) 1568 DR6(Adelman-McCarthyetal.2008) 0.00 0.80 7.0 2.4±2.5 Rahmanietal.(2014) 2347 DR7(Abazajianetal.2009) 0.02 0.74 6.7 −2.1±1.6 Thiswork(2015) 13175 DR12(Alametal.2015) 0.04 1.00 7.9 0.9±1.8(b) (a)ForaΛCDMcosmologywithH0=67.8kms−1Mpc−1,Ωm=0.31andΩΛ=0.69fromPlanck+WMAP-9+BAO(PlanckCollaborationetal.2014). (b)Note:SincewehavealargersamplethanGutie´rrez&Lo´pez-Corredoira(2010),weexpectafactor≈ 2.5ofimprovementintheerrorjustfrompurely statisticalreasons.InFigs9and10,itisshownthattheerrorisdominatedbytheskysubtractionalgorithm,whichsuggeststhattheperformedanalysishave reachedthemaximumprecisionwiththeavailabledata. fundamentalconstantsastheyaremeasuredtoday.Therefore,they importantonebeingionizationandchemicalhomogeneity.These should be considered as characterizing the state of the Universe assumptionsmayinducesystematicbiasesonthevalueofα. (Uzan2003).Therearemanycurrenttheoreticalframeworkswhich Inthisarticle,weusethemethodbasedonthe[Oiii]emission allowforsuchvariationofthefundamentalconstants,forinstance, lines,firstproposedbyBahcall&Salpeter(1965),whichislessaf- stringtheory(Maeda1988),modifiedgravityandtheorieswithex- fectedbysystematics.Inparticular,thereisnoneedforassuming tradimensions(e.g.Cliftonetal.2012).Moreover,theexperimental ionizationandchemicalhomogeneity,sincethestudiedlineshave boundsontheirvariationhavebecomeastringenttestforthosethe- thesameprofile(thetransitionsoriginateatthesameupperenergy oreticalmodels(e.g.Thompson2012;Lealetal.2014).Themost level).Furthermore,theemission-linemethodsuffersofmuchless studiedfundamentalconstantsarethefine-structureconstantα,the spectral distortion, since the measurements of ∆α/α are done on Newton gravitational constantG and the electron-to-proton mass a spectral window ∼ 100Å as compared to ∼ 1000Å when the ratioµ(Uzan2003,2011;Garc´ıa-Berroetal.2007). MMmethodisused.Withalargeensembleofquasarsand/orus- Thefine-structureconstantgovernstheelectromagneticcou- ing high-resolution spectroscopy, the uncertainty can be reduced plingbetweenphotonsandchargedparticlesα=e2/((cid:126)c).Thecur- significantly, and will compete with the absorption method when rentconstraintonitsrelativevariation∆α/α,overgeologicaltime- usinghigh-resolutionspectroscopy. scales,is|∆α/α| < 7×10−8 uptoz ≈ 0.15(2Gyrago);obtained ThebeginningoftheSDSSsurveyopenedaneweraofpre- fromtheOklophenomenon(e.g.Petrov&etal.2006).Ithasalso cision, allowing us to use big samples of quasars; thus, reducing beenreported|∆α/α|<3×10−7uptoz≈0.45(4−5Gyrago)from the statistical uncertainty of the measurement of ∆α/α (see Ta- meteorites(Oliveetal.2002);whichalsoexcludespossiblevaria- ble1).Here,weextendtheseworksbyusingtheSDSS-III/BOSS tionsonthescalesoftheSolarsystem.Ontheotherhand,thereare Data Relase 12 (SDSS-DR12; Alam et al. 2015), which covers also constraints, |∆α/α| (cid:46) 10−2, based on the cosmic microwave the full Baryonic Oscillation Spectroscopic Survey (BOSS) sur- background (CMB; Landau & Sco´ccola 2010; Planck Collabora- veyfootprintwithanareacoverageof10000deg2.Incontrastto tionetal.2014)atz ≈ 1100andfrombigbangnucleosynthesis, thesepreviousinvestigations,weusespectraobtainedwiththecur- thelatterbeingmodel-dependent.Bymeasuringfine-structuremul- rent BOSS spectrograph (Smee et al. 2013) instead of the previ- tipletsatdifferentredshiftintheabsorptionoremissionspectraof ousSDSS-I/IIinstrument,makingourBOSSsampletotallyinde- galaxiesandquasars,locatedatdifferentdirectionsinthesky,one pendentfrompreviousworks.Moreover,thespectralrangeofthe can measure an estimate of the variation of α with time or space BOSSspectrographallowsanextensionoftheredshiftintervalfor overcosmologicalscales. the [Oiii] doublet from z = 0.8 to z = 1. The number of quasar Thefirstmeasurementsonthevariationofαfromastronom- spectraisincreasedbyafactorof5withrespecttoSDSS-DR7.All ical observations reached an accuracy of ∆α/α ≈ 10−2 − 10−3 thesespectrahavebeenvisuallyinspectedandclassifiedasquasars (Savedoff1956;Bahcall&Salpeter1965;Bahcall&Schmidt1967; bytheBOSScollaboration,andtheirproductsareprovidedinthe Bahcall, Sargent & Schmidt 1967). Since then, the methodology SDSS-III/BOSS Data Release 12 Quasar catalogue (DR12Q; see andunderstandingofsystematicshasdramaticallyimproved.Cur- Paˆrisetal.2015).Forthefinalconstrainton∆α/α,wecombinein rentmeasurementsofabsorptionmultipletsalongthelineofsight this work the BOSS sample with the previously studied SDSS-II ofthreequasarsaroundredshift1.5,observedwithspectralresolv- DR7quasarsample. ingpowerR≈60000atUVES/ESO-VLT,reachedthe≈ 5×10−6 There are several emission doublets, in addition to [Oiii] level (Evans et al. 2014). Using emission lines, an accuracy of (λλ4960,5008Å),thatcanbeusedtomeasure∆α/αasnotedby ≈ 2×10−5 was achieved analysing 1500−2300 quasar spectra Bahcalletal.(2004),andfirstusedbyGrupeetal.(2005).Gutie´rrez at z ≈ 0.6 (Gutie´rrez & Lo´pez-Corredoira 2010; Rahmani et al. &Lo´pez-Corredoira(2010)analyseddifferentdoubletsandfound 2014),takenwiththeSloanDigitalSkySurvey(SDSS)R ≈ 2000 that the [Neiii] (λλ 3869,3968Å) and [Siii] (λλ 6719,6733Å) spectrograph. doubletsappearinquasarspectrawithsufficientfrequencytohave The measurements on absorption features on a quasar spec- ameaningfulsample.Resultsfor[Siii]areconsistentwithnovari- trum are currently limited by the precision in the absolute wave- ationofthefine-structureconstant,althoughtheuncertaintyisan lengthcalibrationofthespectra,i.e.,50−200ms−1 usingspectra order of magnitude bigger than for [Oiii] , and this doublet can with R ≈ 60,000 (Molaro et al. 2013; Evans et al. 2014; Whit- only be used at low redshift < 0.4 for optical spectra. However, more&Murphy2015).Furthermore,theso-calledmany-multiplet they obtained a positive variation of the fine-structure constant, (MM)methodusedinEvansetal.(2014),althoughmoreprecise, ∆α/α=(34±1)×10−4,whenthe[Neiii]linesareused.Noexpla- remains controversial as several assumptions are made, the most nationwasfoundforthispositivevariation.Inthiswork,wealso (cid:13)c 2002RAS,MNRAS000,1–15 Fine-structureconstantwithBOSS 3 1400 60 S(cid:144)N >10 1200 @OIIID 50 S(cid:144)N >25 @OIIID 1000 0 40 ars S(cid:144)N@OIIID>50 200 30 uas 800 J q Dec 20 o.of 600 N 10 400 0 200 -10 0 0 50 100 150 200 250 300 350 0.0 0.2 0.4 0.6 0.8 1.0 RAJ2000 Redshift Figure1.Left-handpanel:skydistributionofthefullSDSS-III/BOSSDR12Qquasars(297301)inJ2000equatorialcoordinates.Right-handpanel:number ofquasarswith[Oiii]emissionlinesinourfiducialsample(10363quasars)in∆z = 0.05bins.S/N[Oiii]5008 > 10(10363quasars),blacksolidline; S/N[Oiii]5008>25(4015quasars),bluedashedline;andS/N[Oiii]5008>50(1498quasars),reddottedline. analysethe[Neiii]linestocheckwhetherthesameeffectispresent plegalaxyandquasarspectrumtemplatesforallallowedredshifts. inourBOSSquasarsample. Then,avisually-inspectedquasarcatalogueisbuiltfromtheseob- There are investigations which use Si iv absorption lines jects.OurfiducialsampleisobtainedfromtheDR12Qcatalogue (λλ1394,1403Å)toobtainaprecisionof4×10−6 (Chandetal. version(Paˆrisetal.2015). 2005). This method also avoids the assumption of ionization and ThewavelengthcoverageoftheSDSS-III/BOSSspectrograph chemicalhomogeneity.However,sincetheseparationbetweenboth is 3600-10400 Å and that of the SDSS-II spectrograph is 3800- linesisonly≈9Å,thewavelengthprecisionneededinthelabora- 9200 Å. The BOSS sample is homogeneous since all the spectra toryfortheseparationbetweenbothlinesisfivetimeshigherthan have been obtained with the same instrument, and it is indepen- using[Oiii]lines.Nevertheless,theseconstraintsapplytothered- dentfrompreviousinvestigations.Thewidercoverageofthenew shift interval 1.59 < z < 2.92, which does not overlap with our spectra allows consideration of higher redshifts (up to z = 1 for range,thustheyarecomplementarytotheonesreportedinthisre- [Oiii]doublet)thaninthepreviousSDSS-IIanalysisbasedonthe search. samemethod(seeTable1).TheBOSSspectrographhastwochan- Finally,inthelightoftheupcominglargegalaxysurveys,like nels (blue and red) whose wavelength coverage is 3600-6350Å eBOSSandDESI,thatwillprovidemillionsofhigh-redshiftgalaxy and5650-10400Å,respectively.Theresolvingpowerrangesfrom spectra, we also discuss using galaxies instead of quasars to set 1560at3700Åto2270at6000Å(bluechannel)andfrom1850at constraintsonthefine-structureconstant. 6000Åto2650at9000Å(redchannel).Foroursample,the[Oiii] Thepaperisorganizedasfollows.First,inSection2,wede- linesfallintheredchannelfor> 96%ofthequasars.Thenum- scribe the data set used for our analysis. Next, in Section 3, the berofpixelsofeachspectrumisabout4600fortheBOSSspectra methodologyispresented,theemission-linemethodisexplained, and3800fortheSDSS-I/IIspectra.Thepixelspacingisuniformin andthecodeandsimulationstoanalysethespectraaredescribed. log-wavelengths(∆logλ = 10−4dex).Morecompleteinformation In Section 4, we study several samples to check for systematics. abouttheSDSS-I/IIandBOSSspectrographscanbefoundinSmee Then,ourresultsarepresentedinSection5.Finally,weprovidein etal.(2013). Section 6 a summary of the main conclusions achieved with this researchproject. 2.1 Dataselection The SDSS-III/BOSS DR12Q catalogue contains 297301 objects. 2 SAMPLEDESCRIPTION Fig. 1 (left-hand panel) shows the quasar distribution in the sky. Wesummarizebelowthemainselectioncriteriainordertodefine All the spectra used in this investigation were downloaded from ourfiducialsamplefromthiscatalogue. the SDSS Database. This survey (York et al. 2000), which began takingobservationsin1998,consistsofamassivecollectionofop- (i) Redshift< 1.Thislimitationisimposedbythewavelength ticalimagesandspectrafromastronomicalobjectsincludingstars, range of the BOSS optical spectrograph and the position of the galaxies and quasars. For this purpose, there is a dedicated 2.5- [Oiii] lines. This criterion decreases the sample down to 45802 m wide-angle optical telescope at Apache Point Observatory in quasars. NtheirwdMpheaxsiecoof(UthSiAs;prfoojremcto(rSeDdSeSta-iIlIsI,;sEeeiseGnusnteninetetala.l2.020061)1.)Tihne- (ii) S/N[Oiii]5008 > 10. We impose a mild constraint on the signal-to-noiseratio(S/N)ofthestronger[Oiii]line(5008Å)in cludes BOSS (Dawson et al. 2013) among its four main surveys. ordertopreservealargenumberofspectra.Constraintsontheex- ThedataanalysedinthisresearchwereprovidedbyBOSS,andit pectedwidthandamplitudesofthelineshelpinavoidingmisiden- isusedformeasuring∆α/αforthefirsttime.TheSDSS-III/BOSS tificationsofthe[Oiii]doublet(seeSection4).Thisselectionre- pipeline(Boltonetal.2012)classifiestheobjectsasquasarswith ducesthesamplefrom45802to13023objects. aχ2minimizationproceduretofittheobservedspectrumtomulti- (cid:13)c 2002RAS,MNRAS000,1–15 4 F.D.Albaretietal. ΛHÞLHobservedL 6000 7000 8000 9000 4 SDSSJ112639.11+543826.5 50 z=0.612 L -1 3 @OIIIDlines 20 Þ -1 10 -2s 2 ÞL m H 5 c ΕÈÈ erg 1 2 -170 0 1 H 1 Λ f -1 0.2 0.4 0.6 0.8 1.0 7980 8000 8020 8040 8060 8080 8100 Redshift ΛHÞLHobservedL Figure2.Left-handpanel:datapoints(1416)forwhich|(cid:15)| = |δλz/(1+z)−δλ0|,namelytheabsolutevalueofthedifferencebetweenthemeasuredline separationatredshiftzinrestframeandthelocalone,isbiggerthan1Åplottedasafunctionofredshift(andthewavelengthobservedfor[Oiii]4960).We comparewithatypicalskyspectrum:the[Oiii]positionsforthesespectracorrelatewiththeskyemissionlines.Hence,thesehighvaluesof|(cid:15)|areduetobad skysubtractionsand/orlowS/N.Thesespectraareremoved.Right-handpanel:aspectrumremovedfromthesamplebecauseoftheskyemission-linecriteria. Forthisquasar,weget(cid:15) =1.2±0.6Å.Theweak[Oiii]lineisaffectedbythesubtractionofthe7995ÅOHskyemissionline,indicatedbytheverticalred dashedline. (iii) Non-converging fits. Since we analyse spectra with low Weobtainaweightedmeanfor∆α/αusingasweightstheuncer- S/N,therearesomecaseswheretheGaussianfittothelinesdoes taintyin∆α/αcomputedwiththestandarderrorsforthepositionof notconverge.1244spectraarediscarded,leavinguswith11779 thelinesderivedfromtheGaussianfits.ThecontaminationofHβ spectra. isautomaticallytakenintoaccount.Forinstance,abroadHβline nearthe[Oiii]4960linemeansabadGaussianfit.Thus,weobtain (iv) Sky emission lines. Strong atmospheric lines, for instance theOi5578Åline,arepoorlyornotcompletelyremovedbythe largererrorsinthepositionofthelinecentroidsand,consequently, in∆α/α.InSection4,weanalyseseveralsampleswheretheS/N SDSSskysubtractionalgorithm.Thismayleadtoawrongiden- Hβ tificationofthe[Oiii]linesandtoincludelowS/N[Oiii] spectra isconstrainedtocheckthattheHβcontaminationhaslittleweight (Gutie´rrez & Lo´pez-Corredoira 2010). Both effects will produce onthefinalconstraintvalue. An electronic table is published along with the paper which outliers.WeusetheSDSSskymaskforLyαforeststudieswhich contains all the information of each spectrum from our fiducial contains 872 lines (see Delubac et al. 2015, for more details) to removespectrawhose[Oiii]linesliewithinaparticulardistance sampleof10363quasars(seeAppendixA). The distribution of the selected quasars in redshift accord- from the strongest sky lines. Even though we vary the distance [Oiii]–skylines,usedifferentsetofskylines(accordingtotheir ingtotheirselectedS/N[Oiii]5008 isplottedinFig.1(right-hand intensity),orevaluateotherconditions(S/N,fiterrors,etc.)tore- panel). Fig. 3 (left-hand panel) displays a composite image built moveaffectedspectra;weusuallyeliminate3−5goodspectrafor withallthespectrafromourfiducialsamplesortedbyredshift.The right-handpanelshowsthe[Oiii]doubletinrestframe. eachbadspectraeliminated.Thus,thesetestsdecreasesignificantly thenumberofquasarswhilenotbeingveryeffective:typically50% oftheoutliersarenotremoved.Thus,wedecidedtoeliminateall spectraforwhichtheseparationbetweenbothlinesdifferbymore than1Åfromthelocalvalue(seethelastparagraphinSection3.3). 3 METHODOLOGY Fig.2(left-handpanel)showsthatthedistributionoftheseoutliers 3.1 Measurementmethod is correlated with a typical sky spectrum. From a visual inspec- tion, we observed that these spectra have low S/N, and they are Tofirstorder,thedifferencebetweentheenergylevelsofanatom in fact contaminated by sky emission line subtraction (see right- isproportionaltoα2.Transitionsbetweenenergylevelsofthesame handpanelofFig.2).Thiseffectcausesustodiscard1416spec- atomatagivenionizationlevel,withthesameprincipalquantum tra(12%oftheprevious11779quasars).Finally,wehave10363 number and different total angular momentum J, have an energy quasarspectra(our‘fiducialsample’). differenceproportionaltoα4.Thesegroupsoftransitionsarecalled fine-structuremultiplets.Savedoff(1956)firstrealizedthatthefine The presence of broad Hβ emission line (4861 Å) near the structureoftheseenergylevelscouldbeusedtobreakthedegener- weak [Oiii] line 4960 Å could produce a blueshift in the deter- acybetweentheredshifteffectandapossiblevariationofα. mination of the [Oiii] line position. This could mimic a positive The value of the fine-structure constant can be measured variationonthefine-structureconstant.Therefore,aconstrainton throughtheseparationbetweenabsorptionoremissionmultiplets the strength and/or width of the Hβ emission line has been im- inthespectraofdistantquasars(Uzan2003)as posedonpreviousinvestigations(Bahcalletal.2004;Gutie´rrez& Lo´pez-Corredoira2010;Rahmanietal.2014).However,wedonot ∆α 1(cid:40)[(λ −λ )/(λ +λ )] (cid:41) (z)≡ 2 1 2 1 z −1 , (1) restrict any characteristic of the Hβ line in our fiducial sample. α 2 [(λ −λ )/(λ +λ )] 2 1 2 1 0 (cid:13)c 2002RAS,MNRAS000,1–15 Fine-structureconstantwithBOSS 5 CIIID MgII @NeVD@OIID@NeIIID HΓ HΒ @OIIID @OIIID4960 @OIIID5008 0.9 0.9 0.8 0.8 0.7 0.7 hift hift s s d d e e R R 0.6 0.6 HΑ 0.5 0.5 0.4 0.4 0.3 0.3 0 0 4000 5000 6000 7000 8000 9000 10000 4940 4960 4980 5000 5020 ΛHÞLHobservedL ΛHÞLHrestframeL Figure3.Compositeimagewithourfiducialsampleof10363BOSSquasarspectrasortedbyredshift.Left-handpanel:thewholerangeofwavelengthsis shown.Fromrighttoleft,thestrongestemissionlinesareHα6565Å;[Oiii]λλ4960,5008Å;Hβ4861Å;Hγ4341Å;[Neiii]λλ3869,3968Å;[Oii]3730 Å;[Nev]3426Å;Mgii2796ÅandCiii]1906Å.Thenarrowstraightlineat5579Åisthestrong[Oi]atmosphericline.Right-handpanel:wavelength intervalcentredatthe[Oiii]doubletinrestframe. where λ (λ > λ ) are the wavelengths of the transitions and lengthseparation,indirectlywiththetheta-pinchdischargeanddi- 1,2 2 1 subscript 0 and z stand for their value at redshift zero (theoreti- rectly with the Michelson interferometer, are in good agreement, cal/laboratoryvalues)andatredshiftz,respectively.Forillustrative being the Michelson interferometer more accurate with an error purposes,expression(1)canbeapproximatedby <5×10−4Å. ∆α (cid:15) From equation (2), a determination of (cid:15) with a precision of ≈ , (2) 1Å allows for an uncertainty of 10−2 in ∆α/α when using the α 2δλ 0 [Oiii] doublet. The precision from the NIST atomic data allows whereδλ0 = [λ2−λ1]0 isthelocalz = 0separationbetweenboth for a determination of ∆α/α up to 10−5, which is a bit less than wavelengths,and(cid:15) = δλz/(1+z)−δλ0 isthedifferencebetween the uncertainty in our result. One could perform a blind analysis themeasuredlineseparationatredshiftzinrestframeandthelocal inordertosearchforapossiblevariationonα,wheretheabsolute one.Thus,inprinciple,thelargerthedifferencebetweenthepairof wavelengthvaluesarenotrequired,ifonehadalargeenoughsam- lines,thebettertheprecisionformeasuring∆α/α. pledistributedinredshift.However,theprecisionontheabsolute Concerning emission lines, the most suitable pair of lines is wavelengthslimitstheusefulnessofhigh-resolutionspectroscopy the[Oiii]doublet,whichisoftenpresentinquasarspectrawithrel- untilbettermeasurementsofthe[Oiii]lines(orjusttheirsepara- ativelyhigh-S/N.Thevacuumvaluesforthe[Oiii]doubletwave- tion)areavailable. lengthsare λ[Oiii] =4960.295Å λ[Oiii] =5008.240Å (3) 1 2 3.2 Implementation δλ[Oiii] =47.945Å, (4) 0 The code developed for the analysis of the quasar spectra fol- whicharepublishedintheNISTAtomicSpectraDatabase.1These lows the one described in Gutie´rrez & Lo´pez-Corredoira (2010), although there are some modifications and more information has transitionsareforbidden(theycorrespondtomagneticdipoleand beenextractedfromtheanalysis.Wedescribethemaincharacter- electric quadrupole transitions), and they are not observed in the isticsofourcodebelow. laboratory.Thewavelengthexperimentalvaluesareobtainedindi- rectly by first computing the energy levels from observed wave- lengthsusingatheta-pinchdischarge(Pettersson1982).Thewave- 3.2.1 Wavelengthsampling lengthseparationhasdirectlybeenmeasuredintheinfraredfromH iiregionsusingaballoon-bornetelescopeandMichelsoninterfer- Weconsideronlytheexperimentaldatatogetherwiththeirerrors ometer(Moorwoodetal.1980).Bothmeasurementsofthewave- asprocessedbytheSDSSpipelinetoobtaintheconstraintonthe possiblevariationofα.Wedonotresamplethewavelengthrange by using an interpolation method. Since the pixel spacing is uni- 1 http://physics.nist.gov/PhysRefData/ASD/linesform.html forminlog-wavelengths,agivenrangeofwavelengthsinrestframe (cid:13)c 2002RAS,MNRAS000,1–15 6 F.D.Albaretietal. SDSSJ124144.40+601619.7 SDSSJ120839.61+340406.4 SDSSJ082251.37+455946.2 Þ-1L 15 z=0.291 Þ-1L 15 z=0.404 Þ-1L 25 z=0.679 -1s -1s -1s 20 -2cm 10 -2cm 10 -2cm 15 g g g er er er 10 -170 5 -170 5 -170 H1 H1 H1 5 fΛ 0 H∆HΓ HΒ@OIIID HΑ fΛ 0 H∆HΓ HΒ@OIIID HΑ fΛ 0 MgII H∆ HΓ HΒ@OIIID 4 4 4 2 2 2 fΛ 0 fΛ 0 fΛ 0 D D D -2 -2 -2 -4 -4 -4 4000 5000 6000 7000 8000 9000 10000 4000 5000 6000 7000 8000 9000 10000 4000 5000 6000 7000 8000 9000 10000 ΛHÞLHobservedL ΛHÞLHobservedL ΛHÞLHobservedL Figure4.Seventh-orderpolynomialfits(red)tothecontinuumspectrumwiththeirresidualsforthreetypicalquasarspectraatdifferentredshift.Thegapsin theresidualsarethemaskedregionscorrespondingto(fromrighttoleft)Hα,the[Oiii]doublet,Hβ,Hγ,HδandMgii(blackdashedlines). (λ−,λ+)hasthesamenumberofpixelsN,i.e. emissionlinesina15Åwindowaroundtheexpectedlocationof N ∝(cid:90) λ+(1+z)d(cid:0)logλ(cid:1)=logλ+(1+z) =logλ+ , (5) tthhee [rOedsiihi]ifltinbeass.edThoenSaDχS2Sfiptiptoelidnieffeprreonvtidteems paladteetse;rmwienarteifoenr otof λ−(1+z) λ−(1+z) λ− Boltonetal.(2012)formoredetails.Theseredshiftestimateshave andisindependentoftheredshiftoftheobject.Allthewavelength errors between 10−4 and 10−5, which are sufficient for our pur- intervalswiththesamewidthinrestframewillhavethesamenum- poses. Moreover, there is also a visual redshift estimation which berofexperimentalpoints. can be found in the quasar catalogue DR12Q (Paˆris et al. 2015). Thedifferencebetweenbothredshiftestimates(ifany)isusually 3.2.2 Fitofthecontinuumspectrum |z −z |≈5×10−4.Wedecidedtoadoptthevisualredshifts. vis pipe First,wefitaseventh-orderpolynomialtosubtractthecontinuum Thecentroidpositionsofthe[Oiii]emissionlinesaredeter- spectrumwhilemaskingregionswherestrongandwideemission minedbyfourdifferentmethods. linesarepresent(Hα,Hβ,Hγ,Hδ,Mgiiandthe[Oiii]doublet). OurmethoddiffersfromGutie´rrez&Lo´pez-Corredoira(2010)in (i) Gaussianprofilemethod. that they use a cubic local spline to fit the continuum masking First, we search for the maximum flux value in an ∼ 15(1+ strongemissionlines.Thechosenorderofthepolynomialprovides z)Å window around the expected position of the line (according enoughdegreesoffreedomtoreproducedifferentcontinuumfea- totheredshiftprovidedbytheDR12Qcatalogue).Thisprocedure tures.InSection3,wetesthowourmeasurementfor∆α/αisaf- automaticallyerasesanybiasproducedbytheredshiftvalue.Then, fectedbychangingthepolynomialorder.Hundredsofcontinuum wemakeaninitialGaussianfitaroundthepositionofthemaximum spectra fits were checked by eye. The residuals from the fits are fluxvalueusingafixedwidthof∼10(1+z)Å.Fromthisfirstfit, smallerthantheerrorsonthefluxdensities.Fig.4showsthreedif- weobtainanewpositionforthelinecentroidandaGaussianwidth. ferentspectrawiththeircontinuumfitandresiduals. These values are used as initial parameters for the final fit of the lines;namely,thewavelengthrangeconsideredtoperformthefinal 3.2.3 Signal-to-noiseratio fitiscentredaroundthepositionofthelinecentroid,anditisfour WefollowGutie´rrez&Lo´pez-Corredoira(2010)forthedetermina- timestheGaussianwidthofthelines.Thisapproachmeansthatwe tionofS/N.Hence,wecomputethestandarddeviationoftheflux considerpixelsupto2σawayfromthecentreoftheline.Hence, between 5040(1+z) and 5100(1+z)Å (where z is the redshift somelinesarefittedusing∼4−5pixels,whileotherswith∼15−20 of the quasar) where there are no strong emission or absorption pixelsdependingonthelinewidth.Thefittakesintoaccountthe lines. Then, we search for the maximum of the [Oiii] 5008 line, fluxerrorsforeachpixel,i.e.,weusetheivarcolumnfoundineach anddetermineS/N[Oiii]5008astheratiobetweenthemaximumof spectrumasweightsforthefit.Ourfinalcentroidmeasurementfor thelineandthepreviouslycomputedstandarddeviation.Although eachconsideredlinecorrespondstothecentroidoftheGaussianfit for a more reliable determination of the S/N, it is better to use a doneinthelaststepoftheadoptedprocedure.Wealsoderivean Gaussianfittotheline.Thisprocedureavoidspossibleissuesre- errorfor∆α/αusingthestandarderrorsforthecentrepositionof latedwhenfittingdatawithverylowS/N.ThisS/Nisusedinthe theGaussians.Thisisourmainmethodformeasuringα. criterionii(Section2)tobuildourfiducialsample. (ii) Voigtprofilemethod. FollowingthesameprocedurethanwhenusingaGaussianpro- 3.2.4 Measurementoftheemission-linewavelengths file,wemakethefitwithaVoigtprofileinsteadofaGaussian.More Tomeasurethewavelengthsofthe[Oiii]doublet,ourfittingcode precisely,weuseapseudo-Voigtprofilewhichisalinearcombi- needs as input an accurate estimate of the redshift of the quasar, nationofaGaussianandaLorentzianprofile.Then,wehaveone at least with an error ∆z<3×10−3. This allows a search for the moreparameter,i.e.theamplitudeoftheLorentzianfunction,while (cid:13)c 2002RAS,MNRAS000,1–15 Fine-structureconstantwithBOSS 7 350 3500 SDSSJ121417.80+293143.4 SDSSJ121417.80+293143.4 z=0.063 300 z=0.063 3000 @OIIIDlines @NeIIIDlines -1L 2500 -1L 250 Þ Þ -1 -1 -2s 2000 -2s 200 m m c c g g er 1500 er 150 -17 -17 0 0 1 1 H 1000 H 100 Λ Λ f f 500 50 0 0 5260 5280 5300 5320 5340 4100 4120 4140 4160 4180 4200 4220 ΛHÞLHobservedL ΛHÞLHobservedL Figure5.[Oiii](left-handpanel)and[Neiii](right-handpanel)linesforSDSS-J121417.80+293143.4,atredshiftz = 0.063.Themeasured∆α/αforthis quasaris∆α/αG[Oausisii] =(2.3±7.6)×10−4,∆α/αV[Ooigitii] =(3.3±12.6)×10−4and∆α/αG[Nauessiii] =(39±8)×10−4,∆α/αV[Noigetiii] =(37±9)×10−4.The measured∆α/αfor[Neiii]isnotconsistentwithzeroregardlessoftheprofile;seeFig.14andlastparagraphofSection4fordiscussion.Eachpanelshows thefluxdensityforeachpixelwiththeirrespectiveerrorbars(solidsymbols),togetherwiththeGaussianfit(dottedredcurve)andthepseudo-Voigtprofile (thickgreycurve)toeachofthelines.Thefittingprocedure(describedinthetext)onlytakesintoaccounttheexperimentaldata(solidsymbols)weightedby theirerrorbars.NoticehowthedeviationofthelinecentroidpositionderivedfromourGaussianfit(verticalsolidline)withrespecttotheexpectedposition oftheline(verticaldashedline)accordingtothevisualredshiftprovidedbytheDR12Qcataloguearewellcorrelatedforthesamepairof[Oiii]and[Neiii] lines,andforthedifferentsetoflines.Thegreenshadedverticalareashighlighttheuncertaintyfortheexpectedpositionofthelinesduetothequasarredshift error(≈ 5×10−4).Alsoshownisafourth-ordersplineinterpolationtothespectrumaftersubtractingthecontinuum(thinsolidline).The[Neiii]linesare weakerbyoneorderofmagnitudethanthe[Oiii]lines,whichisusuallythecaseforallthespectrashowingbothpairoflines.Theweaklinenearthestronger [Neiii]lineisblendedwithHei(3889.75Å)andHζ(3890.16Å). its width and its position are the same as those for the Gaussian 3.3 Simulatedspectra profile. In Fig. 5, we depict the [Oiii] and [Neiii] lines for the In order to test the robustness and accuracy of our methodology, we generate realizations of quasar spectra using as noise a nor- same quasar spectrum to illustrate the Gaussian and Voigt fitting mal distribution centred at the flux value, and taking the error in methods. each pixel as the standard deviation. From our fiducial sample (iii) Integrationmethod. (10363 quasars), we simulate 100 realizations for each spectrum Here, the centroids of the lines are obtained by integrating (> a million in total). This number of realizations provides rea- around 1σ from the position of the fitted Gaussian profile. This sonablestatisticstoderiveanerrorfromthestandarddeviationof techniqueprovidesindicationsofwhetherthereisHβcontamina- themeasurementsontherealizationsofeachrealspectrum,while tion.However,duetothemid-resolutionofthespectraR ≈ 2000, thecomputationtimeremainsreasonable(∼2d)usingastandard- thismethodisnotveryaccurate. size computer. The estimated error derived from the simulations (iv) ModifiedBahcallmethod. ∆(∆α/α) includes In Bahcall et al. (2004) the authors used a different approach sim tocomputethelinepositions.Theyperformedathird-orderspline ∆(∆α/α)2 =∆(∆α/α)2 +∆(∆α/α)2 +∆(∆α/α)2 , (6) interpolationtothestronger[Oiii]5008line,thenfittedthisinter- sim fit continuum code polation to the weaker 4960 line by adjusting the amplitude and where∆(∆α/α) istheerrorderivedfromtheGaussianfits,which fit separationoftheprofile.Wehavemodifiedthismethodbyusinga is our error estimate for each real spectrum; ∆(∆α/α) is continuum Gaussianfittothestrongerlineratherthanathird-orderspline. theerrorfromdifferentcontinuumsubtractionduetotheGaussian noise,and∆(∆α/α) isthesystematicerrorofourcode.Then, code Althoughwehavedescribedfourdifferentmethods,themain weexpect∆(∆α/α) >∆(∆α/α) andtheirdifferencewillbean sim fit resultsfor∆α/αpresentedinthisworkarebasedontheGaussian indicationofthecontinuumandsystematicerrors. fittingmethod,whiletheotherthreeareusedonlyforcomparison Fig.6(left-handpanel)showsthecorrelationbetweentheer- (seeSection4). rorin∆α/αfromtheGaussianfitsofeachrealspectrumandthe Finally,ourfinalresultfor∆α/αanditserrorisobtainedinthe standard deviation for ∆α/α of its 100 realizations. The standard samewayasinChandetal.(2005),namelywecomputeaweighted deviationsfromthesimulationsarewithinafactorof0.5−2ofthe meanandaweightedstandarddeviation,wheretheerrorsfor∆α/α standarderrorsfromthefitsfor97%(84%)ofthecaseswhenboth ofeachspectrumareusedasweights. quantitiesare< 5×10−3 (< 50×10−3).Thisshowsthatourcode (cid:13)c 2002RAS,MNRAS000,1–15 8 F.D.Albaretietal. ΛHÞLHobservedL 6000 7000 8000 9000 5 3-10DDΑΑ´H(cid:144)LHLsim 1234 DDΑΑH(cid:144)Lsim 0.00.0001.11101 ´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´ 0 0 1 2 3 4 5 0.2 0.4 0.6 0.8 1.0 DHDΑ(cid:144)ΑLfitH´10-3L Redshift Figure6.Left-handpanel:errorsfor∆α/αobtainedfromthesimulations(standarddeviationofthe∆α/αmeasurementson100realizationsofeachreal spectrum)andstandarderrorsfromtheGaussianfitsforourfiducialsample.Thesolidlinerepresentsaone-to-onecorrespondence,whilethedashedlines haveslopesof2and0.5.Onlythesimulationandfiterrorssmallerthan<5×10−3areshown.Right-handpanel:errorsestimatedfromthesimulationsasa functionofredshift.Spectrawith∆(∆α/α) >5×10−3areshownasredcrosses(24%ofthetotal).Thereisacleardivisionbetweentwodifferentsetof fit spectrawhichcorrelateswiththeskyemissionlines(seediscussioninthemaintext). 10 10 -3DΑΑ´H(cid:144)LHL10Voigt --10505 ´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´ -3DΑΑ´H(cid:144)LHLError10Voigt 02468 ´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´´ -10 -5 0 5 10 0 2 4 6 8 10 HDΑ(cid:144)ΑL H´10-3L ErrorHDΑ(cid:144)ΑL H´10-3L Gaussian Gaussian Figure7.Left-handpanel:measurementsof∆α/αusingGaussianandVoigtfittingprofiles.Non-compatiblemeasurementsat1σareshownasredcrosses (6.5%ofthetotal).Right-handpanel:errorsfromtheGaussianandVoigtfitting.Non-compatiblemeasurementsat1σareshownasredcrosses. andthecontinuumsubtractiondonotintroducenoticeablesystem- Asafurtherproof,wealsosimulaterealizationsofthe1416 aticerrorscomparedtotheGaussianfitting.However,thereisaset dropped spectra because of sky emission lines (criterion iv, see ofspectra(9%ofthetotal)forwhichthesimulationsprovidemuch Section 2). We found that more than 80% of the spectra have largererrors∆(∆α/α) > 0.1.Fig.6(right-handpanel)showsthe ∆(∆α/α) > 0.1.Thisconfirmsthatthesespectrahaveverylow sim errorsfromthesimulationsasafunctionofredshiftforourfidu- S/N and/or large pixels error due to the proximity of the lines to cialsample.RedcrossesstandforspectrawhoseGaussianfiterror strongskyemissionlines. ∆(∆α/α) > 5×10−3 (24%). The errors are distributed in two fit clouds of points. For the cloud with ∆(∆α/α)sim ∼ 1, the virtual 3.4 GaussianversusVoigtfittingprofiles realizationsofeachspectrumseemtodiffersignificantlyfromthe TheresultsobtainedwhenusingVoigtprofilesinsteadofGaussian realspectrum.Sinceweusetheerrorineachpixeltobuildthere- onesarecomparedinFig.7.TheVoigtandGaussianmeasurements alizations,therelativeerrorislargeforthesespectra,whichisan indicationofalowS/Nratioorlargeabsoluteerrorsinthepixels, are1σ-compatibleforthe93.5%ofthecases(98.3%at2σ).Re- gardingtheerrors,thereisnoclearimprovementwhenusingeither forinstanceinwavelengthregionswithskyemissionlines.Infact, ofbothmethods.However,Voigtprofileshaveonemoreparameter thecloudwithbiggererrorsmimicstheskyspectrum.Notealsothe andrestrictthenumberofdegreesoffreedom.Duetothespectral strongcorrelationbetweenthiscloudofpointsandthespectrawith mid-resolution and the fact that the [Oiii] lines are very narrow, largeGaussianfittingerrors(redcrosses).Theothersetofpoints with ∆(∆α/α) ∼ 10−3 are close to our error estimation on the thereareoftenonlyafewpixelstofit,whichfrequentlyleadtonon- sim measurementof∆α/αbasedontheGaussianfits. convergentfits.Thisreducesthequasarsamplein≈1000quasars. FurtherdiscussionaboutbothprofilescanbefoundinSection4. (cid:13)c 2002RAS,MNRAS000,1–15 Fine-structureconstantwithBOSS 9 4 SYSTEMATICS Table2.Resultsfor∆α/αconsideringseveralsampleswithdifferentcon- straints.Thenumberofquasarspectra,themeanandstandarddeviationof In thissection, weexamine thepossible unnoticedsystematic er- rorsbyanalysingdifferentquasarsamples.Table2summarizesall theredshiftandthevaluefor∆α/αareshown. thesamplesconsideredtogetherwiththeirmeanredshiftsandthe measuredvaluefor∆α/α. σ4960/5008−1 No.ofquasarspectra Redshift ∆α/α(×10−5) Weconsiderthefollowingsourcesofsystematicerrors. <50% 10028 0.56±0.21 1.6±2.3 <25% 8877 0.56±0.21 1.9±2.3 (i) Misidentificationofthelines.Theexpectedlinewidthsand <10% 5846 0.56±0.21 1.7±2.5 amplitudesareusefultoavoidmisidentificationofthe[Oiii]emis- <5% 3458 0.54±0.22 −0.9±3.0 sionlines.(a)Linewidths:sincebothlinesoriginateonthesame [Fλ×σ]5008/4960 No.ofquasarspectra Redshift ∆α/α(×10−5) upperenergylevel,theirwidthmustcoincide.Wecheckthatthis isthecasebyconsideringquasarswhose[Oiii]linewidthsarethe 2.98±0.50 8327 0.56±0.21 1.8±2.4 2.98±0.25 5761 0.55±0.21 −0.4±2.6 samewithinarelativefraction.Formorethanhalfofourfiducial sample,the[Oiii]linewidthsdifferbylessthan10%(seeTable2). 2.98±0.10 2658 0.54±0.21 0.0±3.4 2.98±0.05 1411 0.52±0.22 5.2±4.6 (b)Amplituderatio:atomicphysicsstatesthattheamplituderatio between the [Oiii] 5008 and [Oiii] 4960 lines is 2.98 (Storey & Fitwidth No.ofquasarspectra Redshift ∆α/α(×10−5) Zeippen2000)(asquotedinSection5,weobtain2.96±0.02syst). 2σ 10363 0.56±0.21 1.4±2.3 Thus,weconsiderdifferentsampleswherethisratiodiffersbyless 3σ 10252 0.59±0.20 5.5±2.5 thanacertainamountfrom2.98(seeTable2).Allthesamplescon- 4σ 9978 0.59±0.20 7.1±2.7 sidered in this test yield results for ∆α/α compatible with zero. 5σ 9726 0.59±0.20 5.3±2.6 Fig.8displaystheGaussianwidthsandfluxesofboth[Oiii]emis- S/NHβ/[Oiii]4960 No.ofquasarspectra Redshift ∆α/α(×10−5) sionlinesforourfiducialsample. (ii) WindowsfortheGaussianfits.Weuseawavelengthrange <5 10338 0.57±0.21 1.4±2.3 of 2σ around each [Oiii] line in order to obtain the final Gaus- <2 9831 0.57±0.21 0.6±2.3 <1 8162 0.57±0.21 0.1±2.5 sian fit to the line profiles. We study how our results depend on <0.5 5831 0.58±0.21 −0.7±2.8 thischoice.Byconsideringalargerwavelengthinterval,theresults aremoreaffectedbytheHβcontaminationandpossibleasymme- Pol.order(cont.) No.ofquasarspectra Redshift ∆α/α(×10−5) triesonthelinewings.Thedifferencesinthenumberofspectrafor 3 10528 0.57±0.21 1.0±2.3 thesesamples[whichareobtainedbyapplyingtheselectioncrite- 5 10550 0.57±0.21 1.3±2.3 ria (i)−(iv) discussed in Section 2.1] arise because of the criteria 7 10363 0.56±0.21 1.4±2.3 concerningthenon-convergingfitsandtheskyemissionlinesde- 9 10471 0.56±0.21 −1.1±2.3 scribedinSection2. R2(bothfits) No.ofquasarspectra Redshift ∆α/α(×10−5) (iii) Hβcontamination.Weanalysesampleswheretheratiobe- tthweevenalSue/NfoHrβ∆aαnd/αS/dNec[Oreaiisi]es49a6s0wisecpolnascteraminoerde.sDtreinspgietnetthcoenfsatcrtatihnatst >>00..997 69024554 00..5566±±00..2211 12..58±±22..47 >0.99 2301 0.54±0.21 2.0±3.5 onHβ,itisalwaysconsistentwithnovariationinαwithintheer- >0.995 845 0.51±0.22 −0.4±4.8 rors.Thisanalysisdemonstratesthatthestrengthand/orwidthof theHβlinedonotaffectsubstantiallytheresultfor∆α/αwhena [Oiii]5008(kms−1) No.ofquasarspectra Redshift ∆α/α(×10−5) weightedmeanisadopted. <1000 10353 0.56±0.21 1.4±2.3 (iv) Continuumsubtraction.Weuseaseventh-orderpolynomial <500 8990 0.56±0.21 0.2±2.4 tosubtractthecontinuumspectrum.Weexamineifthepolynomial <300 2798 0.52±0.22 −6.8±3.9 order has important effects on our measurements. Our values for <200 150 0.52±0.24 21±18 ∆α/αandtheirerrorsareonlyslightlyaffectedbythechosenpoly- Method No.ofquasarspectra Redshift ∆α/α(×10−5) nomialorder. (v) Goodness of Gaussian fits. We quantify the quality of the Gaussian(weighted) 4537 0.58±0.20 −0.4±2.8 GaussianfitsbytheR2coefficient.Alltheconsideredsamplesshow Gaussian 4537 0.58±0.20 1.2±4.5 valuesfor∆α/αconsistentwithnovariationinα. Integration 4537 0.58±0.20 3.6±4.8 ModifiedBahcall 4537 0.58±0.20 0.8±4.4 (vi) Broadlines.Wealsostudysampleswherethewidthofboth Median 4537 0.58±0.20 1.8±1.4 lines is less than a certain value (in kms−1). These samples are consistentwithnovariationofα.Samplesbuiltfromnarrowlines GaussversusVoigt No.ofquasarspectra Redshift ∆α/α(×10−5) < 300kms−1 maybemoreaffectedbymisidentificationof[Oiii] Gaussianprofiles 8485 0.55±0.19 0.4±2.5 linesasskylines. Voigtprofiles 8485 0.55±0.19 −1.1±2.8 (vii) Different methods for measuring the [Oiii] line position. Mixedprofiles 8485 0.55±0.19 1.3±2.4 We compare the results obtained by the methods to measure the position of the [Oiii] lines described in Section 3.2.4. Since not allthemethodsprovideanerrorforthemeasurement,wecannot (viii) Gaussian versus Voigt profiles. We compare the results calculateaweightedmean,anditisnecessarytoselectamorere- for 8485 quasars from our fiducial sample after dropping 1878 strictedsample.Then,weconsiderasamplewherethedifference spectra with non-converging Voigt fits (this reduction increases betweenthewidthsofthelinesislessthan25%,theamplitudera- the statistical error). We also compute a ‘mixed’ value for ∆α/α tioisconstrainedtodifferfromthetheoreticalvalue2.98(Storey& where for each spectrum we use the value for the variation of Zeippen2000)bylessthan0.5,andtheS/N issmallerthanhalf thefine-structureconstantwithsmallererror,either(∆α/α) or Hβ Gauss theS/N[Oiii]4960. (∆α/α)Voigt. (cid:13)c 2002RAS,MNRAS000,1–15 10 F.D.Albaretietal. 1500 L -1 s 1000 ms(cid:144)L 1000 -2cm 500 kH 700 rg 0 e II496D 500 -17H10 15000 OI 60 @ 9 M 300 4 WH DIII 10 O F 200 @ 5 x u 150 Fl 150 200 300 500 700 1000 1500 10 50 100 500 1000 5000 FWHM@OIIID5008Hkm(cid:144)sL Flux@OIIID5008H10-17ergcm-2s-1L Figure8.Left-handpanel:Gaussianwidths(inkms−1)forboth[Oiii]lines.Bothlinesoriginateonthesameupperenergylevel,thentheirwidthsmust coincide(reddashedline).Right-handpanel:fluxesforboth[Oiii]lines.Thetheoreticalfluxratiois2.98(reddashedline).Theentirefiducialquasarsample isshown. Table3.DetailedinformationaboutthebinsinFig.10. Redshiftinterval No.ofquasarspectra Redshift ∆α/α(×10−5) 0.390−0.460 817 0.42±0.02 −5.2±6.8 0.460−0.520 723 0.49±0.02 5.5±8.9 0.520−0.580 757 0.55±0.02 0.4±9.2 0.580−0.625 843 0.60±0.01 40.4±9.4 0.625−0.675 988 0.65±0.01 −3.5±7.4 0.675−0.715 1299 0.69±0.01 −8.2±7.4 0.715−0.765 1117 0.74±0.01 1.7±7.1 0.765−0.820 1444 0.79±0.02 18.1±8.3 0.820−0.880 644 0.84±0.02 4.7±9.2 0.880−1.000 580 0.93±0.03 17.0±13.3 Thisvalueisconsistentwiththepreviousresultsreportedindiffer- entinvestigationsbasedonthesamemethod:Bahcalletal.(2004), Figure9.Errorsfor∆α/αderivedfromtheGaussianfits(greypoints)for Gutie´rrez&Lo´pez-Corredoira(2010),andRahmanietal.(2014). ourfiducialsample,movingmeanofthetheseerrors(blueline)usingover- TheredshiftdependenceofthemeasurementsisshowninFig.10 lappingbins(100spectraperbin,∆z≈0.025),movingstandarddeviation (left-hand panel), where several bins have been made taking into of ∆α/α measurements using the same bins (red line) and a typical sky accounttheredshiftintervalsaffectedbythesky(shadedzones).In spectrum. theright-handpanel,weshowtheresultsobtainedfromthesimu- lationsdescribedinSection3,usingthesameredshiftsintervalsfor thebins.Themaindifferencesbetweentherealresultsandthesim- We have also analysed the standard deviation and errors of theresultsfor∆α/αasafunctionofredshift(Fig.9).Eventhough ulationsareintheregionswheretherearestrongskylines(shaded regions),whilebeinginagreementintheremainingzones.Detailed we have imposed a constraint on our initial sample based on the informationabouteachbinfortherealdatacanbefoundinTable3. skyemissionlines,thestandarddeviationanderrorsstillcorrelate Our results are little affected by the specific constraints im- withthesky.Inparticular,forthecorrelationwiththemovingstan- posed in our sample as discussed in Section 4. For instance, we darddeviation,thismeansthattheprecisioninourmeasurementof ∆α/αalongthewholeredshiftintervalislimitedbytheskysub- varythewidthfortheGaussianfits,thecontaminationofHβ,the polynomialorderusedtofitthecontinuumspectrum,thequalityof tractionalgorithm. theGaussianfitsandtestdifferentmethodstomeasure∆α/α.The mostimportanteffectfoundisthatbyconsideringbroaderwidths 5 RESULTS fortheGaussianfits,theresultsaremoreaffectedbythecontamina- 5.1 [Oiii]lines tionfromHβandpossibleasymmetriesinthelinewings.Wehave alsocheckedforpossiblemisidentificationsofthe[Oiii]emission We used a total of 10363 quasar spectra, drawn from the SDSS- linesusingtheirexpectedwidthsandamplituderatio. III/BOSS DR12Q catalogue, after applying the selection criteria Table4containstheresultsfor∆α/αwhenthelowerbound (i)−(iv) (see Section 2), to measure the possible variation of the on the S/N[Oiii]5008 is increased. All the results remain consis- fine-structureconstant.Thefollowingmeasurementisobtained: tent with no variation of the fine-structure constant. In Fig. 11, ∆α the measured ∆α/α for our fiducial sample as a function of the α =(1.4±2.3)×10−5. S/N[Oiii]5008areplottedtogetherwiththeirerrors. (cid:13)c 2002RAS,MNRAS000,1–15