1 98lAp J . . .247 9K calibrations ofSandageandTammann(1976),deVau- The AstrophysicalJournal,247:9-16,1981July1 of theextragalacticdistancescale.Thesemi-independent couleurs andBollinger(1979),Aaronsonetal(1980), © 1981.TheAmericanAstronomicalSociety.Allrightsreserved.PrintedinU.S.A. Branch (1977,1979)yieldmeanvaluesfortheHubble constant whichdifferfromeachotherbyasmucha may beresolvedbycalibratingyetothertypesofextraga- factor of2.Whileitispossiblethatthiscurrentimpasse lactic distanceindicators,moremaybelearnedthrougha critical reanalysisofthecurrentdistanceladders,sup- paper isthelastinaseriesaimedatevaluatingand ported bynewsetsofimprovedobservationaldata.This Tammann (hereafterS-T)ladder,thediameters is operatedbytheAssociationofUniversities forResearchinAstron- recalibrating oneoftheweakeststepsinSandageand b c;hereafterPapersI,II,andIIIrespectively),itwas and luminositiesofthelargestHnregionsingalaxies. diameters andfluxesforthefirst-rankedHnregions. earlier S-Tdistances(S-T1974c),butthissmallchangeis Adoption ofthenewdatasuggesteda15%decreasein substantially bymeasuringphysicallybasedphotometric shown thattheS-Tdistancescalecouldbeimproved omy, Inc.,undercontractwiththeNational ScienceFoundation. dominated byalargeuncertaintyintheHnregion because onlyonegiantScspiral,M101,possessesan distances tobrightspirals,±25%ormore,introduced Vaucouleurs 1978;andMould,Aaronson,Huchra independently determineddistance.Theuncertaintyin- distance ofM101(see,forexample,S-T1974fr;de troduced intothedistancescalebycontroversial Ha bandpassphotographsof M101andNGC3631, largest HnregionsinM101.Figure1(Plate1)displays 9 1980) isexacerbatedbytheextraordinarycharacterof 1 Great controversycurrentlysurroundsthecalibration In thepreviouspapersofthisseries(Kennicutt1979a, VisitingAstronomer,KittPeakNational Observatory.KittPeak © American Astronomical Society • Provided by theNASA Astrophysics Data System Virgo clusterof31?4±0^25,basedontheSandageandTammannlocaldistancescale,but independent oftheirdistancetotheM101group.TheVirgoHnregionsappearpossessaverylarge data tothecalibrationforLocalandM81groupgalaxiesyieldsatruedistancemodulus representative offieldgalaxies,distancesderivedtoindividualgalaxiesusingHnregionswillbe dispersion inHnregionproperties,evenamonggalaxiesofthesametype.Observationalerrorsand selection effectscannotaccountforthisscatter,andwearguethatitisreal.Ifthedispersion Subject headings:cosmology—galaxiesclustersofnebulaegeneral subject tolargerandomerrors,andothermethodsshouldbeusedinstead. We presentHaphotometryforthebrightestHnregionsin21VirgoclusterScgalaxies.Fittingthese I. INTRODUCTION H iiREGIONSASEXTRAGALACTICDISTANCEINDICATORS. Mount WilsonandLasCampanasObservatories,CarnegieInstitutionofWashington Received 1980October20;accepted1981January16 IV. THEVIRGOCLUSTER 1 Robert C.Kennicutt,Jr. ABSTRACT galaxy. SincetheHnregiondistancesarebasedon another giantSc.ThelargestfiveHnregioncomplexesin their luminosityrelativetotheotherHnregionsin M101 areuniqueintheircomplexityofstructureand measurements ofthethreelargestobjects,peculiarity any distancescalebasedonthem.Figure2,takenfrom of theM101complexesseverelydamagesintegrity plotted asafunctionofDDOluminosityclass.Again,the polations ofthedataforfaintergalaxies.Aprioriwehave regions intheLocal,M81,andM101groupgalaxies, the diametersandluminositiesofcalibratingHn Papers IIandIII,furtherillustratesthispoint.Shownare nonlinearity, sharedbyallgiantScgalaxies,orwhetherit no wayofjudging,however,whetherthisreflectsareal point isacrucialonebecauseitthegiantScI-ScII means theHnregionsinM101areabnormallylarge.The M101 HIIregionsdeviatesignificantlyfromlinearextra- constant intheS-Tscheme.chosetoignore linear extrapolationoftheHnregiondataforother, systems whosedistancesultimatelydeterminetheHubble range ofcalibrations. fainter systems.Earlierpapersinthisseriesconsidereda M101 data;theirdistancestogiantspiralsarebasedona describes anattempttosolvetheproblembydirectly measuring theshapeofHnregioncalibrationover obtaining Haphotographicandphotometricmaterialfor the fullrangeofgalaxyluminosity.Thiswasdoneby 20 ScgalaxiesintheVirgocluster(§§II,III,andIV).The data withthoseavailablefornearby galaxies(e.g.,Fig.2), galaxies studiedspanarangeinluminosityfromgiant systems ofluminosityclassIII-IV.Bycombiningthese cluster isobtained,againindependent ofthedistanceto the M101group(§VI). M101. Alongtheway,animproved distancetotheVirgo it ispossibletocalibratethe H nregionsindependentof Sc IsystemsanalogoustoM101,intermediate-dwarf This situationisclearlyunsatisfactory,andthispaper 1 98lAp J . . .247 9K 1 io Virgo cluster.Thespecificrequirementswere: in §VII. considerably larger.Theimplicationsofthislargedisper- problem hasbeenbasedonabiasedsampleofgalaxies. function ofDDOluminosityclass.M101isthepointatclassI. sion foradistancescalebasedonHnregionsisdiscussed H iiregionpropertiesinthismorecompletesampleis this study.Allofthepublishedworktodateon 4536), whichareprobablemembers,wereincluded,but cluster (deVaucouleurs1961).Iftherangeofpositionin selected onthebasisoftheirprobablemembershipin faint Hiiregionswerenotincluded,inordertoguarantee Resolution ofHnregionswasarequirementforinclu- cluster. TwoadditionalScIgalaxies(NGC4303and the lineofsightiscomparabletothisradius,mostgalaxies Virgo clustersample,however,theselectioncriteriaare sion intheS-Tsamples.Galaxieswithunresolved,very km s". they lieslightlyoutsidethe6° circle. should liewithin±10%ofthemeandistanceto truly distance-limitedsample(S-T1974c).Inthepresent that thedistancescalederivedwouldbebasedupona before anythingwasknownofthepropertiestheirHn on thebasisoftheirlocationinskyandvelocity, completely different.Galaxiesforthisstudywerechosen much morerepresentativecrosssectionofScgalaxies regions. Consequently,wehaveintheVirgosamplea region populations.Notsurprisingly,thedispersionof and, wehope,amuchmorerepresentativerangeofHn Sandage system(S-T1974c, 1980). There isonemoreimportantreasonforundertaking As describedabove,galaxiesforthisstudywere Fig. 2—Isophotaldiameters(left)andHaluminosities(right)forthemeanofthreelargestHnregionsincalibratinggalaxies,plottedasa 2. Allpossessgalactocentric velocities lessthan2500 3. AllareclassifiedasScor laterontheHubble- 1. Thegalaxiesliewithina6°radiusofthecenter © American Astronomical Society • Provided by theNASA Astrophysics Data System v II. SELECTIONOFTHESAMPLE 1 500 300 Q 400 800 700 600 200 100 0 _l 1I n ir x LUMINOSITY CLASS of V IV O M8I x M101 • LOCAL / O / • / * ^ J L KENNICUTT m -1 Three referencesourceswereavailable,vandenBergh Galaxies whoseassignedtypesdifferedbymorethana classes agreedwitheachothertowithinhalfofaclass. full classwerethrownoutofthesample. avoid, asmuchpossible,systematicerrorsinthe of 60°orlesstotheplanesky.Thiswasdone lacking intheLocalGroup(Fig.2)sample. classifications andtolimittheeffectsofinterstellar (1960) andS-T(1974c,1980).Generally,theluminosity absorption ontheHaphotometry. procedure wasidenticaltothatdescribedinPaperI.The on theKittPeak2.1mtelescope.Theobservational each luminosityclassfromScItoIII.Essentiallyno according tothecriterialistedabove,wemadenoattempt of 21galaxies,listedinTable1.Whilethelistiscomplete data wereobtainedfordwarfswithluminosityclassIV-V goal wastoselectanadequatenumberofspiralswithin to defineatrulymagnitude-limitedsample.Themain permitting exposuresof30-45 withnegligibleskyfog. interference filterslentbyDr. PaulHodgewereused, only changewasadifferentimage tubewhichproduceda for galaxiesofM33typeandbrighter,justtherange for theHnregionversusgalaxyluminositycalibration difficult, andMalmquist-typeselectioneffectswouldbe axies attheVirgoclusterdistancewouldbeextremely slightly highergainthanbefore. Asetofnarrow,20 severe. Instead,ourgoalistoobtainanapproximateform slightly largerplatescale(13.0 ±0'.'2mm)anda (i.e., M>—18).DetectionofHnregionsinsuchgal- 1979 Aprilwithatwo-stageCarnegieimage-tubecamera b 1 U 39 4. AllpossessluminosityclassesontheDDOsystem. We obtainedHa-bandpassimage-tubephotographsin These requirementseliminatedallbutacandidatelist 5. Thesamplewaslimitedtogalaxieswithinclinations o o* C5 40- 41 38 37 III. OBSERVATIONSANDDATAREDUCTION i nr j ii_ LUMINOSITY CLASS V IVIIIIII O M8I x MIOI • LOCAL Vol. 247 1 98lAp J . . .247 9K 2 No. 1,1981 identify foregroundstars,backgroundgalaxies,andstel- lar associationsontheHaplates.AportableKPNO Continuum exposureswerealsoobtainedinorderto portion ofeachimage-tubeplate.Digitalmicrophotome- process maybefoundelsewhere(Kennicutt1978,1979a). aid ofthePDSmicrophotometeratKittPeakand tric mapswerethenproducedforeachHaplatewiththe sensitometer wasusedtorecordcalibratedspotsona galaxies. ExamplesofafewareshowninFigure3(Plate software developedatKPNO.Detailsofthecalibration photometry wasobtainedforaregionof2arcminin This representsachangeintechniquefromearlierwork. each ofthe21programgalaxieswithaSITVidicon on theplates,photoelectricdatawerealsoobtained.Ha the duPont2.5mtelescopeofLasCampanasObserva- Photographs wereobtainedfor18ofthe21candidate The extremefaintnessandsmallsizesoftheHnregions camera onthePalomarObservatory1.5mtelescope. tory witha90mmITTimagetube. galaxies, madethepreviousmethodofaperturepho- in thesedistantspirals,especiallythosethefainter 2). Inaddition,aplateofNGC4536wasobtainedon tometry impracticalhere. pixels isused,atascaleof0.55sperpixel.Thecamera Kent (1979).Inthedirectmode,aformatof256x photometric nightsin1979March,April,andJune.The pixel-to-pixel variationsinboththesensitivity(“flat minutes forthephotometryinthisprogram.Inmode, operated inanintegratingmode,normallyfor10-20 positions werechosenwhichgenerallyincludedtwoor would becountedinthreeofmoreHnregions.Observa- exposure timesselectedsothatatleast10,000photons more ofthebrightestHnregionsinSITfield,with One ortwoframeswereobtainedforeachgalaxy; same Hafilterswereemployedasforthephotography. field ”)andthebackground(“eraselevel.Aftercalibrat- the accuracyofphotometrywaslimitedto in thephotometryofstandardstarsaverage±3%rms, ing individualexposureframesfortheseeffects,residuals photometry, includingcorrectionsforfiltertransmission, tions ofstandardstars(PaperI)werealsomadeand certainly adequateforthisapplication. the extinction.Detailsofabsolutecalibration served tocalibrateboththeinstrumentalzeropointand found inPaperIandKennicutt(1978). stellar Ha,nebular[Nn],andcontinuumcanbe providing forthecalibrationofindividualframes,rou- developed byBillSebokatCaltech.Inadditionto aid ofaninteractiveimage-processingsoftwaresystem H iiregionsonthestoredSITframeswasdonewith within circularaperturesof arbitrary size.Theselatter because ofthehighlystructured backgroundsurrounding capabilities wereessentialfor the Hnregionphotometry around eachindividualHn region andmeasurefluxes tines madeitpossibletodeterminebackgroundlevels many oftheobjectsinterest. Fluxesweremeasuredfor In ordertoobtainabsolutefluxesandisophotallevels Photometry oftheprogramspiralswasdoneonthree The SITcamerasystemhasbeendescribedindetailby The reductionofaperturephotometryindividual © American Astronomical Society • Provided by theNASA Astrophysics Data System H iiREGIONSASDISTANCEINDICATORS 2 ranging from8-17"indiameter,dependingonthesizeof three tosixHnregionsoneachframe,inapertures calibrated image-tubeplateswasthenobtainedbyinte- the individualobject.Themicrophotometricscaleof grating overanidenticalareaonthedigitalPDSmaps. we mightbeabletomeasureaccurateisophotaldiameters for thelargestthreeHnregionsineachVirgogalaxy. program galaxies(e.g.,NGC4303and4254),the Except forthelargestHnregionsinafewofbrightest Figure 3(Plate2)presentsreproductionsofsixthe Ha platesanddemonstrateswhythisisnotpossible. only slightlylargerthantheseeingdisk.Insomeof H iiregionsareonlymoderatelyresolved,withdiameters the faintestgalaxies,Hnregionsareessentially deconvolution techniques(Kennicutt1978,1979a).But, due toseeingweresuccessfullyremovedwithsimple stellar. Inthepreviouswork,smallamountsofsmearing in thoseinstances,theHnregionswereconsiderably was onlyasmallone.Inthepresentcircumstances, larger thantheseeingdisk;deconvolutioncorrection ience, wefeelthattheuncertaintiesinstellarphoto- the correctionswouldbelarge,and,basedonourexper- useless. Instead,weconcentrateonthemuchmoreeasily graphic profileswouldrenderanyresultsvirtually measured luminositiesofthelargestHnregions. were obtainedunderconditionsofonlymoderate>2" photographs obtainedformanyoftheseobjectsunder lent seeing,perhapswithalineararraydetector? ment touseasadistanceindicator.Onedrawbackofthe largest Hnregionsareasfeasiblediametermeasure- conditions ofexcellentseeinginChile,andtheemission seeing. Couldaccuratediametersbemeasuredinexcel- luminosities istheirlargeintrinsicdispersion(relativeto knots arestillonlymarginallyresolvedinmanyofthe Probably not.Wehavebeenabletoexamineblue their muchstronger1/rdistancedependence.When cal onetosolvefromtheground.TheSpaceTelescope,of fainter objects.Thustheproblemis,atbest,animpracti- calibrated usingthesameproceduresasfordiameters, the diameters),butthisismorethancompensatedforby discourage anysucheffort. course, couldeasilyresolvetheHnregions,but hence wewillpresentouranalysis intermsofthefluxes. diameters. Theresolutionofthepresentdatasetis the Hafluxesyieldeddistanceswhichagreedtowithin results reportedinthefollowingsectionsmaywell adequate formeasuringaccurate Hnregionfluxes,and three objectsineachgalaxy wereselectedbyvisual the sameasthatdescribedin PaperIII.Thebrightest ±15% rmswiththosedeterminedfromisophotal At theoutsetofthisinvestigation,wehadhopedthat We shouldnotethattheHaplatesusedinthisstudy In PaperIIIitwasshownthatHaluminositiesofthe The procedureforreducingthe Hnregionfluxeswas b) LuminositiesoftheLargestHnRegions a) DiametersoftheLargestHnRegions IV. THEDATA 11 1 98lAp J . . .247 9K inspection. First-orderfluxeswerethenobtaineddirectly necessary, asmall“aperture”correction,alwayslessthan much deeper,digitalPDSphotographicmaps,and,if field. TheHnregionboundarieswerethentracedonthe field, thephotographicfluxwasconvertedtoabsolute 20%, wasapplied.ForHnregionslyingoutsideaSIT from theSITphotometry,ifanHnregionlayin units usingthemeanSIT-platecalibration.Inadditionto were measuredwiththeSIT,NGC4152and4178.The the 19galaxiesphotographed,twoadditionalobjects background levelaroundeachHnregion.Comparison individual Hnregionsissomewhatlargerbecauseofthe the faintergalaxies,uncertaintiesinphotometryof from theVidiconframes. H ilregionfluxesinthesegalaxiesweremeasureddirectly lower fluxlevelsandthedifficultyinassigninganaccurate The Hafluxquotedisthemeanoflargestthree magnitudes andluminosityclassesoftheparentgalaxies. of photographicandphotoelectricfluxesrepeat applied forcontaminationofthephotometrybyback- measurements ofthesameobjectsindicatethatflux ground continuumand[Nn]A/16648,6684inthewingsof H iiregions.Ineachcase,astatisticalcorrectionhasbeen 12 rms forthebrightestregions,to±25%faintest, any individualHnregionhasanuncertaintyof±10% photometry itself,allcontributetotheoveralluncertainty the filterbandpasses.AsdescribedinPaperIII,uncer- at aboutthe10%level.Inpresentsample,especiallyin tainties inthesecorrections,aswellthephotoelectric from thelattereffectsalone.Theoveralluncertaintiesare shown inTable1. 4152 II12.14-13.26±0.07 4303 I9.86-12.27±0.05 4254 I10.11-12.46±0.05 4536 I10.44-13.10±0.07 4535 I10.12-12.97±0.06 4321 I9.79-12.98±0.07 4689 II-III11.22-13.60±0.07 4568 II-III11.13-13.11±0.07 4212 II-III11.44-13.06±0.08 4654 II10.68-12.88±0.05 4189 II12.20-13.06±0.05 4178.... I11.26-12.94±0.07 4237 Ill12.11-13.82±0.08 4571 II-III11.49-13.82±0.09 4567 II-III11.67-13.40±0.07 4595. Ill12.79-13.48±0.07 4540 III-IV 12.51-13.72±0.08 4299 III-IV 12.56-12.67±0.05 4298 Ill11.61-13.54±0.07 4294 HI12.02-13.02±0.06 4561 IV 12.61-13.12±0.06 a b-2_1 Table 1liststhederivedfluxes,alongwithcorrected FromS-T1980. ErgscmsinHa. © American Astronomical Society • Provided by theNASA Astrophysics Data System 0,iab NGC LCBlog T3 H iiRegionData TABLE l KENNICUTT regions plottedasafunctionofapparentmagnitudeforVirgoSc galaxies. fluxes plottedasafunctionofgalaxianluminosityclass. the largestHnregionsinVirgoversusapparent The expectedtrendofincreasingHiiregionluminosity magnitude oftheparentgalaxy.Figure5showssame with increasinggalaxyluminosityisapparent,butavery The spreadisreal;thefluxesplottedinFigures4and5are served is±0.4dexrms.Figure3(Plate2)illustratesthe large dispersionaboutthemeanrelationsisalsoevident. accurate to±0.05-0.10dex,whilethedispersionob- luminosity class,followingSandage and Tammann(1980). phenomenon. properties isprobablythemostsignificantresultofthis general populationoflate-typespirals,itimpliesthat distances derivedtoindividualfieldgalaxiesusingthis study because,ifthisVirgosampleisrepresentativeofthe method willneverbemoreaccuratethan~±50%at -21 Figure 4displaysaplotofthemeanapparentHaflux Fig. 4.—Meanflux(ergscmsinHa)ofthreelargestHn Fig. 5.—Sameas4,exceptplotted asafunctionofDDO The discoveryofsuchalargedispersioninHnregion A v o o* IO a s: nzniii a) TheHiiRegionDispersion V. ANALYSIS LUMINOSITY CLASS Vol. 247 No. 1, 1981 H il REGIONS AS DISTANCE INDICATORS 13 best. This result would effectively render the technique 4. The cluster environment in Virgo causes the disper- useless for most applications, since far less elaborate sion. This possibility is a difficult one to evaluate, though methods may be used to derive distances at the 50% level our own view is rather skeptical. Despite the considerable (van den Bergh 1980). It is important, therefore, to try to literature on the subject of environmental effects on the understand the cause of the dispersion and to eliminate evolution of cluster spirals, there appears to be no any possible selection effects which might introduce an consensus at all as to whether the cluster environment is artificial scatter in Figures 4 and 5. important, and certainly none regarding the Virgo cluster. Furthermore, even if evolutionary effects were present in Virgo, it is unclear how they might alter the b) The Cause of the Dispersion properties of the large H n regions. Our own data on the Before discussing the factors which might account for a total ionized gas contents of the Virgo spirals (Kennicutt real dispersion in the properties of the largest H n regions 1981, in preparation) suggests that they do not differ in galaxies, let us discuss those effects which might significantly from field galaxies. Thus while we cannot introduce a superficial scatter. completely dismiss this explanation, we feel that other 1. Errors in the photometry or in the luminosity classes. causes may be more important. Those are described As discussed above, the very large dispersions in Figures 4 below. and 5 are too large to be accounted for by errors either in 5. A dispersion in morphological type. All of the gal- the Ha or the galaxy photometry. While errors in as- axies in this study are classified as Sc on a homogeneous signed luminosity classes could produce a more serious system (ST 1980). If, however, the luminosities and sizes scatter, the equality of the dispersion in Ha flux in both of H ii regions are very sensitive functions of morphologi- the luminosity class (± 0.37 dex) and the apparent mag- cal type, small “ errors ” in assigned Hubble type might nitude plots (±0.38 dex) seems to argue against this introduce considerable scatter into the H n region cali- interpretation. As aptly noted by the referee, however, bration. The limited available data suggests that this stranger things have been known to happen. indeed may be the case. Sérsic (1960) measured diameters 2. Many of the galaxies in the sample are either fore- by eye for the largest H n regions in 65 nearby spirals and ground or background objects. This is not a likely cause for irregulars, using large-scale 5.0 m and 2.5 m telescope at least two reasons. First, the area of the Virgo cluster plates. Table 2 lists the mean diameter as a function of the surveyed was intentionally limited to the 6° circle in order Hubble type of the parent galaxy. The absolute values to minimize this problem. Of the 21 galaxies surveyed, differ from those in Sérsic’s own table; we have adopted a 1 1 only one (NGC 4178), to our knowledge, has been different distance scale (H0 = 50 km s~ Mpc" ), and we suggested as a possible foreground object (de Vau- have restricted the sample to giant spirals (luminosity couleurs 1961). In order to further check this point, an classes I-II) only. While systematic errors in the absolute inspection was made of large-scale photographs of as measurements may be large (Paper II), there is a clear many of the galaxies as were available, in order to see trend in the relative diameters; early spirals possess whether the resolution characteristics of the images of smaller H n regions. As might be expected, the gradient in any belied their membership. No obvious nonmembers H ii region luminosity is even steeper. Preliminary Ha were isolated in this way, though two galaxies, NGC 4178 photometry was obtained for the largest three H n re- and NGC 4152, appeared somewhat abnormal for their gions in six Sb galaxies, using the SIT detector on the assigned luminosity classes (Sc II). If these two galaxies Palomar 1.5 m telescope. Table 2 shows that the mean are removed from the sample, however, the rms scatter in flux in the Sb galaxies is roughly a factor of 6 lower than Figure 5 actually increases! Secondly, even if a few in a large sample of Sc galaxies drawn from this paper nonmembers were present in the sample, they would not and from Paper III. appreciably alter the dispersion in Figure 4. The H n It therefore seems plausible to argue that, if there were a region-galaxy magnitude relation is nearly degenerate sprinkling of Sbc galaxies present among the Virgo Sc with a distance line ; an error in the presumed distance of a galaxies, for example, the dispersion in type could easily galaxy would simply move it along a vector nearly introduce a dispersion in the H n region calibration of a parallel to the mean relation, leaving its deviation factor of 2-3. However, there is only circumstantial unchanged. evidence to support the argument. Those galaxies whose 3. Ha fluxes are unreliable. H il region diameters must H ii regions lie above the mean relation in Figure 4 are a T be used. While accurate isophotal diameters could not be bit bluer than those below [(B — V)0 = 0T56 versus measured, an inspection of the plates shows that those H ii regions which are anomalously bright/faint for their TABLE 2 galaxy type also appear to be anomalously large/small (see Fig. 3). In a more local sample, where both diameters H ii Region Properties versus Galaxy Type and fluxes could be measured (Papers II, III), the fluxes Sérsic Diameter log
© American Astronomical Society • Provided by the NASA Astrophysics Data System 1 98lAp J . . .247 9K ouleurs’ system(=5.7versus4.7,deVaucouleurs, 0T61, deVaucouleurs,andCorwin1976] 14 Vaucouleurs, andCorwin),butthedifferencesaresmall, ical innature,especiallythecaseofclassifications. and areclassifiedsystematicallylateronthedeVauc- and thecorrelationstosomedegreeareprobablytautolog- expect toencounterelsewheresincetheyarebasedpri- classifications usedhere,infact(S-T1980),areprobably closer to“perfect”forourpurposesthanmostwewould system willnotreducethedispersiondrastically.The In otherwords,evena“perfect”galaxyclassification which shouldcorrelatestronglywiththeHnregion marily onthecriterionofdiskresolution,aparameter population. Thus,inadditiontoapossibledispersion dispersion inHnregionpropertiesispresent. galaxy type,wesuspectthatanadditionalstochastic capable ofintroducingsomedispersionintothecalibra- present sample.Weareforcedtoconclude,therefore,that accounting fortheverylargescatterobservedin the factorsdiscussedabove,especiallylast,seem indicators. tion ofthelargestHnregions,noneseemcapable regions whichrendersthemunsuitableasdistance dispersion inthesizescalesoflargestgiantHn the scatterisreal,thattherepresentanintrinsic that theobserveddispersionispurelyastatisticalartifact, diameters ofthelargestthreeHnregionsinamuch arising becauseweonlymeasuretheluminositiesor diameter functionsoftheHnregionsinindividual larger population.Theexpectederrorfromsampling galaxies andthenapplyingbinomialsamplingstatistics alone canbederivedbymeasuringtheluminosityand diameter functionsintwogalaxies,M33andNGC628 in themeandiameteroflargestthreeHnregionswas largest rankedobjects.Thiswasdoneforobserved to derivetheexpecteddispersioninpropertiesof in thepresentsamplingofdifferentgalaxies. higher, oforder±20%,butagainfarlessthanisobserved of order±10%;thedispersioninfluxesisconsiderably function ofHnregionsdoesnotitselfvary.Inspection above onlyappliestoagroupofgalaxiesiftheluminosity tion ofsubluminousHnregions.Whetherthemean case. GalaxieswhoselargestHnregionsareanoma- tion oftwoeffects.First,thestatisticalanalysisdescribed (Kennicutt 1978).Inbothcases,theexpecteddispersion the Haphotographssuggeststhatthisissimplynot NGC 2366,describedbyKennicutt, Balick,andHeck- is NGC4299,showninFig. 3.Localexamplesinclude in agalaxy,mostoftenthegalaxiesofintermediateor in Figures4and5.Thesecondeffectistheoccasional measured, suchgalaxieswilloccupyadiscrepantposition lously faintusuallyappeartopossessanentirepopula- low luminosity.(Thebestexample inthepresentsample appearance ofanextraordinarilylargenebularcomplex the brightestthreeor30Hnregionsis man 1980,orNGC5471inM101.) Itseemsquiteclear that thedistributionofsizein the verylargestHnregions 6. Hiiregionsarenotstandardcandles.Whileafewof Before elaborating,itisimportanttodismissthenotion What thencausesthescatter?Wesuspectacombina- © American Astronomical Society • Provided by theNASA Astrophysics Data System KENNICUTT 1/2 is considerablymorestochasticthancanbepredictedby parently boththemassspectrumandmassesof largest nebularcomplexesinparticularareextremely the distributionfunctioninindividualgalaxies.Ap- in theintroduction,presentsampleisessentially distance maybedeterminedusingHnregionproperties. and 5imposesthebasiclimitonhowaccuratelyagalaxy’s this interpretationiscorrect,thenthespreadinFigures4 sensitive functionsofgasdensityandgalaxydynamics.If partly onthebasisofHnregionresolvability.Hence, Tammann werebasedonasampleoffieldspiralsselected previous workinthisseriesandtheofSandage complete forScgalaxiesbrighterthanM^—19.The evidence foranabnormalitybroughtaboutbythecluster environment. Thepresentsamplewouldhavetobe of thebiasisdifficulttoestimate.Itpossible,course, axies withfaint,smallHnregions,thoughthemagnitude older fieldsamplesareundoubtedlybiasedagainstgal- chosen sampleoffieldgalaxiesforthispossibilitytobe compared withamuchlargerandmoresystematically that thespreadinVirgoHnregionsisactually worst, thatHnregionsarenotnearlyasprecise dismissed. Tentatively,however,wemustassumethe paper. of thisconclusionarediscussedinthelastsection Virgo tomakeareasonabledistancedetermination. standard candlesaswemayhavehoped.Theimplications Absolute Hnregionfluxes,galaxymagnitudes,and individual galaxies,wehavealargeenoughsamplein luminosity classesareavailablefor17calibratinggal- words, fitFigs.2and5together).Theslopeoftheother follow theprocedureofS-T(1974c)andfitVirgodata axies, fiveintheLocalGroup,sixM81group,and data consistprimarilyofgalaxieswithLCIII-IVor derive theVirgodistanceindependentofM101 six intheM101group(PaperIII;seealsoFig.1),andwe Virgo datawerefittedtotwoarbitraryinterpolating brighter. Sincetheregionofoverlapisquitesmall,we luminosity class(LC)II-IIIorfainter,whiletheVirgo the remainingcalibrationsetconsistsofgalaxies distance orHnregionproperties.IfM101isthrownout, a degeneratedistanceline(slope=0.4)tobeofuse. to theHafluxversusluminosityclasscalibration(inother B functions, alogarithmicform(log)versusLC].Distanceswere except forthetwo,extreme,class Vdwarfs(IC1613and form, plottedinFigures2and5,fitsthedatafairlywell, then derivedforbothformstoassessthesensitivityof result ontheinterpolatingfunction.Thelogarithmic to beobservedinVirgo,theywere assignedzeroweightin Ho I)inthecalibratingsample. Sincetheyarefartoofaint shown) alsofitsthedatafairly wellandmakessense 3 3 Why isthedispersiononlyevidentnow?Asdiscussed In spiteofthelargedispersioninHnregionfluxesfor For thereasonsgiveninintroduction,wewishto VI. THEDISTANCETOVIRGOCLUSTER Vol. 247 No. 1, 1981 H il REGIONS AS DISTANCE INDICATORS 15 because H n diameter scales roughly linearly with lumin- osity class, while H n region flux scales approximately with the square of the diameter. The Virgo and calibration data were then fitted separately by linear least squares, in the variables 40 described above. In each case, half-weight was given to a galaxy whose H n regions’ fluxes exceeded the mean relations by many standard deviations, NGC 2366 in the calibrating sample and NGC 4299 in Virgo. (Both possess a single, very bright region which accounts for the 39 departure.) The Virgo distance was then derived by matching the calibration and Virgo regressions at lumin- K> osity class III, where the best overlap between samples AO occurs. The distances derived, using the local distance X V scale calibration of S-T (1976) and adjusted to a Hyades o» modulus of 3T30 are: o 38 logarithmic fit 18.7 ±1.5 Mpc , quadratic fit 19.5 ± 2.7 Mpc , and Virgo cluster distance 19 ± 2 Mpc . 37 This corresponds to a true distance modulus (m — M) = 31.4 ± 0.25 . Fig. 6.—Combined Local and Virgo H n region calibration, assum- The uncertainties quoted above represent the range of ing a distance of 19 Mpc to the Virgo Cluster. distances that could be derived by attaching different weights to NGC 2366 and 4299, by including or deleting dwarf members of the M101 group, etc. Naturally, if the sharply bounded Strömgren regions, easily measured, somewhat smaller distance scales of de Vaucouleurs and, from the small sample available, they seemed to (1978) or van den Bergh (1980) are used, the Virgo dis- possess a very small dispersion, at least among galaxies of tance will decrease proportionally. the same type and luminosity. Those assumptions have Assuming a distance of 19 Mpc to the Virgo cluster, we been critically examined in this series of papers, and most have plotted the combined H n region flux and LC no longer appear to be valid. The majority of the regions relation in Figure 6. This single regression is to be are not classical Strömgren spheres; their apparent preferred over the “ maximum,” u minimum,” “ best ” fit diameters are strongly dependent on the measuring mess in Papers II and III. Coincidentally and quite technique employed and can only be accurately happily, the extrapolation to bright luminosity classes in determined by careful photographic and photoelectric the Paper III “ best ” fit agrees quite closely with Figure 6. photometry. If the Virgo cluster sample is representative Hence, the field galaxy distances derived in that paper are of the field, then even carefully measured fluxes and not changed significantly. In light of the dispersion in diameters will often yield distance errors of a factor of 2. Figure 6, however, any distances to field galaxies derived In that case, it may be far more practical to employ one of in Paper III (or elsewhere) must be treated very cau- the many more readily available distance indicators, such tiously. The best example of this is M101 itself. While the as galaxy diameters or magnitudes calibrated as a func- M101 H ii regions are within the range of those in Virgo, tion of luminosity class or H i line width. Figure 6 shows that they are probably about a magnitude At times, a number of investigators have discussed the brighter than average. Had one previously based a dis- feasibility of using the Space Telescope to measure H n tance scale for Sc I galaxies on M101 alone, the distances region sizes in very distant galaxies, in order to determine derived would have been roughly 60% too high. the Hubble constant at cosmological distances. We hope that the results of this paper will discourage any such futile enterprise. Measuring the mean distance to a VII. CONCLUSIONS—IMPLICATIONS FOR THE sample of galaxies to an accuracy of ±10% would DISTANCE SCALE require observation of approximately 20-30 galaxies, Van den Bergh (1980) recently speculated that uncer- depending on how representative the scatter found here tainties in the existing data on H n region sizes and really is. An even larger sample might have to be galactic luminosity classes render the H n regions unsuit- measured in order to ensure that selection biases did not able for measuring accurate distances. The results of this seriously affect the data. It seems clear that precious paper seem to confirm those suspicions. When H n Space Telescope observing time will have to be spent on regions were first applied by Sérsic (1960), Sandage standard candles with a smaller cosmic dispersion than (1962), and S-T (1974a), they appeared to be ideal as the H ii regions. distance indicators. They were bright, they appeared to be When these results are combined with the recent
© American Astronomical Society • Provided by the NASA Astrophysics Data System 1 98lAp J . . .247 9K 1- 16 discussion oftheluminosityclass-absolutemagnitude entire upperpartoftheS-Tladderappearstobe calibration byTammann,Yahil,andSandage(1979),the ment ofthecompetingdistanceladdersAaronsonetal. questionable. Shouldthisbeinterpretedasanendorse- not. Indeed,thedistancederivedtoVirgocluster above correspondstoavelocitydistanceratioof~55 km s"Mpc,closetotheoriginalS-Tvalue.On (1980) ordeVaucouleursandBollinger(1979)?Wethink with theS-TPopulationIdistanceladder,“rung” other hand,itappearsthatifanyprogressistobemade presently occupiedbyHnregionswillhavetobe Aaronson, M.,Mould,J.,Huebra,Sullivan,W.T.,Schommer,R.A., bright supergiantstars. candle, perhapsbySpaceTelescopeobservationsof replaced orcircumventedbyamorereliablestandard .1979,M.N.R.A.S.,186,609. de Vaucouleurs,G.,A.,andCorwin,H.G.1976,Second .1978,Ap.J.,224,710. de Vaucouleurs,G.1961,Ap.J.Suppl,6,213. Branch, D.1977,M.N.R.A.S.,179,401. Robert C.Kennicutt,Jr.:DepartmentofAstronomy,SchoolPhysicsandUniversityMinnesota,116 de Vaucouleurs,G.,andBollinger,G.1979,Ap.J.,233,433. Church Street,SE,Minneapolis,MN55455 .1979a,Ap.J.,228,394(PaperI). -.1919b,Ap.J.,228,696(PaperII). Kennicutt, R.C.1978,Ph.D.thesis,UniversityofWashington. .1979c,Ap.J.,228,704(PaperIII). .1981,inpreparation. and Bothun,G.1980,Ap.J.,239,12. Reference CatalogueofBrightGalaxies(Austin:UniversityofTexas Press). © American Astronomical Society • Provided by theNASA Astrophysics Data System KENNICUTT REFERENCES assisted inthepreparationofthismaterial.Specialthanks are owedtothenightassistantsatKittPeakandPalomar Mountain, especiallytoSkipStaplesatPalomar.The and JohnHoessel.BillSebok’simage-processingsystem P-60 directSITisthecreationofJimGunn,SteveKent, Newton, andJohnBedke.AllanSandageprovided photometry. Preparationofthemanuscriptwasexpertly aided invaluablyintimelyandaccuratereductionofthe handled byHelenCzaplicki,PamGilman,Nancy -.1980,ARevisedShapely-AmesCatalogofBrightGalaxies Kennicutt, R.,Balick,B.,andHeckman,T.1980,Pub.A.S.P.,92,134. prints inadvanceofpublication,andPaulHodgekindly classifications andvelocitiesloanedhigh-resolution Sandage, A.R.1962,inIAUSymposium15,ProblemsofExtragalactic Kent, S.M.1979,Pub.A.S.P.,91,394. lent hisHafilterset.Thisworkwascompletedwhile .19746,Ap.J.,194,223. Mould, J.,Aaronson,M.,andHuchra,J.1980,Ap.238,458. and LasCampanasObservatories. the authorheldaCarnegieFellowshipatMountWilson .1976,Ap.J.,210,7. .1974c,Ap.J.,194,559. Sandage, A.R.,andTammann,G.1974a,Ap.J.,190,525. Tammann, G.A.,Yahil,andSandage,A.1979,Ap.J.,234,775. .1980,Ap.J.,235,1. van denBergh,S.1960,Pub.DavidDunlapObs.,vol.2,no.6. Sérsic, J.L.1960,Zs.Ap.,50,168. Research, ed.G.C.McVittee(NewYork:MacMillan),p.359. (Washington: CarnegieInstitutionofWashington). Heartfelt thanksareowedtothemanypeoplewho 1 98lAp J . . .247 9K (east) edgeofthemainfigure. Kennicutt (seepage9) Fig. 1.—Habandpassphotographs of M101(top)andNGC3631(bottom).Insetinphotographsshows NGC5471,whichisjustofftheleft © American Astronomical Society •Provided bythe NASAAstrophysics Data System *% V ¥ 1 PLATE 1 1 98lAp J . . .247 9K Kennicutt (seepage11) has beenprintedatthesamescaleto facilitatecomparison.Brightforegroundstarsaredenotedbycrosses. PLATE 2 Fig. 3.—HabandpassphotographsofsixVirgoclusterScgalaxies,chosen toillustratethedispersioninHnregionproperties.Eachphotograph © American Astronomical Society •Provided bythe NASAAstrophysics Data System