Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed20January2009 (MNLATEXstylefilev2.2) Mid-IR period-magnitude relations for AGB stars I.S. Glass1, M. Schultheis2, J.A.D.L. Blommaert3, R. Sahai4, M. Stute5 and 9 S. Uttenthaler3 0 0 1South African Astronomical Observatory, PO Box 9, Observatory 7935, South Africa 2 2Observatoire de Besan¸con, 41bis, avenue de l’Observatoire, F-25000 Besan¸con, France 3Instituut voor Sterrenkunde, K.U. Leuven, Celestijnenlaan, 200D, B-3001 Leuven, Belgium n 4NASAJPL, 4800 Oak Grove Drive, Pasadena, CA 91109, USA a 5Department of Physics, University of Athens, Panepistimiopolis, 15784 Zografos, Athens, Greece J 0 2 Submitted2008;accepted ] R S ABSTRACT . h Asymptotic GiantBranchvariablesare found to obey period-luminosity relations p inthemid-IRsimilartothoseseenatKS (2.14µm),evenat24µmwhereemissionfrom - o circumstellar dust is expected to be dominant. Their loci in the M, logP diagrams r are essentially the same for the LMC and for NGC6522 in spite of different ages and st metallicities.Thereisnosystematictrendofslopewithwavelength.Theoffsetsofthe a apparent magnitude vs. logP relations imply a difference between the two fields of [ 3.8 in distance modulus. The colours of the variables confirm that a principal period with log P > 1.75 is a necessary condition for detectable mass-loss. At the longest 1 v observedwavelength,24µm,many semi-regularvariables have dust shells comparable 2 in luminosity to those around Miras. There is a clear bifurcation in LMC colour- 3 magnitude diagrams involving 24µm magnitudes. 0 3 Key words: Stars: AGB and post-AGB, stars: variables: other, stars: mass-loss, . stars: late-type, stars: fundamental parameters 1 0 9 0 : 1 INTRODUCTION sometimes with additional periods about 10 times as long v i astheprincipalones.Theoriginofthelongperiodsisnotyet X The existence of near-infrared period-luminosity relations understood.Atleast onecaseisknowninwhichavariation inLMCAsymptoticGiantBranch(AGB)starsisnowwell- r of this kind starts as a momentary decrease in amplitude a known.TheyhavealsobeenfoundintheNGC6522Baade’s and develops in width over subsequent cycles (Blanco 26; Window of the inner Bulge by Glass & Schultheis (2003) see Glass & Schultheis, 2002). andhavesincebeenextendedtoglobularclustersandother members of the Local Group. The LMC and the NGC6522 One of the most important aspects of the long-period fields differ markedly in metallicity but contain large num- variables is their high mass-loss rates. Though Mira vari- bers of stars at approximately uniform distances, making ables are major contributors of matter to the interstellar them highly suitable places for studying population differ- medium, the ISOGAL survey in the galactic plane (Omont ences. et al 2003), which made use of the ISO infrared satellite, Among the AGB stars, many of the semi-regular vari- showed that many other late-type giants also shed large ables have small amplitudes so that, in contrast to the Mi- quantities of dust. The [7] - [15] vs [15] colour-magnitude ras,theycanrevealusefuldistanceinformation fromsingle- diagrams that resulted from ISOGAL show a continuous epochobservationsiftheirperiodsareotherwiseknown.In- distribution of mass-losing stars from the ‘blue’ end of the fraredstudieshavetheadditionaladvantagethattheeffects colour range to the ‘red’ (Glass et al. 1999). Because these of interstellar absorption are minimized; Aλ in the mid-IR stars are comparatively numerous they rival the Miras in can beas low as ∼AV ×0.04 mag. their total dust output. Using an objective prism survey of Examinationoftheperiod-luminosityrelationsofAGB theNGC6522fieldbyBlanco(1986)itwaspossibletoshow starsoffersthepossibilityofgaininginsightintothephysics thatthemass-losingstarsareallgiantsoflaterspectraltype of their outer atmospheres. Many of the semi-regulars ex- thanaboutM5(Glass&Schultheis2002).Moreover,thanks hibitmultipleperiodicities. Theprincipalpeaksintheirpe- to variability data generated as a by-product of the MA- riodograms lie in the range 10-200 days, usually accompa- CHO,OGLEandsimilar gravitational-lensing experiments, nied by other peaks within a factor of two in period and it is also known that they are nearly all semiregular vari- 2 I.S. Glass, M. Schultheis, J.A.D.L. Blommaert, R. Sahai, M. Stute, S. Uttenthaler ables,withtheadditionofafewMiras(Glass&Alves2000; ifluxestolevelsbelowtheMACHOdetectionthreshold.At Alard et al. 2001; Glass & Schultheis2002). leasttwoobjectsmoreluminousthantypicalAGBstarsare The Spitzer infrared satellite has now surveyed the present in our sample. One is a listed supergiant (WOH S Large Magellanic Cloud (SAGE; Meixner et al. 2006) and 286).Themostapparentlyluminousofall,atM =–15.25 24 theNGC6522field1 atmid-infraredwavelengthswithsensi- is an Hα object (LHA 120-N 132E) which has a MACHO tivitiesunobtainablefromtheground.Eachofthesesurveys counterpart with logP = 2.661 (it is too bright to appear used the IRAC camera at 3.6, 4.5, 5.8 and 8 µm and the in Fig 1). It should be mentioned, in view of the compar- MIPS camera at 24 µm. The SAGE data have been pub- isonswemake,thatUttenthaleretal.(inpreparation)have liclyreleased;theNGC6522datawerereducedbyM.Stute reduced the SAGE data using their methods and find no (Uttenthaler et al., in preparation). Because the NGC6522 systematicdifferenceswiththeresultsofBlumetal(2006). field has beenso well studied in thepast,it was an obvious The NGC6522 field does not contain any objects more candidate for a Spitzer survey. It has importance as a fidu- luminous than the Miras nor, unlike the LMC, does it pos- cial field for application to more heavily extincted areas of sess any AGB stars with carbon chemistry. Saturation (the the inner Bulge where visible region data are impossible to criterion being that the aperture and PSF fluxes disagree obtain. at > 7% level) sets in at 320, 330, 2300 1200 and 220 mJy Schultheis, Glass and Cioni (SGC2004) compared the for the IRAC/MIPS bands, corresponding at dm = 14.7 to JHKS properties of AGB stars in the Milky Way galaxy magnitudesof–7.34,–7.86,–10.45,–10.36,–3.98.Thusmea- and Magellanic Clouds. In summary, they formed complete surementsofthebrighterstarsoftheNGC6522 fieldin the samplesbyselectingallstarswithMKS,0 <∼–5.0intherel- [3.6] and [4.5] bands, though included in the diagrams, are evantareasfromthe2MASSCatalog.Thislimitisabout1.5 likely to be somewhat affected. magbelowthetipoftheRGB.Theobjectstheyfoundwere Wehaveincluded a numberof AGBstars from theSo- carefully cross-identified with the MACHO survey at r and lar Neighbourhood for comparison. Because local AGB gi- i. Each MACHO light curve was Fourier analysed to yield ants are too bright to be measured by Spitzer, synthetic its periods and associated amplitudes. The sizes of the two IRACphotometryofanumberofthemhasbeenderivedby fields of interest in this study, NGC6522 at 1032 arcmin2 Marengo, Reiter and Fazio (2008) from ISO-SWS spectra. and the LMC at 271 arcmin2, were chosen to yield about Periodsforthese,whereavailable,havebeentakenfromthe 1800starsfromeach.Theobjectswerealsoinvestigatedus- GCVS as theystand. Hipparcos(The NewReduction) par- ing7µmand15µmphotometryfromISO,whereavailable. allaxes(post-September2008)wereobtainedfromtheCDS. However, theSpitzerinstrumentsnow offerhigh photomet- Onlystarswithparallaxesπ>5σπhavebeenincluded.Most ric precision at the faint end and therefore the possibility ofthestarsthatsatisfiedthiscriterionareM-type;onlytwo of continuing the ISOGAL study to stars with lower mass- C stars appear. loss rates. Spitzer also extends the long-wavelength end of the survey from 15 to 24 µm, where radiation in excess of thatexpectedfromthephotospheresshouldbemoreclearly 3 PERIOD-MAGNITUDE DIAGRAMS apparent. Figure 1 contains the fiveM vs. logP diagrams for the two fields, superimposed to show their essential similarity. Dots correspond to the dominant periods. Miras are stars with 2 THE DATA MACHO amplitudes > 1.0. Note that the scatter in abso- TheNGC6522 andLMCfieldsdiscussedhereusingSpitzer lutemagnitudes due todepth effects is much greater in the data are sub-fields of those investigated by SGC2004. In NGC6522 than in theLMC field. the current work, theNGC6522 vs. MACHO and LMC vs. VersionsofFig1werepreparedforseveraldifferentval- MACHO lists from SGC2004 were cross-identified with the uesofthedmoftheNGC6522fieldandcompared visually. Spitzer photometry. The IRAC and MIPS fluxes were con- The fit of the shortest-period SRVs to the corresponding vertedintomagnitudesbytakingzero-pointsinthe3.6,4.5, sequence in the LMC appeared to be the clearest discrim- 5.8, 8 and 24 µm bands of 2.809 ×108, 1.797 ×108, 1.15 inant and was best with dm = 14.7 to 14.8. Increasing or ×108,6.41×107and7.14×106µJyrespectively.Theredden- decreasing the dm by 0.1 mag produced noticeable devia- ing of the NGC6522 field was assumed to be EB−V =0.5. tions. This value appears to be similar to other determina- The distance modulus (dm) for the LMC was taken to tions of the distance modulus of the Galactic Centre based be 18.5. Following trials of several values for the dm of the on AGB stars but is higher than those found using other NGC6522 field, 14.7 was adopted (see section 3). objectsandmethods.Groenewegen,Udalski&Bono(2008) In theLMC oursurvey covers about 0.00154 times the summarize the current position and find dm = 14.50 based area of SAGE (Blum et al 2006). Because we only discuss onRRLyraevariablesandPopulation-IICepheids.Noclear objects listed in the MACHO survey, a number of very red explanation for this discrepancy has yet emerged. It can be objectsdetectedbySpitzerdonotappearinourlists.They remarked that, because they are infrared-based, the AGB- rangein24µmmagnitudefrom8.5to15.3andin[3.6]–[24] based determinations should belargely unaffectedby inter- colour from 4.1 to 9.0. Many of them may be AGB stars stellar reddening. withcircumstellarshellsthickenoughtodepresstheirrand The five Spitzer period-magnitude diagrams clearly show that the sequences seen in the KS, logP diagrams bySGC2004 also appear at longer wavelengths, though the 1 AspartoftheprogrammeID2345:‘ASpitzerSurveyofMass LMCsampleislimitedat[24] bythesensitivityofMIPSto LosingStarsintheGalacticBulge’,P.I.R.Sahai M[24] brighterthan–9.Inspiteofthepresenceofdustshells Mid-IR period-magnitude relations for AGB stars 3 magnitudesarenotavailableandabsolute25µmmagnitudes derived from IRASare plotted instead. 3.1 Slopes of the relations In order to determine whether the slopes of the period- magnitude relations change with wavelength, straight lines were fittedtoseveral of thesequencesvisible in Fig. 1. The periods were assumed free of errors. The nomenclature of thesequencesisbased onthatof Itaetal. (2004). Onlythe LMC data were used for thispart of theinvestigation since theyshowtheclearestseparations.NotallofIta’ssequences areseparateenoughtoincludeinthisanalysis.Evenamong thosethatareincluded,B+andB−areratherpoorlydefined and may be contaminated. The parts of period-magnitude spaceusedforfittingareshowninFig.2andtheresultsare giveninTable1.TheKS vs.P dataforLMCstarsgivenby SGC2004 have been included in the analysis for complete- ness. As in Fig. 1, stars with MACHO r amplitudes greater than 1.0 were taken to be Miras. This level was chosen to include carbon Miras, which tend to have lower amplitudes thanM-types(seee.g.Glass&LloydEvans2003).MostMi- ras fall in the C region, though smaller amplitude stars are alsoincluded.ObjectswithJ−KS >1.6wereassumedtobe carbon stars (though this is not an infallible criterion) and were not included in the linear fits. Previous work such as thatof Glass et al.(1987) and Feast et al.(1989) suggested that O- and C-type Miras obey essentially the same rela- tions at KS, though differing at J and H and consequently in M . The carbon Miras in Fig. 2 seem to lie below the Bol M-star linear fit at KS but to be above it at longer wave- Figure 1.log P vs. Spitzer magnitudes. Red dots arethe LMC lengths. However, it should be noted that the sample is a SRVs (principal periods), red ×s the LMC Miras, magenta dots the LMC subsidiary long periods and black and green the same small one and that the KS and IRAC data were not taken simultaneously. respectivelyforNGC6522. TheprincipalM,logP sequences are clearlyvisible.NotethattheLMCdataarelimitedinsensitivity Table 1 shows that the stars show little scatter (σ) at 24µm. Bright NGC6522 points (longest periods) suffer from aroundthederivedperiod-magnituderelations,especiallyin saturation at 3.6 and 4.5µm (see text). Local AGB stars from the A and B sequences. The SRVs of these sequences have syntheticphotometrybyMarengoetal.(2008)areshowninblue the smallest amplitudes; the C sequence, however, contains with×forM-typesand+forC-types. somelarge-amplitudevariables(Miras)andcanbeexpected to show more scatter. For each of the sequences A+, A−, B+, B−, C and D thereis no systematic changein slope with wavelength. around the longer-period stars at this wavelength, there is still some tendency towards a relation. The fluxesfrom the shorter-periodvariablesremainphotospheric.FromKS to8 4 PERIOD-COLOUR DIAGRAMS µm essentially the same M – logP sequences are repeated. Though SGC2004 presenteda[7]vslogP diagram and sug- Thecoloursofthesamplesareplottedagainst theprincipal gested that the sequences may persist at this wavelength, and the very long periods in Fig. 3. The most conspicuous theywerenotasobviousashereduetothesmallernumbers trendis noticed in thelog P vs.K0,S - [24] diagram, where of stars, the limited depth of the photometry and possibly the SRVs show a very sharp onset of excess radiation at P higher scatter in thecolours. ∼ 60d. This is, in effect, an exaggerated version of what is As expected from theoretical models, the LMC stars, seen in the 15µm excess vs. logP diagram of Alard et al. being of lower metallicity, tend toreach higher luminosities (2001). Noteworthy again is the tendency for many SRVs than the galactic ones. This effect was previously remarked to have very red infrared colours, comparable to, or even on by SGC2004 for the KS band. Groenewegen and Blom- greaterthan,thoseoftheshorter-periodMiras.Itshouldbe maert (2005) point out also that the LMC contains higher noted, however, from Fig 1 that they are not necessarily as mass stars than the NGC6522 field. There are many ad- luminous. ditional subtle differences between the M, logP relations The SRVswith secondary verylong periods show clear which are unfortunately difficult to quantify because of the evidence, particularly at 24 µm, for the existence of dust increased scatter dueto deptheffects in NGC6522. shells.Unfortunately,theLMCsampleislimitedtothemost In the case of the Solar Neighbourhood sample, 24µm luminous cases bythesensitivity of MIPS. 4 I.S. Glass, M. Schultheis, J.A.D.L. Blommaert, R. Sahai, M. Stute, S. Uttenthaler Table 1.LinearfitstotheM,logP sequences Seq. Slope Const.term No σ KS-band A− -3.11±.11 -1.67±.14 214 0.13 A+ -3.46±.19 -1.43±.29 67 0.15 B− -2.78±.14 -1.57±.22 96 0.14 B+ -3.85±.16 0.19±.29 70 0.13 C -3.56±.29 0.86±.66 34 0.29 D -3.61±.18 3.20±.51 90 0.29 3.6µmband A− -2.99±.10 -1.99±.14 207 0.13 A+ -3.50±.19 -1.58±.29 59 0.14 B− -2.89±.17 -1.56±.26 89 0.15 B+ -4.05±.17 0.17±.31 76 0.15 C -4.00±.26 1.46±.59 40 0.29 D -3.78±.19 3.44±.53 88 0.30 4.5µmband A− -2.95±.10 -1.92±.13 240 0.14 A+ -3.75±.25 -1.00±.39 48 0.16 B− -2.90±.17 -1.42±.26 109 0.17 B+ -3.81±.17 -0.14±.31 72 0.14 C -3.62±.25 0.63±.56 43 0.31 D -3.74±.19 3.43±.54 89 0.30 5.8µmband A− -3.22±.11 -1.68±.14 230 0.14 A+ -3.58±.20 -1.36±.32 53 0.14 B− -2.71±.18 -1.74±.28 101 0.18 B+ -3.67±.20 -0.46±.36 83 0.19 C -4.08±.25 1.542±.57 41 0.30 D -3.98±.21 4.01±.58 87 0.33 8µmband A− -3.11±.13 -1.87±.17 184 0.15 Figure 2. IRAC and 2MASS KS absolute mags vs. periods for A+ -4.21±.23 -0.50±.36 49 0.14 theLMCsample.BasedonJ−KScolour,thepointsrepresenting B− -2.80±.17 -1.72±.26 92 0.17 oxygen-richstarsareblackandthoseforthecarbon-richarered. B+ -3.75±.21 -0.44±.39 71 0.17 Miravariablesareshownasasterisks.Theregionsusedformaking C -4.35±.26 1.98±.58 36 0.28 thelinearmagnitude vs.logP fits (tothe O-richstarsonly)are D -4.15±.25 4.32±.68 86 0.38 giveninred.Thecoefficientsofthefittedlinesingreenarelisted Note:‘No’isnumberofstarsinsample. inTable1. 5 THE [3.6] – [24] VS. M [24] A similar trend was previously observed by SGC2004 COLOUR-MAGNITUDE DIAGRAM in ISO data. In their work, only 7 and 12µm photometry Inthisdiagram(Fig4),AGBstarsfromboththeLMCand was available for the Magellanic Clouds and 7 and 15µm NGC6522 fields are shown, together with thelocal sample. for the NGC6522 field. The present use of identical bands As earlier, LMC stars with J −KS > 1.6 are taken to be obviatesanysuggestion thatthedifferencebetweenthetwo C-typeand theremainder M-type. fields arises from the different spectral responses of ISO in The LMC points (red) are conspicuously bifurcated in the12 and 15µm bands. the range –8.5 > [24] > –11. Those on the right side of the Again, for local AGB stars, 25µm magnitudes from fork (redder colours) are mainly O-rich and comparable in IRAS have been used. These objects appear to lie slightly positiontotheNGC6522stars.Theleftforkcontainsmany to the left of their NGC6522 counterparts. It is not clear carbon stars. whetherthisisduetotheuseoftheIRAS[25]bandinstead Thecarbonstarsareclearlyseparatedfromoxygen-rich of the MIPS [24], to a calibration error or to a truly phys- ones in the sense of being less red in [3.6] – [24] colour for ical effect such as higher luminosities in the Marengo et al. a given M . A similar effect is seen whichever m – [24] (2008) sample.ThetwolocalCstarslietotheleftoftheM [24] colour is used on the abscissa but not if say M is used stars;largersampleswouldofcourseberequiredtoconfirm [8] astheordinateinstead ofM .Though thedifferencemay thisresult. [24] ariseinpartfromthehigherluminositiesofthecarbonstars, A recent study of variables found by comparing two thereisalackofemissionfromcarbon-richdustintheregion epochsoftheentireSAGEsurvey(Vijhetal.2008)presents covered by the 24µm band. The IRAS [25] – [60] vs. [12] – a [24] vs. [4.5] - [24] magnitude-colour plot (their Fig. 4). [25] colour-colour diagram (e.g. van der Veen and Habing, This is, however, dominated by the class of ‘extreme AGB 1988) showsquiteclearlythatgalacticC-richstarsseparate stars’; SRVs would occupy mainly the bottom left corner fromO-richonesat25µm.Inthelattercase,thispartofthe (low-luminosity and moderately red variables) and would spectrum corresponds to thebroad 18µm silicate feature. beextremely numerousif all were detected. Mid-IR period-magnitude relations for AGB stars 5 Figure4.[24]magvs.[3.6]-[24]colourforLMCandNGC6522 AGB stars, including non-variables. This diagram also incorpo- ratesnearbystarsfromMarengoetal.(2008)withπ>5σπ using IRAS [25] mags instead of Spitzer [24]. The LMC sample is di- vided into carbon stars (red + symbols) and O-rich stars (red dots). Stars not detected by MACHO are omitted. Note the bi- furcation which we ascribe to a deficiency of 24–25µm flux in Figure3.logP vs.KS,0-Spitzercolours.ReddotsaretheLMC C-richdustandpossiblyalsotohigheroverallluminosities. SRVs (principal periods), red ×s the LMC Miras, magenta dots the LMC subsidiary long periods and black and green the same for NGC6522 respectively. The brightest stars in the NGC6522 REFERENCES field,suchastheMiras,sufferfromsaturationeffectsinthelower AlardC.etal,2001,ApJ,552,289 threediagrams. BlancoV.M.,1986,AJ,91,290 BlumR.D.etal,2006,AJ,132,2034 Feast M.W., Glass I.S., Whitelock P.A., Catchpole R.M., 1989, MNRAS,241,375 GlassI.S.etal,1999,MNRAS,308,127 GlassI.S.,AlvesD.R.,2000,InISOSurveysofaDustyUniverse, edsLemkeD.StickelM.WilkeK.,SpringerLectureNotesin 6 ACKNOWLEDGMENTS Physics548. WethankDr.M.Marengoforprovidinguswithhisdetailed Glass I.S., Catchpole R.M., Feast M.W., Whitelock P.A., Reid, syntheticphotometryresults.Weacknowledgetheuseofthe I.N.,1987,inKwokS.,PottaschS.(eds),LateStagesofStellar Evolution,Reidel,Dordrecht Simbad and Vizier on-line resources of CDS, Strasbourg. GlassI.S.,LloydEvansT.,2008MNRAS,343,67 ISG’s travel was supported by the CNRS/NRF agreement. GlassI.S.,Schultheis M.,2003,MNRAS,345,39 Dr.M.Groenewegenkindlycommentedonanearlierversion GlassI.S.,Schultheis M.,2002,MNRAS,337,519 of the ms. GroenewegenM.A.T.,BlommaertJ.A.D.L.,2005,A&A,443,143 MStissupportedbytheEuropeanCommunity’sMarie GroenewegenM.A.T.,UdalskiA.,BonoG.2008,A&A,481,441 Curie Actions - Human Resource and Mobility within the ItaY.etal,2004,MNRAS,353,705 JETSET (Jet Simulations, Experiments and Theory) net- Marengo M., Reiter M., Fazio G.G., 2008, AIP Conference Pro- work under contract MRTN-CT-2004 005592. The data re- ceedings,1001, 331 ductionoftheNGC6522SpitzerdatawascarriedoutbyMSt MeixnerM.etal,2006,AJ,132,2268 in collaboration with RS at the Jet Propulsion Laboratory, OmontA.etal,2003, A&A,403,975 Schultheis M., Glass I.S., Cioni M.-R., 2004, MNRAS, 427, 945 California Institute of Technology, under a contract with (SGC2004) theNationalAeronauticsandSpaceAdministration.RSre- VanderVeenW.E.C.J.,HabingH.,1988, A&A,194,125 ceived partial support for this work through JPL/Caltech VijhU.P.etal,2008,arXiv0811-0408v1 SpitzerSpace Telescope GO award MN0710076.