Baryonic Tully-Fisher Relation for Extremely Low Mass Galaxies

0.23 MB·English

by  Ayesha Begum

Mon.Not.R.Astron.Soc.000,000–000(0000) Printed2February2008 (MNLATEXstylefilev2.2)
Baryonic Tully-Fisher Relation for Extremely Low Mass Galaxies
Ayesha Begum1⋆, Jayaram N. Chengalur2, I. D. Karachentsev3 and M. E. Sharina3
1InstituteofAstronomy,UniversityofCambridge,MadingleyRoad,Cambridge,CB30HA,UK
8 2NationalCentreforRadioAstrophysics,PostBag3,Ganeshkhind,Pune411007,India
0 3SpecialAstrophysicalObservatory,NizhniiArkhys369167,Russia
0
2

n
a
J

3
ABSTRACT
2 We study Tully Fisher relations for a sample that combines extremely faint (MB < −14.)
galaxies along with bright (i.e. ∼ L∗) galaxies. Accurate (∼ 10%) distances, I band pho-
]
h tometry,andB-Vcolorsareknownforthemajorityofthegalaxiesinoursample.Thefaint
p galaxiesaredrawnfromtheFaintIrregularGalaxyGMRTsurvey(FIGGS),andwehaveHI
- rotationvelocitiesderivedfromaperturesynthesisobservationsforallofthem.Forthefaint
o
galaxies,we findthateventhoughthe medianHIandstellar massesare comparable,the HI
r
t masscorrelatessignificantlybetterwiththecircularvelocityindicatorsthanthestellarmass.
s
a Wealsofindthatthevelocitywidthatthe20%level(W20)correlatesbetterwithmassthanthe
[ rotationvelocity,althoughthedifferenceisnotstatisticallysignificant.Thefaintgalaxieslie

systematicallybelowtheIbandTFrelationdefinedbybrightgalaxies,andalsoshowsignif-
1
icantlymoreintrinsicscatter.Thisimpliesthattheintegratedstarformationinthesegalaxies
v
hasbeenbothlessefficientandalsolessregulatedthaninlargegalaxies.Weestimatethein-
6
trinsicscatterofthefaintgalaxiesabouttheIbandTFtobe∼1.6mag.Wefindthatwhilethe
0
6 faintenddeviationisgreatlyreducedinBaryonicTully-Fisher(BTF)relations,theexistence
3 ofabreakatthefaintendoftheBTFissubjecttosystematicssuchastheassumedstellarmass
. to light ratio. If we assume that there is an intrinsic BTF and try to determine the baryonic
1
massbysearchingforprescriptionsthatleadtothetightestBTF,wefindthatscalingtheHI
0
massleadsto a muchmoresignificanttighteningthanscalingthe stellar masstolightratio.
8
Themostsignificanttighteningthatwefindhowever,isifwescaletheentirebaryonicmass
0
: of the faint (but not the bright) galaxies. Such a scenario would be consistent with models
v
wheredwarf(butnotlarge)galaxieshavealargefractionofdarkor“missing”baryons.Inall
Xi caseshowevertheminimumintheχ2 curveisquitebroadandthecorrespondingparameters
arepoorlyconstrained.
r
a
Key words: galaxies: dwarf – galaxies: kinematics and dynamics – galaxies: individual:
FIGGSradiolines:galaxies
1 INTRODUCTION galaxymasses.Reheatingpriortotheepochofreionizationwould
leadtosmallhaloshavingareducedbaryonfraction(e.g.Gnedin
Oneofthestringenttestsofgalaxyformationmodelsistheirabil-
(2000)), and small dark matter halos are also expected to easily
itytoreproducethetightnessoftheTully-Fisher(TF)relation,i.e.
losebaryonsbecauseofenergyinputfromthefirstburstofstarfor-
thefactthattherotationvelocityandabsolutemagnitudeofbright
mationandsupernovawinds(e.g.Tassisetal.2007;Dekel&Woo
spiralgalaxiesaretightlycorrelated.Forsuchgalaxies,theintrinsic
2003;Efstathiou2000).Largegalaxiesontheotherhandaremore
scatterintheIbandTFhasbeenestimatedtobeassmallas0.2mag
abletokeeptheiroriginalshareofbaryons.Consistentwiththisex-
(e.g.Sakaietal.(2000)),whileVerheijen(2001)arguesthatthere
′ pectation,observationsshowthatdwarfgalaxieswithrotationve-
isprobablynointrinsicscatterintheK bandTFrelation.Thisim- locitieslessthan∼90kms−1liebelowtheTFrelationdefinedby
pliesthatthedarkandvisible(i.e.stellar)mattercontentofgalax-
brightergalaxies(McGaughetal.2000;Begum&Chengalur2004;
ies are tightly correlated, or, if one assumes that the visible mat-
McGaugh,S.2005),i.e. they areunder luminous for the velocity
tertracesthebulkofthebaryons,thatthenonbaryonicandbary-
width.Ontheotherhand,Gehaetal.(2006)usinganindependent
onic content of galaxies are tightly correlated. Theoretically, one
SDSSbasedsampleofdwarfgalaxiesdidnotfindanybreakinthe
wouldnotexpectthiscorrelationtoholdallthewaydowntodwarf
IbandTFrelation,although,inagreementwithearlierstudiesthey
findanincreasedscatterabouttheTFrelationatlowluminosities.
⋆ E-mail:[email protected] Further,asfirstnotedbyMcGaughetal.(2000)the“break”seen
2 Begum et al.
atthefaintendoftheBbandTFrelationisremovedifoneworks malizedtotheHolmbergradiusofthegalaxy.Column(10)theB-V
withthetotalbaryonicmass(i.e.thesumofstellarandgas)instead colour, Column (11)the velocity widthof theglobal HI profileat
oftheluminosity. 20%level,derivedfromtheGMRTdata,Column(12)thedistance
AccurateextensionsoftheTFrelationtofaintdwarfgalaxies tothegalaxyfromKarachentsevetal.(2004)andColumn(13)the
poses anumber of practical problems. Firstly, dwarf galaxies are method used to derive the distance e.g. the tip of the red giant
largely a field population, and very few of them have accurately branch(rgb),Tully-Fisherrelation(tf),theHubblerelation(h)and
knowndistances.Secondly,beingfaint,andlargelyoflowsurface themember ship of the group (grp). 22 of the 29 galaxies in our
brightness, accuratephotometryisdifficult.Finally,incontrastto samplehaveindependent distancesmeasuredusingtheTRGB.In
thesituationforbrightspirals,thepeakrotationalvelocitiesoffaint our analysis below, where appropriate, wetreat the galaxieswith
(MB > −14) dwarf galaxies is generally comparable in magni- independentdistancesseparatelyfromthosewhichdonot.
tudetothevelocitydispersionofthegas(e.g.Begumetal.2003; Thesourcesoftheabsolutemagnitudesofoursamplegalaxies
Begum&Chengalur 2004).Insuchasituationitisunclearifthe arelistedinthetable.Thephotometryforthosegalaxiesforwhich
velocitywidth(e.g.atthe20%level,W20,whichisthecommonly there isno separate notation is froman HST(WFPC2and ACS)
used indicator of the circular velocity in TF relations), is a good surveyofLocalVolumedwarfgalaxies.Thephotometryofgalax-
tracerofthecircularvelocity. ieswithlargeangulardiameterswasdoneusingSDSSimages.The
InthispaperwediscusstheTFrelationusingadatasetwhich HSTphotometrywasdoneusingtherecipesandequationsgivenin
addressalloftheaboveissues.Accurate(∼10%)TRGBdistances Holtzmanetal.(1995)andSiriannietal.(2005).Detailsonboththe
areknowntomostofthegalaxiesinoursample,andwealsouseac- HSTand SDSSbased photometry can befound in(Sharinaetal.
curateIbandmagnitudes.Finallyallofthegalaxieshavebeenob- (2007)).Photometryforthefewgalaxiesinoursamplethatarenot
servedinHIusingtheGiantMeterwaveRadioTelescope(GMRT) alsoinSharinaetal.(2007)wasdoneusingSDSSimagesandan
as part of the FIGGS survey (see Begumetal. (2008)). Pressure identicalproceduretothatfollowedbySharinaetal.(2007).Trans-
correctedrotationcurvesareavailableforallofthegalaxies;these formation to the standard Johnson-Cousins system was done us-
can be compared with W20 to see which is the better observable ingtheempiricalcolortransformationsgivenbyJordietal.(2006).
to use for TF studies. As a comparison sample we use the sam- Note that these magnitudes are those measured within a limiting
pleofbrightgalaxiesforwhichCephieddistancesweremeasured diameter(tabulatedinSharinaetal.(2007)).Inprincipleonecould
aspartoftheHSTkeyprojectontheextragalacticdistancescale insteaduseextrapolatedtotalmagnitudesderivedfromassumingan
(Sakaietal.(2000)). exponential stellar disk. However, in practice, the surface bright-
nessdistributionispoorly fitbyanexponential disk, andthereis
substantial uncertainty in the scale length and location of center
of thestellardisk.Consequently, theerrorsintheextrapolated to-
2 THEDWARFGALAXYSAMPLE
tal magnitudes are large, and, in fact, will agree within the error
ThedwarfgalaxysampleisapartofanHIimagingstudyoffaint barswiththoselistedinTable1.Wehenceusethemagnitudesas
dwarf galaxies with the GMRT − the Faint Irregular Galaxies tabulatedbySharinaetal.(2007).
GMRT Survey (FIGGS; Begumetal. (2008)). The FIGGS sam- As a comparison bright galaxy sample we use a sample of
ple is a systematically selected subsample of the Karachentsev 21galaxieswithCepheiddistancestakenfromSakaietal.(2000).
et al.(2004) catalog of galaxies within 10 Mpc (see Begumetal. Rotationcurvesareavailableinliteratureforonlyfewgalaxiesin
(2008) for detailson thesample selectionand observations). The thissample,hence,weusetheinclinationcorrectedvelocitywidths
FIGGS galaxies represent the extreme low-mass end of the dIrr at 20% level of the peak emission ( W20), given in Sakaietal.
population,withamedianMB ∼−13andamedianHImass∼3× (2000). The velocity widths were corrected for the line broaden-
107 M⊙.OurGMRTobservationsshowthat,contrarytothegen- ingduetotheturbulentvelocitydispersionusingtheprescription
eralbeliefthatextremelyfaintdIrrgalaxieshave“chaotic”veloc- giveninVerheijen&Sancisi(2001).Notethatalthoughthereexist
ityfields(Loetal.(1993)),mostofthemhavecoherentlargescale severalintermediatemassdwarfgalaxiessamples(e.g.Schombert
velocitygradients(seee.g.Begumetal.(2006)).Formanygalax- et al. 1997) that could have been used as templates, we have re-
ies, these large scale gradients are consistent with the systematic strictedouranalysisheretogalaxysampleswithaccurate,Hubble
rotation. Rotation curves were derived using a tilted ring model, flowindependentdistanceestimates.
details can be found in Begum et al.(2008, in preparation). For
thecurrentstudy,wehaveconsideredonlythosegalaxiesfromthe
FIGGSsamplewhicharewellresolved(i.e.aremorethan6inde-
3 RESULTSANDDISCUSSION
pendentbeamsacross)andforwhichtiltedringanalysisgavesen-
siblefitstothedata.Examplescanbeseenin,fore.g.Begumetal. Table2 liststhe correlationcoefficients between different indica-
(2003); Begum&Chengalur (2004); Begumetal. (2005, 2006). torsofthemassandthecircularvelocity.WeuseeitherVrotorW20
ThegalaxiesfromtheFIGGSsampleselectedforthecurrentstudy aftercorrectionforthelinebroadeningduetotheturbulentveloc-
aregiveninTable1.Thecolumnsareasfollows:Column(1)the itydispersion usingtheprescriptiongiveninVerheijen&Sancisi
nameofthegalaxy. Column(2)Vrot,therotationvelocityatthe (2001).Thestellarmass,M∗ wascomputedusingthestellarmass
lastmeasuredpointoftherotationcurve,correctedforthepressure to light ratio given by Belletal. (2003) for the case of a “diet”
support,(see e.g. Begumetal. 2003; Begum&Chengalur 2004). SalpeterIMF.Twoestimatesofthetotalbaryonicmasswereused;
Column(3)Vder,thederivedrotationvelocityatthelastmeasured the first (Mbar) was computed using the sum of the stellar mass
point of therotation curve, Column(4) Verr, theerror on thede- and 1.33 times the HI mass to correct for primordial He (Ott et
rived velocity, Column(5) the absolute B magnitude, Column(6) al.(2001);Coteetal.(2000)).Nocorrectionforthemoleculargas
theabsoluteImagnitude.Column(7)theHImassderivedfromthe was made, which is probably justifiable for the FIGGS sample,
GMRTdata,Column(8)theluminosityinBband,Column(9)the since dwarf galaxies do not seem to have a substantial reservoir
extent of the HI disk at a level of NHI = 1×1019cm−2, nor- of molecular gas (Taylor,Kobulnicky&Skillman (1998)). While
BTF RelationforExtremelyLow MassGalaxies 3
Table1.FIGGSdata
Galaxy Vrot Vder Verr MB MI MHI LB DDHHIo B-V W20 D Method
(kms−1) (kms−1) (kms−1) (mag) (mag) (106M⊙) (106L⊙) (mag) (kms−1) (Mpc)
UGC685 51.67 45.84 2.75 −14.31 −15.07 56.15 82.41 1.6 0.52 74.0 4.5 rgb
KKH6 27.98 19.93 3.66 −12.42 −13.53 10.18 14.45 2.9 0.43 40.0 3.73 rgb
KK14a 21.32 19.27 1.52 −12.13 −12.84 21.92 11.06 1.5 0.42 36.5 7.2 grp
KKH11a 56.29 52.39 3.14 −13.35 −14.06 52.87 34.04 4.2 0.40 106.0 3.0 grp
KK41 32.3 21.91 1.97 −14.06 −15.27 67.19 65.46 3.3 0.63 55.0 3.9 rgb
UGCA92 37.08 31.00 3.00 −15.65 −15.71 156.05 283.13 4.5 0.15 70.0 3.01 rgb
KK44 20.00 7.00 2.10 −11.85 −13.03 12.0 8.57 2.3 0.58 32.6 3.34 rgb
KKH34 20.11 10.50 1.38 −12.30 −13.23 10.44 12.94 2.6 0.28 25.0 4.6 rgb
UGC3755 20.19 13.05 2.73 −14.90 −16.00 73.99 141.90 1.7 0.55 49.0 6.96 rgb
DDO43 36.04 30.72 2.16 −14.75 −15.49 203.02 123.59 2.8 0.45 50.0 7.8 rgb
KK65 18.49 16.71 2.90 −14.29 −15.57 38.85 91.20 2.3 0.54 50.2 7.62 rgb
UGC4115 58.54 56.08 2.58 −14.27 −15.95 285.52 79.43 4.0 0.47 105.0 7.5 rgb
UGC5456 39.75 37.55 5.29 −15.08 −16.23 58.95 167.49 1.5 0.39 88.0 5.6 rgb
UGC6145a 40.90 36.94 2.90 −13.14 −13.85 27.02 28.05 1.6 0.40 52.0 7.4 h
NGC3741 46.70 46.70 2.50 −13.13 −13.94 157.99 27.79 8.8 0.37 101.0 3.0 rgb
UGC7242 39.90 35.11 3.03 −14.06 −14.77 45.75 65.46 2.1 0.40 74.0 5.4 rgb
KK144 27.82 22.38 3.95 −12.59 −14.32 81.14 16.90 3.1 0.38 49.0 6.3 h
DDO125 22.92 14.61 2.74 −14.16 −15.24 31.87 71.77 1.7 0.51 42.0 2.5 rgb
UGC8055 65.72 60.25 2.72 −15.49 −16.46 782.63 244.34 3.0 0.37 95.5 17.4 tf
UGC8215 19.95 12.37 4.00 −12.26 −13.30 21.41 12.47 3.5 0.38 31.0 4.5 rgb
UGC8508 40.10 31.17 2.30 −12.98 −13.99 29.07 24.21 3.3 0.46 64.5 2.6 rgb
DDO181 29.92 22.17 4.00 −13.03 −13.90 27.55 25.35 3.3 0.46 49.0 3.1 rgb
DDO183 24.66 21.35 1.85 −13.17 −13.97 25.90 28.84 2.7 0.42 48.0 3.24 rgb
UGC8833 26.68 20.65 3.01 −12.42 −13.27 15.16 14.45 2.3 0.42 40.0 3.2 rgb
P51659a 30.50 22.00 2.58 −11.83 −12.54 52.99 8.39 2.7 0.40 55.5 3.6 rgb
KK246 35.18 29.93 4.00 −13.69 −14.87 63.39 46.47 2.9 0.58 80.0 7.83 rgb
KK250b 50.00 50.00 3.00 −14.54 −16.14 121.0 120.26 3.2 0.91 110.0 5.6 grp
KK251b 38.00 34.00 4.00 −13.72 −14.32 78.0 44.05 2.6 0.37 79.0 5.6 grp
DDO210c 17.00 8.00 2.00 −11.09 −11.90 2.80 4.30 1.3 0.24 29.1 1.0 rgb
aIbandmagnitudecomputedusingtherelation(B-V)=0.85*(V-I)(Makarova(1999))assumingB-V=0.4andBmagnitudesfromKarachentsevetal.(2004).
bIbandmagnitudefromBegum&Chengalur(2004).cIbandmagnitudefromLeeetal.(1999).
themoleculargasprobablymakesasubstantialcontributiontothe FIGGSsamplethan for the combined HST+FIGGSsample. This
total gas mass in the case of the bright spirals in the Sakai et al. is partly because of the substantially larger baseline of the com-
(2000) sample, the baryonic mass of these systems is dominated bined sample, but as as shown below, it is alsopartly because of
bythestellarcontribution, andtheerror wemakeinignoringthe genuinely increased scatter in the low mass range. Finally, W20
molecular gas is likely to be subsumed by the uncertainty in the correlatesbetterwiththemassindicators thanVrot,although the
stellarmasstolightratio.Thesecondestimateofthebaryonicmass correlationcoefficientsoverlapwithintheerrorbars.
(Mbar1)usesthesameestimateofthetotalgasmassasbefore,but Thebestfitregressionoflog(MI)onlog(W20)computedus-
thestellarmasswascomputedfromBelletal.(2003)foraBottema
ingabivariateleastsquaresfitforthedatafromSakaietal.(2000)
IMF.ForagivenB-Vcolor,themasstolightratiofortheSalpeter
giveslog(MI)=-8.66(0.67)log(W20)+0.231(1.7)andisshown
“lite”IMF and BottemaIMF differ only by amultiplicative con-
in Figure 1 as a solid line. The error bars account for errors in
stant,so thecorrelationbetween thestellarmasscomputed using
both the distance measurement and the photometry. The two dif-
aBottemaIMFandthecircularvelocityindicatorsareidenticalto
ferent panels show where the FIGGSgalaxies lie with respect to
thatcomputedusingtheSalpeter“lite”IMF.ThesetwoIMFswere
thisline,inPanel[A]W20 isusedastheestimatorofthecircular
chosenasbeingrepresentativeoftwoextremesofthestellarmass
velocityandinPanel[B]oneusesVrot.Ascanbeseen,relativeto
to light ratio(Belletal. (2003)). The correlation coefficient error
theTFrelationforbrightgalaxies,theFIGGSgalaxiesareunder-
barswerecomputedusingbootstrapresampling.Thelistedvalues
luminous for their velocity width. However, the offset is signifi-
are for the entireFIGGS sample; if one uses only the subsample cantlylargerwhenoneusesVrot(0.93±0.06mag)thanwhenone
of22galaxieswithTRGBdistances,thecomputedcorrelationco- usesW20(0.31±0.06)mag.IfonedoesaregressionfitusingW20
efficientsagreewithintheerrorbarswiththetabulatedvalues.For
tothe entireFIGGS+HSTsample, thebest fit relationislog(MI)
the galaxies in the FIGGS sample alone, a few interesting points
= -8.77 (0.1) log(W20)+0.741(0.3). The scatter of the FIGGS
emerge.ThefirstisthateventhoughtheHIandstellarmassesof galaxies about this relation is ∼ 1.6mag, while that of the HST
thesegalaxiesarecomparable(themedianratiooftheHItostel- samplegalaxiesis∼0.36mag.ThescatterfortheFIGGSgalaxies
lar mass is∼ 1.25) theHI mass issignificantlybetter correlated ismuchlargerthanthetypical estimatedtotalmeasurement error
withthecircularvelocityindicatorsthanthestellarmass.Infact, of∼ 0.37mag,(fromtheerrorsinthedistancemeasurement,the
theHImasscorrelateswiththecircularvelocityindicatorsatleast
photometry and the error in the velocity width scaled by the TF
as well the baryonic mass estimates. The second isthat the error
slope) and indicates that essentially all of the observed scatter at
bars in the correlation coefficients are significantly larger for the
thefaintendisintrinsic.
4 Begum et al.
A B
Figure1.[A]TheabsoluteIbandmagnitudesplottedasafunctionofthelogoftherotationvelocityVrotfortheFIGGSsampleandW20(inclinationand
turbulencecorrected)forgalaxiesfromtheSakaietal.(2000)sample.[B]Sameas[A]exceptthattheabscissaislog(W20)forboththeFIGGSandSakaiet
al.(2000)samples.Ascanbeseen,relativetotheTFrelationforbrightgalaxies,theFIGGSgalaxiesareunder-luminousfortheirvelocitywidth.
Table2.CorrelationCoefficients A B
FIGGS FIGGS+HST
log(Vrot) log(W20) log(W20)
log(MI) -0.505(0.15) -0.629(0.10) -0.931(0.01)
log(M∗) 0.459(0.16) 0.596(0.11) 0.928(0.02)
log(MHI) 0.683(0.10) 0.733(0.07) 0.910(0.01)
log(Mbar) 0.631(0.12) 0.732(0.08) 0.942(0.01)
log(Mbar1) 0.665(0.11) 0.741(0.07) 0.943(0.01)
The systematic deviation from and the increased scatter Figure3.ScatteraboutthecombinedbestfitBTFfortheFIGGS(Panel
around, theTF relation set by bright galaxies, impliesthat dwarf [A])andHST(Panel[B])galaxies.TheBTFusedtogeneratethisfigureis
galaxies have been relatively less efficient at forming stars, and thatbetweentheBaryonicmassderivedfromtheSalpeter“lite”IMFand
thattherehasalsobeconsiderablymorestochasticityintheirinte- log(W20). Different choices ofIMF orcircular velocity indicator donot
qualitatively affecttheplot.Thescatterhasbeenscaledbythecombined
gratedstarformationefficiencies.Leeetal.(2007)findthatdwarf
measurementerrors.
galaxies(i.ewithMB >−15)showalargerscatterinthespecific
star formation rate as compared to bright galaxies. They suggest
that this could be because the processes that regulate star forma- deviatesmuch lessfromthisrelationthatitdoesfromtheIband
tion in bright spirals are not operative in dwarfs, consistent with TFrelation.TheaveragedeviationoftheFIGGSgalaxiesfromthe
the finding of Begumetal. (2006) of a lack of asimple relation- BTFrelationsetbybrightspiralsis-0.16(0.02),-0.11(0.02)forthe
shipbetweengascolumndensityandstarformationrateindwarf Salpeter“lite”andBottemaIMFsandusingVrotasthecircularve-
galaxies.Fig.1indicatesthatevenintegrationovertimeisnotsuf- locityindicator,and0.09(0.02),0.09(0.02)forthesametwoIMFs
ficienttoremovethisscatter,i.e.eventheintegratedstarformation but with W20 as the circular velocity indicator. Clearly, whether
rateofdwarfgalaxiesshowssubstantiallymorescatterthanthatof one sees a deviation from the BTF at the faint end seems domi-
brightspirals.InarecentstudyusingasampleofSDSSgalaxies, natedbysystematicuncertainties(IMF,choiceofcircularvelocity
Gehaetal.(2006)foundasimilarincreaseinscatteraroundtheI indicator).Ontheotherhand,theincreasedscatterabouttheBTF
bandTFrelationatlowluminosities.However,incontrasttowhat atthefaintmassendisindependentofthechoiceofIMForcircular
wefindheretheydonotfindthatfaintgalaxiesaresystematically velocityindicator.Fig.3)comparesthescatteraboutthecombined
under-luminous.Thereasonforthisdiscrepancyisunclear,though best fit BTF for the HST and FIGGS galaxies, the difference in
wedonotethatweareusinggalaxieswithaccuratelymeasureddis- scatterisstriking.DifferentchoicesofIMForcircularvelocityin-
tances,whileGehaetal.(2006)usedHubbleflowbaseddistances dicatordonotqualitativelychangetheplot.
fortheirgalaxies. As discussed above, systematic uncertainties limit the mea-
As mentioned in the introduction, the deviation of dwarf surement of the BTF. Numerous authors (e.g. Pfenniger&Revaz
galaxiesfromtheBbandTFrelationwasoneoftheprimarymoti- (2005); McGaugh,S. (2005) have tried to approach this problem
vationsforsearchingforarelationbetweenthetotalbaryonicmass alongthereversedirection,viz.tousetheBTFitselftodetermine
andthevelocitywidth,i.e.thebaryonicTullyFisher(BTF)relation one or both of the scaling of the HI mass to the total gas mass
(McGaughetal.(2000)).TheBTFrelationfortheHSTsampleis and the stellar mass to light ratio. In this approach, one assumes
showninFig.2asasolidline.Ascanbeseen,theFIGGSsample that an intrinsic BTF relation exists, and that the right choice of
BTF RelationforExtremelyLow MassGalaxies 5
A B
Figure2.ThesolidlinesshowtheBTFforthebrightHSTsample,Panel[A]isforaSalpeter“lite”IMF,whilePanel[B]isforaBotemmaIMF.Thehollow
circlesareforFIGGSgalaxieswithoutTRGBdistances,thefilledcirclesareFIGGSgalaxieswithTRGBdistancesandthefilledsquarearegalaxiesfromthe
HSTsample.
stellar mass to light ratio and/or HI mass scaling will minimize
the scatter of the observed data about the relationship. Pfenniger
&Revaz(2005)foundthattheBTFrelationisoptimallyimproved
whentheHImassismultipliedbyafactor of3; however theac-
tualbestfitvalueispoorlyconstrained,andscalefactorsofupto
∼11leadtoatighteningoftheBTF.Allowingafreescalingofthe
HImassispromptedbymodelswhichhaveadarkbaryoniccom-
ponentwhosedistributioniscorrelatedwiththatofHI(e.gHoek-
straetal.2001).Thesemodelsinturnarebasedonthepossibility
thatsomeofthedarkmatterinthehalosofgalaxiesmaybeinthe
formofcold,densegasclouds(Gerhard&Silk(1996); Pfenniger
& Combes (1994)). Many observations based on dynamical and
stabilityanalysisindicatethatgalacticdisksmaybemoremassive
thaninferredfromitsluminousmass(Fuchs,B.(2003);Masset&
Bureau(2003);Pfenniger&Revaz(2005)).Further,fromtheoret-
icalmodellingandN-bodysimulations,Pfenniger&Revaz(2004)
foundthatifdisksaremarginallystablewithrespecttobendingin-
stabilities,themasswithintheHIdisksmustbeamultipleofthat Figure4.ModelswithHImassscalefactorasafreeparameter.Thegraph
showsχ2 asafunctionoftheHImassscalefactor. Theminimumχ2 is
detectedinHIandstars.WenotethatsincethestellarmassandHI
massarehighlycorrelated(correlationcoefficient∼0.91between foramassscalefactor∼ 6.7.Thehorizontallineshowsthethresholdχ2
fora95%significancethatthemodelwithavariableHImassscalefactor
log(Mstar) and log(MHI), simultaneous minimization of both the providesabetterfittothedata.
stellar mass to light ratio and the scale factor for the HI mass is
degenerate. So instead we first allow the HI mass scale factor to
varykeepingthemasstolightratioofthestellardiskfixed(Fig.4)
LargevaluesofFindicateastatisticallysignificantimprove-
andthenallowthestellarmasstolightratiotovarykeepingtheHI mentinthefit.ApplyingtheF-testtotheχ2 numbersaboveindi-
massscalefactorfixed(Fig.5).InthecaseofavariableHImass
scalefactor(Fig.4),thereisabroadminimumintheχ2,arounda catesthatthisdecreaseissignificantatthe∼2%level,i.e.thereisa
scalefactorof∼ 6.7. Theminimumχ2 is357.2,ascomparedto 2%probabilitythataHIscalefactorof6.7doesnotprovideabetter
fittothedata.ThehorizontallineinFig.4indicatesthethreshold
402.3forthecaseinwhichtheHImassscalefactoriskeptat1.33. valueof χ2 for a5% significance. Ascanbe seenthisleadstoa
Sincethenumberofparametershasbeenincreased,adecreasein
theχ2istobeexpected.Thesignificanceofthedecreasecanbees- relativelybroadrangeinthescalefactor,i.e.from∼ 3.0−13.5.
IfoneexcludestheFIGGSgalaxieswithoutTRGBdistances,the
timatedusingtheF-test.Fornestedmodelswithν1andν2degrees minimumχ2 occurs atanHImassscalefactor of ∼ 5andhasa
opfarfarmeeedtoerms rinestpheectsiveceolynd(pm=odνe1l)−anνd2χ)b2evinaglutehseχe1x2traanndumχb22erroef- ∼5%significance.
FormodelswithafixedHImassscalefactorof1.33butvari-
spectively,theF-testlooksatthevalue
able stellar mass to light ratio, we follow the parametrization of
Belletal.(2003)andassumethatthestellarmasstolightratiois
(χ12−χ22)/p givenbylog(Γ∗) = α+β(B−V),whereαandβ arefreepa-
F = χ22/ν2 rameters.Fig.5showscontoursoftheresultingχ2.Theminimum
6 Begum et al.
Figure 6. Models in which the scale factor of the baryonic mass of the
galaxies in the FIGGS sample is a free parameter. The graph shows χ2
asafunctionofthebaryonicmassscalefactor.Theminimumχ2isfora
massscalefactor∼ 9.2.Thesolidhorizontallineshowsthethresholdχ2
Figure5.Contours ofχ2 asafunction ofthestellarmasstolightratio,
fora95%significancethatthemodelwithavariableHImassscalefactor
log(Γ∗)=α+β(B−V).ThescalefactorfortheHImasshasbeenkept providesabetterfittothedata.Thedashedhorizontallineisthethreshold
fixedat1.33.Thecontourlevelsstartat358andareuniformlyspacedin χ2fora99%significance.
stepsof20.
χ2(356.3)correspondstoα=−1.51,β =1.20.Forcomparison,
α = −0.549,β = 0.824 corresponds totheSalpeter “lite”IMF
and α = −0.749,β = 0.824 corresponds to the Bottema IMF.
The decrease in the χ2 however is significant only at the ∼ 7%
level(ascomparedtothe∼ 2%levelforavariableHImassscale
factor).
Since the HI mass forms a much larger fraction of the to-
tal baryonic mass in the case of galaxies from the FIGGS sam-
ple, the greater sensitivity of the goodness of fit to the HI mass
scale factor suggests that models in which one has a scale factor
forthetotalbaryonicmassofdwarfgalaxieswouldprobablysig-
nificantlyimprovethefit.Whilethis(liketheearliertwomodels)
isaphenomenological model, it would correspond togalaxy for-
Figure7. Baryonic Tully Fisherrelations. Thehollow points are forthe
mation models in which dwarf galaxies have a larger fraction of
FIGGSgalaxieswithstellarmasstolightratiocomputedusingaSalpeter
“dark” (e.g. highly ionized) or missing baryons. For e.g. models
“lite”IMF.Forthesolidpointsthetotalbaryonicmasshasbeenincreased
which account for the reheating and ionization of the gas during by a factor of 9.2. For the HST galaxies (the cluster of points at large
theepochofrecombinationpredictthatlargegalaxiesshouldhave log(W20)) the baryonic mass is not scaled. The two lines show the best
essentiallythecosmicbaryonfraction,whilesmallgalaxies(with fitBTFforthesetwocases.
circular velocity in the ∼ 20 km/s range) could lose up to 90%
oftheirbaryons.Ofcourse,insuchasituation,whenapplyingthe
BTFrelationwiththeobservedvelocitywidths,wehavetomake galaxieshasbeenscaledbyafactorof9.18isshowninFig.7.For
thefurtherassumptionthatthebaryonlossdoesnotaffecttheob- reference, the original BTF relationwithout this rescaling is also
servedvelocitywidth.Theχ2curveforamodelinwhichthescale shown.Rescalingbyafactorof9.18impliesthatover90%ofthe
factorofthebaryonicmassoftheFIGGSgalaxiesisafreeparam- originalbaryonicmassis“dark”or“missing”.
eter(butthescalefactorforthebaryonicmassoftheHSTsample Tosummarize,wepresentTFrelationsforasamplecombin-
iskeptat1.0)isshowninFig.6.Onceagain,theχ2curveshowsa ingextremelyfaintdwarfirregulargalaxiesfromtheFIGGSsurvey
broadminimum,howevertheminimumχ2(289.6,forascalefac- andlargespiralgalaxies.FortheFIGGSsamplealone,wefindthat
torof9.18)issubstantiallylowerthaninthecasethatonlytheHI thoughthemedianHImassandstellarmassarecomparable,theHI
masswasrescaled.TheF-testgivesaprobabilityof∼ 2×10−4 masscorrelateswiththevelocitywidthmuchbetterthanthestellar
thatthisreductioninχ2isbychance.ThehorizontallineinFig.6 mass.Distancesareaccuratelyknownforthebulkofthegalaxies
indicatesthethresholdvalueof χ2 fora1%significance. Ascan in our sample, and we find that the extremely faint dwarfs show
be seen this leads to a relatively broad range in the scale factor, substantialintrinsicscatteraboutboththeIbandTFandtheBTF
i.e.from∼ 2.0−28.8. IfoneincludesonlythoseFIGGSgalax- relations.DeviationsfromtheBTFatthefaintendaresensitively
ieswithTRGBdistances,theminimumχ2 occursatamassscale dependentonsystematics,e.g.theassumedstellarmasstolightra-
factor of ∼ 7.56 and corresponds to a significance of ∼ 10−3. tio. If one assumes that there isan intrinsic baryonic TF relation
Thebestfit“BTF”relationafterthebaryonicmassoftheFIGGS onecouldtrytodeterminethetotalbaryonicmassbysearchingfor
BTF RelationforExtremelyLow MassGalaxies 7
theprescriptionthatleadstothetightestBTF.Wefindthatchang- PfennigerD.&RevazY.,2005,A&A,431,511.
ingthescalefactorofthetotalHImassleadstoamoresignificant Sakai,A.,Jeremy,M.R.etal.,2000,ApJ,529,698
tightening of the BTF than changing the stellar mass to light ra- SharinaM.E.,etal.,2007,arXiv,712,arXiv:0712.1226
tio. Finally, the most significant tightening that we find is when Schombert, J. M.,Pildis,R.A. &Eder, J.A.,1997, ApJS,111,
weallowtheentirebaryonicmassofthefaint(butnotthebright) 233
galaxiestoscalebyaconstantfactor.Thisbestfitrelationisfora Sirianni,M.,etal.,2005,PASP,117,1049
scalefactor ∼ 9. Theexact value ishowever poorly constrained, Tassis, K., Kravtsov, A. V. & Gnedin, N. Y., 2007,
anyvaluebetween∼ 2−28.8leadstoasignificanttighteningof astro-ph/0609763v2
theBTF. Taylor, C. L., Kobulnicky, H. A. & Skillman, E. D. 1998,
A.J.,116,2746-2756.
VerheijenM.A.W.,2001,ApJ,563,694
Verheijen,M.A.W.&Sancisi,R.,2001,A&A,370,765
ACKNOWLEDGMENTS
TheobservationspresentedinthispaperweremadewiththeGiant
MetrewaveRadioTelescope (GMRT).TheGMRTisoperated by
theNationalCenterforRadioAstrophysicsoftheTataInstituteof
FundamentalResearch.
REFERENCES
Begum, A., Chengalur, J.N., Karachentsev I. D., Sharina, M. &
Serafim,S.2008,MNRASsubmitted.
Begum, A., Chengalur, J.N., Karachentsev I. D., Sharina, M. &
Serafim,S.2006,MNRAS,365,1220.
Begum, A., Chengalur, J.N. & Karachentsev I. D. 2005, A&A,
433,L1
Begum,A.,Chengalur,J.N.&Hopp,U.,2003,NewAstronomy,
8,267
Begum,A&Chengalur,J.N.,2004,A&A,424,509
Bell,E.F.,McIntosh,D.H.,Katz,N.&Weinberg,M.D.,2003,
ApJSS,149,289
Cote,S.,Carignan,C.&Freeman,K.C.,2000,AJ,120,3027
Dekel,A.&Woo,J.,2003,MNRAS,344,1131
Efstathiou,G.2000,MNRAS,317,697
Fuchs,B.2003,Ap&SS,284,719
Geha,M.,Blanton,M.R.,Masjedi,M.&West,A.A.,2006,ApJ,
653,240
Gerhard,O.&Silk,J.,1996,ApJ,472,34
GnedinN.Y.,2000,ApJ,542,535
Hoekstra, H., van Albada, T. S. & Sancisi, R., 2001, MNRAS,
323,453
Holtzman, J. A., Burrows, C. J., Casertano, S., Hester, J. J.,
Trauger,J.T.,Watson,A.M.&Worthey,G.,1995,PASP,107,
1065
Jordi,K.,Grebel,E.K.&Ammon,K.,2006,A&A,460,339
Karachentsev I. D., Karachentseva V. E., Huchtmeier W. K. &
MakarovD.I.,2004,AJ,127,2031
Lee M. G.,Aparicio A.,Tikonov N., Byun Y.-I.,KimE.,1999,
AJ,118,853
LeeJ.C.,KennicuttR.C.,FunesJ.G.,J.S.,SakaiS.,Akiyama
S.,2007,arXiv,711,arXiv:0711.1390
Lo,K.Y.,Sargent,W.L.W.&Young,K.,1993,AJ,106,507
MakarovaL.,1999,A&AS,139,491
Masset,F.S.&Bureau,M.,2003,ApJ,596,152
McGaugh,S.S.,2005,ApJ,632,859
McGaugh,S.S.,Schombert,J.M.,Bothun,G.D.&deBlok,W.
J.G.,2000,ApJ,533,99
Ott,J.,Walter,F.,Brinks,E.,VanDyk,S.D.,Dirsch,B.&Klein,
U.,2001,AJ,122,3070
PfennigerD.&CombesF.,1994,A&A,285,94.
PfennigerD.&RevazY.,2004,ASSL,319,307.