Mon.Not.R.Astron.Soc.000,1–5(0000) Printed20January2011 (MNLATEXstylefilev2.2) Evidence of major dry mergers at M > 2 × 1011M from ∗ ⊙ curvature in early-type galaxy scaling relations? 1 1 Mariangela Bernardi1⋆, Nathan Roche2, Francesco Shankar3 & Ravi K. Sheth1,4 0 1Department of Physics & Astronomy, Universityof Pennsylvania, 209 S. 33rd St., Philadelphia, PA 19104, USA 2 2 Dipartimento di Astronomia, Universita´degli Studi di Bologna, viaRanzani 1, I-40127 Bologna, Italy n 3 Max-Planck-Institu¨t fu¨rAstrophysik, Karl-Schwarzschild-Str. 1, D-85748, Garching, Germany a 4 Centerfor Particle Cosmology, Universityof Pennsylvania, 209 S. 33rd St., Philadelphia, PA 19104, USA J 9 1 20January2011 ] O ABSTRACT C For early-type galaxies, the correlations between stellar mass and size, velocity dis- . persion, surface brightness, color, axis ratio and color-gradient all indicate that two ph mass scales, M∗ =3×1010M⊙ and M∗ =2×1011M⊙, are special. The smaller scale could mark the transition between wet and dry mergers, or it could be related to the - o interplaybetweenSNandAGNfeedback,althoughquantitativemeasuresofthistran- r sition may be affected by morphological contamination. At the more massive scale, t s meanaxis ratiosandcolorgradientsaremaximal,andaboveit, the colorsareredder, a thesizeslargerandthevelocitydispersionssmallerthanexpectedbasedonthescaling [ at lowerM∗. In contrast,the color-σ relation,and indeed, most scaling relations with 2 σ, are notcurved:they are well-describedby a single power law,or in some cases,are v almostcompletely flat.When majordrymergerschangemasses,sizes,axis ratiosand 1 color gradients, they are expected to change the colors or velocity dispersions much 0 less. Therefore, the fact that scaling relations at σ > 150km s−1 show no features, 5 whereas the size-M∗, b/a-M∗, color-M∗ and color gradient-M∗ relations do, suggests .1 that M∗ = 2×1011M⊙ is the scale above which major dry mergers dominate the 1 assembly histories of early-type galaxies. 1 0 Key words: galaxies: formation 1 : v i X 1 INTRODUCTION (2009). We show that the curvature in the color−M∗ re- ar Recent work (Bernardi et al. 2010b) has shown that the lation coincides with curvature in other relations with M∗. However,whenM∗isreplacedwithvelocitydispersion,then color-magnitude relation of early-typegalaxies in theSDSS there is little curvature. Although curvature in scaling re- differs significantly from a pure power-law, curving down- lations does not imply a change in the physics which sets wardsatlowandupwardsatlargeluminosities(Mr >−20.5 therelations(e.g.Graham&Guzman2003; Graham2010), andMr <−22.5,respectively).Thisisalsotrueofthecolor- we argue that our findings suggest that major dry mergers size relation, and is even more apparent with stellar mass, where thecorresponding mass scales are M∗ <3×1010M⊙ dominate themass growth at M∗ >2×1011M⊙. and M∗ > 2×1011M⊙, respectively. The upwards curva- ture at the massive end does not appear to be due to stel- Our results are based on two different ways of select- larpopulationeffects.Curvatureinthecolor-luminosity (or ing early-type samples from the SDSS database. One fol- color-M∗) relation at the faint end was noticed before (e.g. lows Hyde & Bernardi (2009): the image must be round Graham 2008; Skelton et al. 2009); the curvature at bright (b/a > 0.6) and the light profile shape must be well-fit by end is the newfindingof Bernardi et al. (2010b). a deVaucoleurs profile (fracDev = 1). The other is a sim- Curvature at the low mass end using other parameters ple cut on how centrally concentrated the surface bright- (e.g.surfacebrightness)hasbeennoticedbeforeaswell(e.g. ness is (Cr > 2.86). The former method produces a sam- Kauffmann et al. 2003; Shankaret al. 2006); the subject of ple that is more purely elliptical; the latter contains many this Letter is to analyse non-linear scaling relations at the edge-on disks. See Bernardi et al. (2010a) for a more de- high mass end, extending the analysis of Hyde & Bernardi taileddiscussion oftheseselection criteria,andoftheSDSS photometric and spectroscopic parameters which we use below. Where necessary, we have assumed a spatially flat ⋆ E-mail:[email protected] background cosmology with energy density dominated by 2 M. Bernardi et. al. Figure 1. Curvature in the correlations between stellar mass and (from top to bottom, left) size, velocity dispersion and surface brightness,andcolor,axisratio,andcolorgradient(toptobottom,right),intheHyde-Bernardisample.Theverticaldashedlinesmark thescaleswheresomeoftherelationschangeslope:M∗=3×1010M⊙and2×1011M⊙,whichcorrespondapproximatelytoMr =−20.5 and −22.5. Scalings in a sample selected to have Cr > 2.86 are shown only where they differ from the scalings in the Hyde-Bernardi sample. a cosmological constant Λ = 0.7, with a Hubble constant the relations curve towards larger sizes, smaller than ex- H0=70 km s−1 Mpc−1 at thepresent time. pectedvelocitydispersions,faintersurfacebrightnesses(left panels), and smaller axis ratios and smaller color-gradients (rightpanels).Thisispreciselythemassscale onwhich the 2 CURVATURE IN RELATIONS WITH M∗, color-M∗ relation curves towards redder colors (top right panel). BUT NOT WITH σ Figure 1 shows correlations between stellar mass and (from Figure 2 shows that when M∗ is replaced with velocity dispersion, then there is little curvature at logσ/kms−1 > top to bottom, left) size, velocity dispersion and surface 2.2. In fact, the correlations with surface brightness, color brightness, and color, axis ratio, and color gradient (top gradient and axis ratio are almost completely flat. (That to bottom, right), in the Hyde-Bernardi sample. (We dis- surface brightness and σ are uncorrelated was noted by cuss how we define the gradient in Section 3.2.) None of Bernardi et al. 2003.) The fact that there is no feature at thesecorrelations arepurepowerlaws. Althoughthecurva- thelargest σ inanyoftheserelations, despiteclearfeatures tfauirnet,atlolwogm10aMss∗/enMd⊙(M<r10≥.5−is2i0n.5te,rleosgti1n0g(M–∗g/aMla⊙x)ies≤a1t0t.h5e) in the scalings with M∗, is what has motivated thisLetter. tend to curve towards bluer colors, larger sizes, fainter sur- Inthiscontext,itisimportanttonotethattherelation face brightnesses, smaller axis ratios and color gradients – between Mdyn ∝ Rσ2 and luminosity or M∗ is very well inwhatfollows, wewillfocusonwhatappearstobeatran- described by a single power-law over the entire range: the sition massscaleathighermasses.Atlog10M∗/M⊙ >11.3, curvature in the sizes and velocity dispersions cancel (Fig- Major dry mergers at M > 2×1011M ? 3 ∗ ⊙ Figure 2. Curvature in the correlations between velocity dispersion and (from top to bottom, left) size, stellar mass and surface brightness, and color, axis ratio, and color gradient (top to bottom, right), in the Hyde-Bernardi sample. The dashed lines mark the scales where one would expect to see a change in the slope of the relations based on Figure 1. Scalings in a sample selected to have Cr >2.86areshownonlywheretheydifferfromthescalingsintheHyde-Bernardisample. ure 3). Presumably, this is because the objects we observe for axis-ratios and color-gradients, and, significantly, that are virialized, whatever their merger histories. thecurvatureis absent when M∗ is replaced by σ. 3.1 Axis-ratios 3 DISCUSSION The b/a − M∗ relation (right center panel of Figure 1) Major dissipationless mergers are expected to change the deserves further comment. Van der Wel et al. (2009) re- sizes in proportion to the masses, but to leave the velocity port that the width of the b/a distribution changes at dispersions and colors unchanged. In contrast, minor dissi- log(M∗/M⊙) ∼ 10.5. They interpret this as evidence that, pationless mergers producelarger fractional changes in size abovethismass,assemblyhistoriesaredominatedbymajor thaninmass,anddecreasethevelocitydispersionsandcol- mergers. Ourresults suggest thisis not thefull story. ors (see Appendix C in Bernardi et al. 2010b for details). InFigure1wehaveshowntwoversionsofthisrelation, Therefore,thecurvatureinthecorrelationsbetweenM∗and because the Hyde-Bernardi selection requires b/a > 0.6. In other parameters as size, σ and color, which have been no- thissample,b/adecreasesatlog(M∗/M⊙)>11.3.However, ticed before, have all been discussed in this context (e.g. noticethatthisdecreaseisevenmoremarkedinthesample Davies et al. 1983; Matkovic & Guzman 2005; Bernardi et selected to have Cr >2.86, where no cut on b/a is applied. al. 2007; Hyde & Bernardi 2009; Bernardi et al. 2010a,b). Compared to the Hyde-Bernardi sample, this sample has What is new here is the recognition that these all occur at considerablysmallerb/aatsmallM∗.Bernardietal.(2010a) the same mass scale, that this mass scale is also important show that this is primarily due to an increased incidence 4 M. Bernardi et. al. of disks and contamination by Sas, because the Cr > 2.86 sample is not as purely elliptical/early-type as the Hyde- Bernardi sample. We believe this change in morphological mixistheprimaryreason whyVanderWelet al.sawwhat they did. Webelievetherealfeatureofinterestisthedropinb/a at log(M∗/M⊙) > 11.3 where (Bernardi et al. 2010a show that)morphologicalmixisnolongeranissue.VanderWelet al.alsoseethisdrop,buttheydismissit.Instead,webelieve the narrowing of the distribution at log(M∗/M⊙) ∼ 10.5 marks the transition from dissipational to dissipationless histories,orachangeinrelativeimportanceofSNandAGN feedback (e.g. Kauffmann et al. 2003; Shankar et al. 2006), while the decrease in b/a at log(M∗/M⊙)>11.3 marks the transitiontomajordrymergers.Thisdecreasehasbeenex- pected for some time (see Gonz´alez-Garc´ıa & van Albada 2005; Boylan-Kolchin et al. 2006; Ragone-Figueroa & Plio- nis 2007; Ragone-Figueroa et al. 2010) – it was first found byBernardiet al.(2008). Thisisthoughttoindicatean in- creasingincidenceofmajorradialmergers,sincethesewould tend to result in more prolate objects. 3.2 Color gradients The right bottom panel of Figure 1 shows that color- gradients — here defined to be the difference between the model and Petrosian colors — are maximal at log(M∗/M⊙) ∼ 11.3. (The former are approximately the colorwithinthehalf-lightradius,whereasthelatteraremore closely relatedtotheratioofthetotalluminositiesing and r, so correspond to larger scales.) This is consistent with Figure B1 of Bernardi et al. (2010b), who used the same definition of color gradient. It is also consistent with Roche Figure 3. No curvature in the correlations between dynamical etal.(2010),whousedadifferentestimatorofthegradient: massandluminosity(top)andstellarmass(bottom). the ratio of the half-light sizes in the g and r bands. As we discuss below, we believe that the appearance of this same massscaleisagainsignalingtheonsetofmajordrymergers. Mergers are not the only way to produce or alter color Whereas major mergers are expected to decrease color gradients. In some models, gradients are related to feed- gradients(e.g.DiMatteoetal.2009),minormergersshould back and winds (Pipino et al. 2010). Our demonstration not change the gradients significantly (Kobayashi 2004) or that gradients scale differently with M∗ than with σ may theymayenhancethemslightly.Thisisbecausethesmaller have interesting implications for such models. In addition, bluer object involved in the minor merger is expected to producingthedownturnweseeatlog(M∗/M⊙)>11.3isan depositmostofitsstarsatlargerdistancesfromthecenterof interesting challenge for such models, as is relating this to theobjectontowhichitmerged.(Ifithaditsowngradient, the changes in size, σ and b/a we have found. Because the thenthebluestofitsstarswouldhavebeendepositedatthe major dry merger model provides a simple framework for largest radii.) understanding all these relations, our results suggest that As a simple check of this argument, note that major M∗ >2×1011M⊙ isthescaleabovewhichmajordrymerg- mergers, which double M∗, do not change σ (Appendix C ers dominate theassembly history. ofBernardietal.2010b).Therefore,aplotofcolorgradient versus σ should show less of a feature than when gradients areplottedversusM∗.ThisisindeedwhatweseeinFigure2 4 IMPLICATIONS (right bottom panel) – the correlation is almost flat at σ > 150km s−1. Such a plot should also show greater scatter, Wehavefoundthatavarietyofearly-typegalaxyscalingre- since a range of merger histories, hence gradients, can all lations–thesize-M∗,b/a-M∗,color-M∗ andcolorgradient- have the same σ. While we do see this increase in scatter, M∗ relations–allshowdeparturesfromapurepower-lawat notethatthescatterinthegradient-M∗ relationgrowseven M∗ =2×1011M⊙,whereasthereisnosuchfeaturewhenM∗ moredramatically –somethingthat isnot easily explained. isreplacedwithσ.Sincemajordrymergersareexpectedto On the other hand, the major merger picture also provides change the sizes, axis ratios and color gradients of galaxies a natural explanation for why none of the scaling relations while leaving the velocity dispersion and color unchanged, in Figure 2 show any feature at logσ/km s−1 > 2.2, and ourfindingssuggestthatthetotalstellarmassinearly-types those that are most clearly sensitive to merger histories are with M∗ > 2×1011M⊙ today must have grown primarily almost completely flat. by relatively recent major dry mergers. In Bernardi et al. Major dry mergers at M > 2×1011M ? 5 ∗ ⊙ (2010b), we argue that such mergers may be required to MBthanksMeudonObservatory,andRKSthankstheIPhT reconcilethez∼1countsofobjectswithM∗ >2×1011M⊙ at CEA-Saclay, for their hospitality during the course of with those at z∼0. this work. MB is grateful for support provided by NASA Thisparticularmassscalealso appearsinanalysesofa grant ADP/NNX09AD02G; FS acknowledges support from local sample (higher quality data, but significantly smaller theAlexandervonHumboldtFoundation;RKSissupported sample), where it is identified with the transition to dry in part by nsf-ast 0908241. mergers (see page 270 and related discussion in Kormendy et al. 2009). It is special in hierarchical models also. Figure 3 of Guo & White (2008) shows that below 1.6×1011M⊙ REFERENCES star-formationhasbeenasignificantpartofthemeanstellar BernardiM.,etal.2003,AJ,125,1849 massgrowthrate(muchofitthroughwetmergers),whereas BernardiM.,HydeJ.B.,ShethR.K.,MillerC.J.,NicholR.C. the stellar mass growth at masses above this occurs only 2007,AJ,133,1741 through dry mergers. See Hopkins et al. (2008) and Eliche- Bernardi,M.,Hyde, J.B.,Fritz,A.,Sheth, R.K.,Gebhardt, K. Moral et al. (2010a,b) for other arguments suggesting dry &Nichol,R.C.2008,MNRAS,391,1191 mergers since z∼1 are a natural and necessary part of the BernardiM.,2009,MNRAS,395,1491 assemblyhistoryatM∗ >2×1011M⊙.Whileitisreassuring Bernardi M., Shankar, F., Hyde, J. B., Mei, S., Marulli, F. & that many lines of study all identify this same mass scale, Sheth,R.K.2010a, MNRAS,404,2087 wefeelitworthemphasizingthatouranalysissuggeststhat BernardiM.,RocheN.,ShankarF.,ShethR.K.,2010b,MNRAS, abovethis mass scale, themergers were not just dry –they submitted(arXiv:1005.3770) were major. In this respect, there is some tension between Bezanson R., van Dokkum P. G., Tal T., Marchesini D., Kriek M.,FranxM.,CoppiP.,2009,ApJ,697,1290 ourconclusions,andrecentworkwhicharguesthatalthough Boylan-Kolchin M.,Ma C.-P.,Quataert E., 2006, MNRAS, 369, the mass in the central kpc or so of early-type galaxies has 1081 notgrown sincez∼2,thehalf-light radiihaveincreasedby Coccato, L., Gerhard, O. & Arnaboldi, M. 2010, MNRAS, 407, more than a factor of two. This suggests that, since z < 2, L26 mass has been added to the outer regions only: this sort of CookM.,LapiA.&GranatoG.L.,2009,MNRAS,397,534 inside-outscenarioforthegrowthismosteasilyunderstood Davies, R. L., Efstathiou, G., Fall, S. M., Illingworth, G. & ifthemergerswereminor(e.g.Lapi&Cavaliere2009;Cook Schechter, P.L.1983,ApJ,266,41 et al. 2009; Bezanson et al. 2009). However, as Tiret et al. DiMatteoP.,PipinoA.,LehnertM.D.,CombesF.,SemelinB., (2010) note,theobservation ofconstant massinthecentral 2009,A&A,499,427 regions does not, by itself, exclude major mergers. In the Eliche-Moral,M.C.,Prieto,M.,Gallego,J.,Barro,G.,Zamorano, J., L´opez-Sanjuan, C., Balcells, M., Guzm´an, R. & Mun˜oz- simulations of Gao et al. (2004), as an object assembles its Mateos,J.C.2010, A&A,519,55 mass through major dry mergers, the mix of particles in Eliche-Moral,M.C.,Prieto,M.,Gallego,J.&Zamorano,J.2010, the central regions can change dramatically, even though ApJ,submitted(arXiv:1003.0686) thetotal mass in the central regions remains constant. Our Gao, L.,Loeb, A.,Peebles, P.J.E.,White, S.D.M.&Jenkins, finding that color-gradients are erased at large masses may A.2004,ApJ,614,17 be indicating that this is indeed what happens at M∗ > Gonz´alez-Garc´ıaA.C.,vanAlbadaT.S.,2005,361,1043 2×1011M⊙. Graham,A.W.&Guzm´an, R.2003,AJ,125,2936 Finally, it is interesting to ask how BCGs, which are GrahamA.W.,2008,ApJ,680,143 amongst the most massive objects in the local universe, fit GrahamA.W.,2010,(arXiv:1009.5002) into this picture? Compared to non-BCGs of similar mass Guo,Q.&White,S.D.M.2008, MNRAS,384,2 Hopkins, P. F., Cox, T. J., Keres, Dusan & Hernquist, L. 2008, or luminosity, their colors are slightly redder (Roche et al. ApJS,175,390 2010;Figure10inBernardietal.2010b),theyhavesmaller HydeJ.B.,BernardiM.,2009, MNRAS,394,1978 color gradients (Rocheet al. 2010), and slightly larger sizes Kauffmann,G.,etal.2003, MNRAS,341,33 (Bernardi2009).Whereasthefirsttwoareinagreementwith KobayashiC.,2004,MNRAS,347,740 ourmajormergerpicture,thelastsuggestsmoresizegrowth Kormendy, J., Fisher, D. B., Cornell, M. E. & Bender, R. 2008, thanisusuallyassociatedwithmajormergers.Hence,itmay ApJS,182,216 bebetter tothink of BCG formation as a two step process. LapiA.&CavaliereA.,2009, ApJ,692,174 In the first, the major mergers which result in the object Matkovi´c,A.&Guzm´an, 2005,MNRAS,362,289 becoming a BCG erase its color gradient (and decrease b/a Pipino A., D’Ercole A., Chiappini C., Matteucci F., 2010, MN- –Bernardietal.2008);thereafter,minormergerspuffedup RAS,inpress(arXiv:1005.2154) Ragone-Figueroa,C.&Plionis,M.2007,MNRAS,377,1785 its size. Tidal stripping during the minor merger may also Ragone-Figueroa, C., Plionis, M.,Merch´an, M.,Gottl¨ober, S. & have contributed to the formation of intracluster light in Yepes,G.2010,MNRAS,407,581 its host halo (Bernardi 2009; Bernardi et al. 2010b). This Roche,N.,Bernardi,M.&Hyde,J.B.2010,MNRAS,407,1231 two step picture is in striking agreement with a detailed Shankar F., Lapi A., Salucci P., De Zotti G., Danese L. 2006, analysis of the age, metallicity and abundance gradients of ApJ,643,14 BCG NGC 4889 (Coccato et al. 2010). Skelton,R.E.,Bell,E.F.&Somerville,R.S.2009,ApJL,699,9 Tiret, O., Salucci, P., Bernardi, M., Maraston, C. & Pforr, J. 2010,MNRAS,inpress(arXiv:1009.5185) VanderWelA.,RixH.-W.,Holden,B.P.,BellE.F.&Robaina, ACKNOWLEDGMENTS A.R.2009,ApJL,706,120 We are grateful to Simona Mei for help, encouragement, andforurgingustoreorganizehowwepresentourfindings.