A Search for Moderate-Redshift Survivors from the Population of Luminous Compact Passive Galaxies at High Redshift
aa r X i v : . [ a s t r o - ph . C O ] D ec Draft version August 17, 2018
Preprint typeset using L A TEX style emulateapj v. 5/2/11
A SEARCH FOR MODERATE-REDSHIFT SURVIVORS FROM THE POPULATION OF LUMINOUSCOMPACT PASSIVE GALAXIES AT HIGH REDSHIFT Alan Stockton, Hsin-Yi Shih, Kirsten Larson, and Andrew W. Mann Institute for Astronomy, University of Hawaii, Honolulu, HI 96822; [email protected], [email protected],[email protected], [email protected]
Draft version August 17, 2018
ABSTRACTFrom a search of a ∼ region covered by both the Sloan Digital Sky Survey and UKIRTInfrared Deep Sky Survey databases, we have attempted to identify galaxies at z ∼ . z ∼ .
5. After isolating good candidates via deeper imaging, we further refine the samplewith Keck moderate-resolution spectroscopy and laser guide star adaptive-optics imaging. For fourof the five galaxies that so far remain after passing through this sieve, we analyze plausible star-formation histories based on our spectra in order to identify galaxies that may have survived with littlemodification from the population formed at high redshift. We find two galaxies that are consistent withhaving formed &
95 % of their mass at z >
5. We attempt to estimate masses both from our stellarpopulation determinations and from velocity dispersions. Given the high frequency of small axialratios, both in our small sample and among samples found at high redshifts, we tentatively suggestthat some of the more extreme examples of passive compact galaxies may have prolate morphologies.
Subject headings: galaxies: formation,—galaxies: kinematics and dynamics,—galaxies: stellarcontent,—galaxies: structure INTRODUCTION
Massive galaxies found at z ∼ . ∼ & M ⊙ at redshifts of 5–10 or more. Vir-tually all of these, as well as a significant fractionof the population of passive galaxies at z ∼ . ∼
45% of the massive K -band-selected galaxies at z ∼ . ∼
15% of the stellar mass in red-sequence galaxies with masses above 10 M ⊙ today wasalready in place in similar-mass galaxies on the red se-quence at z ∼ .
3. In any case, a significant frac-tion of the very oldest stars incorporated into the mostmassive present-day galaxies may have had their ori- Some of the data presented herein were obtained at the W.M.Keck Observatory, which is operated as a scientific partnershipamong the California Institute of Technology, the University ofCalifornia and the National Aeronautics and Space Administra-tion. The Observatory was made possible by the generous finan-cial support of the W.M. Keck Foundation. Now at Department of Astronomy, The University of Texasat Austin, Austin, TX 78712
Figure 1.
Image of the z = 2 .
472 passive compact galaxyTXS 2332+154 ER1, obtained with the laser guide star adaptive-optics (LGSAO) system and the NIRC2 camera on the Keck IItelescope. The upper-left inset is the
Galfit (Peng et al. 2002,2010a) model, the lower-left inset shows the residual after subtrac-tion of the model from the observed image, and the upper-rightinset shows the model without convolution with the PSF. gin in these passive compact galaxies at high redshifts.van Dokkum et al. (2008), using high resolution HubbleSpace Telescope (
HST ) images of the Kriek et al. sam-ple, estimated that ∼ < z ∼ . ∼ . × M ⊙ , a stellar population ageof 1.9 Gyr for an exponentially declining star-formationmodel with an e -folding time of 0.3 Gyr, a S´ersic index n = 3 .
2, and a circularized effective radius of 360 pc.These massive compact galaxies with old stellar pop-ulations at high redshifts have posed some real enigmas:(1) several of these galaxies for which good photometryexists appear to have reasonably well-determined stellarpopulation ages. These indicate that, for at least somemassive galaxies, the bulk of the stellar population wasin place by z ∼
10. Such an early epoch of star for-mation would imply an average star formation rate of ∼ M ⊙ per year sustained for a few × yr, oreven higher rates for shorter periods. (2) The very highstellar density implied by the mass and compactness ofthese galaxies requires an extreme degree of dissipation,probably coupled with extreme starburst and/or quasarsuperwinds. The only detailed mechanisms proposed sofar (merging extremely gas-rich galaxies; Wuyts et al.2010; Sommer-Larsen & Toft 2010) predict that the com-pact galaxies should be surrounded by a faint extendeddistribution of old stars and have high S´ersic indices,neither of which is consistently observed. Extremelydeep HST
Wide-Field Camera 3 (WFC3) images (e.g.,Cassata et al. 2010) show no evidence for the predictedextended envelopes, and some of the best studied high-redshift compact galaxies show low-to-moderate S´ersicindices (Stockton et al. 2004; Stockton & McGrath 2007;van Dokkum et al. 2008). (3) What is the final fateof these galaxies, and why are they so extremely rareat the present epoch? One possibility that has beenput forward is the so-called “inside-out” buildup ofpresent-day massive galaxies (e.g., Bezanson et al. 2009;Hopkins et al. 2009; van de Sande et al. 2013). In thisscenario, envelopes are acquired over time by the mas-sive compact galaxies, which have become the cores ofthe most massive local galaxies. The extreme dearth ofunmodified survivors in the present-day universe indi-cates that such an addition of material cannot be an in-frequent, highly stochastic process, such as major merg-ers (Taylor et al. 2009). Instead, there would have tobe a more-or-less steady accretion of material, possiblyvia many minor mergers. However, there is disagree-ment whether the amount of material that can reason-ably be added is sufficient to explain the necessary sizegrowth (see, e.g., Peng et al. 2010b; Oogi & Habe 2013;van de Sande et al. 2013). Whether this lack of massivecompact galaxies at the present epoch is a serious matteror not must await a more careful comparison of (comov-ing) volume densities between z ∼ . z ∼ with special care to match morphologicalcriteria and mass ranges . See Quilis & Trujillo (2013)for a discussion of this still unresolved issue.At high redshifts, we can determine general spectral- There are of course all sorts of caveats regarding ages of inte-grated stellar populations. Here we refer to ages determined fromthe best-fitting Bruzual & Charlot (2003) models, which even nowappear to be among the more realistic spectral-synthesis modelscurrently available (e.g., Zibetti et al. 2012). We have exploredinstantaneous burst and exponentially decreasing star-formationmodels with 0.4, 1.0, and 2.5 solar metallicities. For galaxies at z ∼ .
5, the universe is only ∼ . relative ages should be roughly reliable. energy distributions (SEDs) from photometry or very lowresolution spectroscopy, and we can estimate global mor-phologies of these galaxies from high-resolution imaging,with either adaptive optics with large ground-based tele-scopes or the HST
WFC3. But it is extremely difficultto do much more than this with current facilities. In aheroic effort, van Dokkum et al. (2009) were able to mea-sure an absorption feature in a galaxy at z = 2 .
19, butthis took 29 hr of observing time on an 8 m telescope.Very few additional galaxies at z ∼ z & z ∼ .
5, galaxies that are essentially unmodified sur-vivors of typical examples found at high redshifts, or evenclose analogs to these, are extremely rare. We estimatethat, for those with masses > M ⊙ , effective radius R e . z < .
6, the surface density on the sky ofsuch galaxies is probably < .
02 deg − and quite possi-bly much less than this. Second, we have to deal with thequestion of how we can know that examples that we domanage to find actually are examples of, or at least rep-resent true analogs to, those at high redshift. We discussthese issues below in some detail, but the bottom line isthat we believe that it is possible to identify a very smallsample of galaxies at z ∼ . z >
5; others among our can-didates may have formed more recently, but have similarhistories, including, critically, extreme dissipation. At z ∼ .
5, we are not only able to explore the morphologiesand color gradients of the galaxies to much lower surfacebrightnesses than we can at high redshift, but we canalso potentially obtain spectroscopy of sufficient S/N todetermine detailed properties of the stellar populations,such as kinematics and metallicities. FINDING NEEDLES IN A VERY LARGE HAYSTACK
There have been a number of recent efforts to iden-tify compact passive galaxies at low redshifts. Taylor etal. (2009) searched the Sloan Digital Sky Survey (SDSS)database for early-type compact galaxies in the redshiftrange 0 . < z < .
12, using both SDSS spectra andphotometric redshifts. They found a number of galaxieswith 1 . < R e < .
5, but these all had indicated masses ∼ × M ⊙ , so they are not really comparable tothe more massive compact galaxies found at high red-shifts. Taylor et al. (2009) conclude that such galaxiesmust be extremely rare at the present epoch and thattheir size evolution cannot be a result of a stochasticmechanism such as major merging. On the other hand,Valentinuzzi et al. (2010) have claimed to have foundsubstantial numbers of massive, old, compact galaxies innearby X-ray-selected clusters, including a small num-ber with M > M ⊙ and R e < . z < . R e ∼ . M ∼ . × M ⊙ ), but thethird had an R e nearly three times larger than the SDSSestimate. Furthermore, all of these galaxies turned out tohave quite complex structures for which single S´ersic fitsleft quite large systematic residuals. Ferr´e-Mateu et al.(2012) have made detailed observations, including spec-troscopy, of another seven of the galaxies identified byTrujillo et al. (2009), finding that for essentially all ofthese galaxies, the light (and sometimes the mass) isdominated by stellar populations younger than 2 Gyr.Poggianti et al. (2013) have searched the Padova Millen-nium Galaxy and Group Catalogue (Calvi et al. 2011) forsuperdense galaxies in the redshift range 0 . < z < . ∼
38 deg of sky and finding some 32 galax-ies that meet their criteria. All but three of these,however, have indicated masses < M ⊙ . Most re-cently, Damjanov et al. (2013) have identified nine com-pact galaxies at intermediate redshift, including two withessentially pure old stellar populations. However, as faras we can determine, none of these searches seem to haveturned up galaxies like the more compact massive onestypically found at high redshift, with R e . M > M ⊙ .The principal goals of our program have been to carryout a systematic search for minimally altered moderate-redshift examples of the luminous, passive, compactgalaxies found at high redshifts, as well as to identifyclosely related objects that might shed more light on thesubsequent evolution of these galaxies. Our primary re-sources for this search have been the SDSS and UKIRTInfrared Deep Sky Survey (UKIDSS; Lawrence et al.2007) databases. Our search has focused on higher red-shifts ( z ∼ .
5) than the searches mentioned above, withthe hope that at this redshift we might have a betterchance of isolating at least a few massive compact galax-ies that had survived from the high-redshift populationrelatively unscathed. On the other hand, this redshift isstill sufficiently low that we can obtain high S/N spec-tra and LGSAO imaging of sufficient quality to allowus to explore in considerable detail morphologies, stel-lar populations, velocity dispersions, and other physicalproperties. Our search is by no means complete, in thatwe have not been able to explore all of the relevant pa-rameter space, nor have we yet been able to follow upevery potential candidate. In addition, a single errantSDSS or UKIDSS magnitude outside the typical errorestimates (which we find to be not terribly uncommon)will throw an object out of our sample. Nevertheless, wefeel that we have done sufficient exploration to confirmthat galaxies with old stellar populations that meet ourcriteria ( R e . M > M ⊙ ) exist at z ∼ . z ∼ .
5, the SDSS almost never has spectra, althoughoccasionally one will have been flagged as a high-redshiftQSO candidate and have a spectrum for that reason. Weindeed find occasional contamination from certain typesof carbon stars and from red QSOs, but over 90% ofthe objects found by our current selection procedures aregalaxies at close to the expected redshift, so we no longerhave found it worthwhile to obtain low S/N spectra tofurther refine the sample as a first step. However, we dofirst have to obtain deeper imaging observations at finerpixel scales, since both the SDSS and UKIDSS have 0 . ′′ R e much greater than 1 kpc (for best-fit single S´ersic profiles). FIVE MASSIVE COMPACT GALAXIES AT Z ∼ . We have carried out our search over ∼ of sky.We have so far identified just five galaxies that meet ourselection criteria. Two of these have been published be-fore (Stockton et al. 2010, Article 1), but we reanalyzethese as well as the new ones here. (Note that Article 1quotes major axis values for R e , whereas in this article weuse circularized values, unless clearly specified otherwise,as in the following paragraph). All of the spectroscopy weshow has been obtained with LRIS (Oke et al. 1995) onthe Keck I telescope, and all of the high-resolution imag-ing has come from the LGSAO system (Wizinowich et al.2006) and the NIRC2 camera on the Keck II telescope.We first show in Figure 2 the Keck II LGSAO im-ages of the galaxies, along with Galfit (Peng et al.2002, 2010a) model fits to these images. Note that weinitially select these by the major-axis effective radius( A e ) rather than by the usual circularized effective ra-dius ( R e = A e p b/a ), which we otherwise quote, in or-der to minimize selection bias against face-on flattenedsystems. We select all candidates with A e ≤ R e . II ] λ Figure 2.
Keck LGSAO images and model fits for the 5 galaxies in our sample with single-S´ersic major-axis effective radii A e ≤ ≥ . M ⊙ , and initial estimated stellar-population ages > Galfit model (Peng et al. 2002, 2010a), the residual from subtracting the model from the data, the
Galfit model without PSF convolution (which should give the best global impression of the galaxy morphology), and, finally, the spectroscopicredshift and fit parameters for each of the components. Insets for SDSS J101305+071636 give lower-contrast versions to show the threediscrete components of this apparently merging system. Each panel is 3 . ′′ !" !" !" !" %"!&"!’"!$"!!"! !" !" !" !" F λ ( − w a tt m − μ m − ) λ ( μ m) Figure 3. pPXF model spectra (red) superposed on observed spectra (black). polynomial fit to the continuum; thus the fit is dependenton spectral features alone. Population fitting with pPXFinvolves choosing a regularization parameter that affectsthe smoothness of the solutions, and it needs to be em-phasized that acceptable fits to the observed spectra canbe obtained with a considerable range of star-formationhistories. Mostly these involve a simple (but asymmet-ric) narrowing or widening of the duration of star-formingevents, but there can also be modest variations of massfractions between different events. Given the strong dis-sipation that must have been involved in the formation ofthese galaxies, we assume that the initial starburst waslikely intense and rather brief. Accordingly, we have cho-sen regularization parameters that tend to give fairly nar-row distributions for the major star-formation episodes,rather than the smoothest possible distribution consis-tent with the spectra. But it should clearly be under-stood that this choice is an assumption rather than aresult.The star-formation histories corresponding to thesemodels are shown in Figure 4, where we have collapsedthe three metallicities we have used (0.4, 1.0, and 2.5solar) to give a clearer indication of the total mass frac-tion as a function of epoch. The average mass-weightedmetallicities of the pPXF-derived populations are 1.4,2.3, 1.8, and 1.0 solar for SDSSJ011436, SDSSJ101305,SDSSJ160720, and SDSSJ232949, respectively. Eventhough they are constrained by the reasonably high S/Nof our spectra, the fitting uncertainties in these metallic-ity estimates are typically at least ∼ µ m, as shown in Figure 3), byplotting the resulting SEDs over the much wider spectralregion covered by the SDSS/UKIDSS photometry. Theseare shown in Figure 5. Because the SDSS photometry isbased on a model fit to the profile and for the UKIDSSwe have used a 2 ′′ -diameter aperture magnitude, we al-low for a small offset between them, which will likelydepend on the actual source light distribution, amongother factors. In order to estimate these offsets, we useHyper- z (Bolzonella et al. 2000) with the BC03 modelsand Calzetti et al. (2000) reddening to fit to the SDSSphotometry plus the UKIDSS Y and J points (only),where the latter magnitudes are given stepped offsetsuntil a minimum χ is found for the best model fit. Itshould be stressed that because of the small wavelengthoffsets between the z , Y , and J points, this determinationis not sensitive to the exact SED used; any other consis-tent fit (e.g., by ignoring reddening) gives essentially thesame offset. The offsets were always in the sense that theUKIDSS flux densities had to be increased. The magni-tude offsets for all galaxies except SDSSJ101305 were 0.1mag or less; for SDSSJ101305 it was 0.35 mag. The off-sets were then applied to all of the UKIDSS magnitudesto give the SEDs shown in Fig. 5. These show that thepPXF model fits over their limited spectral region fit thephotometry over a much wider region remarkably well,allowing for the occasional errant data point (such as theSDSS z -band point for SDSSJ160720). J101305.63+071636.7
J232949.60+151106.3
J011436.33–011438.1
J232949.60+151106.3
J160720.64+235223.8
Stellar Age (Gyr) M a ss F r a c t i on Figure 4.
Star-formation histories, based on pPXF (Cappellari & Emsellem 2004) population fitting for four compact passive galaxies.As described in more detail in the text, these should be taken as indicative only, because of degeneracies, particularly in the durations ofthe star-formation episodes. !" !" -./012345678-59: !" ;;. !"!!" J160720.64+235223.8 F λ ( − w a tt m − μ m − ) J101305.63+071636.7
J232949.60+151106.3
J011436.33–011438.1 pPXF Model λ ( μ m) Figure 5.
SEDs of four compact passive galaxies. The UKIDSS photometry has been slightly scaled as described in the text. The redtrace is the pPXF model derived from the spectra. SURVIVORS FROM THE HIGH-REDSHIFT COMPACTPASSIVE GALAXY POPULATION?
Stellar Populations
For the two galaxies discussed in Article 1(SDSSJ011436 and SDSSJ232949), we concludedfrom single population fits that the stellar populations ofboth galaxies had ages of about 5 Gyr, which would haveplaced their formation epoch at z ∼ .
8. With the wideroptions given by the pPXF fitting procedure, we see thepossibility that at least some of these galaxies formedmost of their mass at high redshifts and therefore maybe only slightly modified survivors of the high-redshiftpopulation. In particular, SDSSJ011436 is an excellentcandidate for such an object, where the pPXF modelhas ∼
1% (by mass) of ∼ z >
6. Quite aside from the spectral fits,the photometric SED shown in Figure 5 reinforces thispicture, where the SDSS u -band point, in particular,indicates the effect of the small admixture of a youngpopulation. A similar scenario applies to the model forSDSSJ160720: the overwhelming bulk of the mass isformed at high redshift, with a couple of more recentepisodes of star formation (perhaps due to wet minormergers) accounting for < ∼
1, 3, and 7.5Gyr) with the three merging components. Whether thisis the case or not, there is evidence that a significantportion of the total mass formed at z ∼
5. Finally, forSDSSJ232949, the pPXF spectral fit (which, as discussedearlier, is dependent only on spectral features, not thecontinuum shape) indicates major star-formation eventsat both z & z ∼ .
2, with some additional starformation at z ∼ .
65, or ∼ Masses
The actual masses of passive compact galaxies are acomplex and somewhat contentious subject (see, e.g.,Ferr´e-Mateu et al. 2012). We can attempt to estimatethe total stellar masses for these galaxies from theadopted stellar population fits, and we can also try toestimate the dynamical masses from the measured veloc-ity dispersions. For the pPXF stellar population fits, wehave used BC03 models with Chabrier (2003) IMFs. Byscaling the mass-fraction-weighted sum of the SSPs con-tributing to the total spectrum, corrected for reddening,to the photometric SED, we can obtain an estimate ofthe stellar mass.We have also used pPXF to fit velocity dispersions tothe spectra, following the prescriptions for setting thebias parameter. We can then obtain an estimate of themass using the virial mass estimator m V = βR e σ /G ,where β is a parameter that takes into account the rela- tion between h r i and R e , as well as that between h v i and σ , where the former in each case are the virialquantities for the stellar distribution and the latter areobservable quantities. For a sample of local ellipticals,Cappellari et al. (2006) found β = 5 . ± .
1, and thisvalue has generally been used, with a slight dependencebeing noted on the S´ersic parameter n (Bertin et al.2002). In Table 1, Columns 4 and 5, we compare es-timates of the stellar mass derived from our stellar pop-ulation fits with the dynamical masses, assuming β = 5.For the dynamical masses, we estimate R e from the el-lipse enclosing half the total light measured on our Gal-fit model (without PSF convolution). We correct ourmeasured σ to the σ that would be measured within R e with the relation given by Equation 1 of Cappellari et al.(2006).There are clear discrepancies between these mass es-timates, always in the sense that the dynamical mass isless than the mass inferred from the stellar populationhistories. There are two main assumptions that couldexplain the difference:1. The mass we have derived from the stellar popu-lations assumes the Chabrier (2003) initial massfunction (IMF). If the actual IMF is more topheavy than we have assumed, these masses couldbe smaller. However, there is recent evidence thatlocal massive elliptical galaxies may instead havea bottom-heavy IMF (e.g., Conroy & van Dokkum2012). Nevertheless, since these compact galaxieswill have formed under rather specialized condi-tions that we do not yet understand very well, wecannot know what sort of IMF would have pre-vailed, and IMF variation may account for somepart of the discrepancy.2. In estimating the mass from the velocity dispersionin Column 4, we have assumed that the parameter β is the same as that for the local ellipticals stud-ied by Cappellari et al. (2006). This means thatwe are assuming homology between our compactgalaxies and these ellipticals, which is unlikely tobe strictly true and may possibly be very far offthe mark. The possibility of a substantial differ-ence in structure is reinforced by the recent workof Peralta de Arriba et al. (2013), who find an ap-proximate empirical relation between β ( K in theirpaper) and the compactness of a stellar system ata given mass. When we apply their correction,we get the dynamical masses given in the last col-umn of Table 1. These masses are all somewhatabove our estimate of the stellar mass, possiblybecause of contributions from dark matter. How-ever, given the simplifications and uncertainties in-volved, these values probably are not significantlydiscrepant. Prolate Morphologies?
The models shown in the last column of Figure 2 have b/a ratios for single S´ersic fits ranging from 0.26 to 0.47(for SDSSJ101305, we consider only the central com-ponent; if we were to include all components, the ra-tio would be even less). These low ratios are similar
Table 1
Galaxy Mass EstimatesSDSS σ R e Dynam. Mass 1 a pPXF Mass Dynam. Mass 2 b (km s − ) (kpc) (10 M ⊙ ) (10 M ⊙ ) (10 M ⊙ )J011436.33 − . ± . .
86 2 .
74 3.50J101305.63+071636.7 290 . ± . .
57 2 .
23 3.95J160720.64+235223.8 268 . ± . .
13 2 .
14 3.26J232949.60+151106.3 258 . ± . .
46 2 .
00 2.80 a Calculated from the virial relation M dyn = βσ R e /G , with β = 5. The measured velocity dispersions given in column 2 have beencorrected to those expected within R e as described by Cappellari et al. (2006). b Calculated from the virial relation, but with β = 6( R e / . − . ( M ∗ / ) . , following Peralta de Arriba et al. (2013), where R e isin kpc and M ∗ is the stellar mass from Column 5, corrected for the difference between our assumed Chabrier (2003) IMF and the SalpeterIMF assumed by Peralta de Arriba et al. (2013), following the prescription of Longhetti & Saracco (2009). to those often found for passive galaxies at high red-shift (Stockton et al. 2004; Stockton & McGrath 2007;van der Wel et al. 2011). While our sample is too smallto be more than indicative of a possibility, it does serveas a basis for speculating that the intrinsic morpholo-gies of many of these galaxies, at both high and lowredshifts, may in fact be prolate. One can easily showthat the distribution of projected axial ratios of an en-semble of disk galaxies with a given intrinsic b/a willbe approximately uniform over the range from the in-trinsic value to 1, whereas the distribution for prolateobjects will be strongly peaked near the intrinsic ratio.For example, formally, with the probability for a ran-domly oriented thin disk to have a projected b/a ≤ . b/a ≤ . b/a = 0 . b/a ≤ . A e , there should be little or no selection bias favoringsmall axial ratios. The possibility that passive compactgalaxies at high redshift might be prolate was briefly con-sidered and rejected by van der Wel et al. (2011), on thegrounds that such objects are not seen locally and thatgalaxy formation mechanisms should “be independent ofcosmic time.” We believe, on the contrary, that while wehave no compelling model for how these massive com-pact galaxies form, the very fact that passive galaxiesin the early universe typically have morphologies quiteunlike those seen locally is an indication that galaxy for-mation mechanisms have evolved over cosmic time. Ifthese galaxies were actually to be found to be prolate,such a morphology would imply an anisotropic stellarvelocity field and structural differences that could verywell account for the discrepancy between the calculatedstellar mass and the dynamical mass estimated from therelation that works well for the population of local ellip-ticals. OVERVIEW
Our search for examples of galaxies with M ∗ > M ⊙ , R e . z ∼ .
5, with the overwhelm-ing bulk of the stellar mass having been formed at z > × M ⊙ . Although our cur-rent sample is small, the fact that all five of our galaxieshave small b/a ratios has led us to tentatively considerthe possibility that these galaxies may have prolate mor-phologies. It will be necessary to identify a larger sampleto test this suggestion.As indicated in Section 1, we still do not seem to havea wholly adequate understanding of the nature of the ap-parently highly dissipative process by which these mas-sive compact galaxies form. If prolate morphologies wereto be confirmed for a significant subset of these galaxies,this finding would almost certainly help point the way toa better understanding of their formation mechanism.After this article was submitted, Trujillo et al. (2013)announced the remarkable discovery that NGC1277, ata distance of only ∼ Facilities: