Tidal Disruption Event Host Galaxies in the Context of the Local Galaxy Population
Jamie Law-Smith, Enrico Ramirez-Ruiz, Sara L. Ellison, Ryan J. Foley
DDraft version November 16, 2017
Typeset using L A TEX twocolumn style in AASTeX61
TIDAL DISRUPTION EVENT HOST GALAXIESIN THE CONTEXT OF THE LOCAL GALAXY POPULATION
Jamie Law-Smith,
1, 2
Enrico Ramirez-Ruiz,
1, 2
Sara L. Ellison, and Ryan J. Foley Department of Astronomy and Astrophysics, University of California, Santa Cruz, CA 95064, USA Niels Bohr Institute, University of Copenhagen, Blegdamsvej 17, DK-2100 Copenhagen, Denmark Department of Physics and Astronomy, University of Victoria, Victoria, BC, Canada
ABSTRACTWe study the properties of tidal disruption event (TDE) host galaxies in the context of a catalog of ∼ ∼ ∼× ∼× g − r ) ≈ . ∼ ), higher S´ersic indices (∆ n g ≈ B/T ≈ .
5) than galaxieswith matched BH masses. We find that TDE host galaxies appear more centrally concentrated and that all have highgalaxy S´ersic indices and
B/T fractions—on average in the top 10% of galaxies of the same BH mass—suggesting ahigher nuclear stellar density. We identify a region in S´ersic index and BH mass parameter space that contains ∼ ≥
60% of TDE host galaxies. The unique photometric properties of TDE hostgalaxies may be useful for selecting candidate TDEs for spectroscopic follow-up observations in large transient surveys.
Keywords: black hole physics—galaxies: active—galaxies: evolution—galaxies: nuclei
Corresponding author: Jamie [email protected] a r X i v : . [ a s t r o - ph . H E ] N ov Law-Smith et al. INTRODUCTIONThe cores of many galaxies undergo intense nuclearactivity during their lifetimes. This activity inevitablyleads to the growth of the central supermassive blackhole (SMBH) but is short lived compared to galacticages and was more prevalent when the universe was only ∼
20% of its current age (Soltan 1982; Ho 2009). Qui-escent SMBHs starved of fuel are common in the localuniverse (Greene & Ho 2007) and are being discoveredin nearby galaxies.The presence of quiescent SMBHs in the nuclei ofgalaxies has been directly inferred from the dynamics ofthe stars and/or gas near their centers (e.g., Kormendy& Ho 2013; McConnell & Ma 2013). For galaxies toodistant to accurately resolve the nuclear stellar or gaskinematics, it is possible to probe the presence of anSMBH with the fate of the central, closely-packed stars.Each star within the nuclear star cluster traces out an in-tricate orbit under the combined influence of the SMBHand other stars. The orbits are gradually altered ow-ing to the cumulative effect of encounters. As a result,stars that are scattered into orbits that pass too closeto the central SMBH can be ripped apart by the blackhole’s tidal field in what is known as a tidal disruptionevent (TDE; Hills 1975; Frank & Rees 1976; Rees 1988).After the star is disrupted, up to half of the stellar de-bris falls back and accretes onto the SMBH (Carter &Luminet 1982; Evans & Kochanek 1989; Lodato et al.2009; Guillochon & Ramirez-Ruiz 2013). The accretionpowers a flare that is a definitive sign of the presence ofan otherwise quiescent SMBH.TDEs are identified by a combination of a rapid in-crease in flux, proximity to a host galaxy’s nucleus, anda decay in luminosity that loosely follows the canoni-cal t − / law, though the most compelling events arethose in which the rise, peak, and decay of the tran-sient are observed with a frequent cadence (e.g., Ko-mossa et al. 2004; Gezari et al. 2009, 2012; Chornocket al. 2014; Arcavi et al. 2014; Holoien et al. 2014; Milleret al. 2015). The (well-sampled) light curves of TDEscontain vital information about the disruption and canbe used to constrain the properties of the SMBH andthe stellar object that was disrupted (e.g., Guillochonet al. 2014; Law-Smith et al. 2017). A few dozen can-didate TDEs have been observed in the optical, UV,and X-ray (for a summary, see Komossa 2015; Auchettlet al. 2017a). Future surveys such as the Large Synop-tic Survey Telescope (LSST) will likely find hundreds tothousands more events (van Velzen et al. 2011).The observed rates of TDEs and, in particular, therelative rates of flares in different galaxy hosts, hold im-portant discriminatory power over both the dynamical mechanisms operating in galactic nuclei and the natureof their underlying stellar populations. However, thedynamical mechanisms that feed stars into disruptiveorbits within nuclear star clusters remain highly uncer-tain. Stellar tidal disruption rates have typically beenstudied under the assumption of a spherical nuclear starcluster that feeds stars to the black hole (BH) through atwo-body relaxation-driven random walk in angular mo-mentum space (Magorrian & Tremaine 1999; Wang &Merritt 2004; Stone & Metzger 2016). However, disks ofstars and gas, if present, could feed stars to the BH at anenhanced rate through collisionless processes or secularinstabilities (Lightman & Shapiro 1977; Rauch & Ingalls1998; Magorrian & Tremaine 1999; Madigan et al. 2009,2011; Merritt & Vasiliev 2011; Vasiliev & Merritt 2013;Antonini & Merritt 2013). A second massive body, suchas an inspiraling moderately massive BH, could also in-duce large-angle scatterings of stars (Ivanov et al. 2005;Chen et al. 2009). These processes and others couldresult in favorable conditions for TDEs and might man-ifest as enhanced rates within particular galaxy hosts(for a review, see Alexander 2017).Understanding the host galaxies of TDEs is thus im-portant; however, this understanding is in its infancy.Many uncertainties in the conditions necessary for tidaldisruption will only be resolved through an understand-ing of the connection between TDEs and their hostgalaxies. This connection will hopefully become clearerwith a larger sample of TDE host galaxies, but the cur-rent sample already shows hints of being highly unusual.TDEs appear to be observed preferentially in rarequiescent Balmer-strong galaxies (also known as post-starburst or K+A galaxies, or more restrictively as E+Agalaxies; Arcavi et al. 2014; French et al. 2016). Ina sample selected from the Sloan Digital Sky Survey(SDSS), French et al. (2016) found that a particularselection of E+A galaxies contained only 0.2% of thesample but more than one-third of observed TDE hostgalaxies, implying a drastic rate enhancement. How-ever, it is important to make a clear distinction betweengalaxies in which TDEs can occur and galaxies in whichTDEs might be observable. We need to discern among(1) the intrinsic TDE rate based on stellar dynamics, (2)the rate of TDEs that produce luminous flares, and (3)the potential selection effects against detecting a TDE.In this paper, we seek to disentangle some of these is-sues.We study the properties of TDE host galaxies in thecontext of a catalog of ∼ idal Disruption Event Host Galaxies B/T ).We take our sample of TDEs from the catalog pre-sented in Auchettl et al. (2017a) and compile galaxyproperties from the SDSS galaxy catalogs of Brinch-mann et al. (2004), Simard et al. (2011), and Mendelet al. (2014). We explore several key properties of TDEhost galaxies, including stellar mass, BH mass, redshift,star formation rate (SFR), bulge colors, surface bright-ness, S´ersic index, bulge-to-total-light ratio, and galaxyasymmetry. We also compare TDE host galaxies to ac-tive galactic nuclei (AGNs) and star-forming (SF) galax-ies across these observables.This paper is organized as follows. We describe ourdata in Section 2. We present the uniqueness of TDEhost galaxies in Section 3. We explore selection effectsin Section 4. We present a possible physical explanationfor the overabundance of TDEs in E+A/post-starburstgalaxies, as well as a new unique feature of all TDEhost galaxies, in Section 5. We discuss and interpret ourfindings in Section 6. We study a few other properties ofTDE host galaxies in Appendix A and show correlationsbetween properties in Appendix B. DATAIn this section, we describe our data sources as wellas some conventions and definitions we use throughoutthe paper. 2.1.
Reference Catalog
Our reference catalog is contained in the SDSS (Yorket al. 2000; Gunn et al. 1998, 2006) DR7 (Abazajianet al. 2009) and is based on the main galaxy sample(Strauss et al. 2002). We make use of the MPA-JHUcatalogs (Brinchmann et al. 2004) of ∼ ∼ g − r , bulge and galaxymagnitude, galaxy half-light radius, galaxy S´ersic index(see Equation 4), bulge fraction ( B/T ), galaxy asym-metry indicator, and inclination measurements fromthe Simard et al. (2011) catalog. We obtain veloc-ity dispersion, H α equivalent width (EW), Lick H δ A , D n (4000), and star formation rate (SFR) measurementsfrom the MPA-JHU catalog. Here, we define H α EW as
H ALPHA FLUX / H ALPHA CONT . We obtain total and bulgestellar masses from the Mendel et al. (2014) catalog. Wecollate these measurements into a catalog of ∼ z > .
01 (to prevent severe aperture bias), reliableH α EWs (
H ALPHA EQW ERR > -1 in the MPA-JHU cat-alog), and median signal-to-noise ratio (S/N) per pixelof the integrated spectrum of greater than 10. Apply-ing these selection criteria leaves us with a final catalogof ∼ β , NII (6584˚A),and H α —of greater than 3. We classify low-S/N AGNand low-S/N SF galaxies as those with a minimum S/Nof less than 3.We use the M bh - σ e scaling from Kormendy & Ho(2013) to estimate galaxy BH masses: M bh M (cid:12) = (cid:0) . +0 . − . (cid:1) (cid:16) σ e − (cid:17) . ± . . (1)Equation 1 yields M bh values with an intrinsic scatter of0.29 dex. We use velocity dispersion measurements fromthe MPA-JHU catalog and perform an aperture correc-tion to obtain the bulge/spheroidal velocity dispersion σ e , using Equation (3) in Jorgensen et al. (1995):log σ ap σ e = − .
065 log (cid:18) R ap R e (cid:19) − . (cid:20) log (cid:18) R ap R e (cid:19)(cid:21) (2)where R e is the effective radius of the bulge or spheroidfrom the Simard et al. (2011) catalog, R ap is the aper-ture radius (1.5 (cid:48)(cid:48) ), and σ ap is the velocity dispersionmeasured within the aperture. Our errors on BH massinclude the error on velocity dispersion and the intrinsicscatter in the M bh - σ e scaling, and are ∼ Law-Smith et al. uncertainties on BH mass are relatively large, particu-larly for galaxies with low velocity dispersions (a fewof our TDE host galaxies have velocity dispersions nearor slightly below the 70 km s − SDSS instrumental res-olution). However, our analysis in this work is pri-marily concerned with differences between properties ofTDE host galaxies and our reference catalog and so doesnot rely on accurate determinations of BH masses—onlythat they are determined homogeneously in the varioussamples we consider. Indeed, we often control for BHmass. We also performed our analysis using M (cid:63), bulge todetermine BH masses using the scaling from Kormendy& Ho (2013), M bh M (cid:12) = (cid:0) . +0 . − . (cid:1) (cid:18) M bulge M (cid:12) (cid:19) . ± . , (3)and M (cid:63), bulge estimates from the Mendel et al. (2014)catalog, as well as using different scalings for M bh - σ ,and our conclusions are insensitive to these choices. Infact, if we replace BH mass with M (cid:63), total throughout ouranalysis, our conclusions remain the same.When we study bulge quantities, such as the bulgecolor, bulge fraction ( B/T ), and bulge magnitude,obtained from the Simard et al. (2011) catalog ofbulge+disk decompositions, we will show measurementsfrom all galaxies in our reference catalog. Note that thiswill include galaxies where a second component is notstatistically justified in the fit. We can isolate a rela-tively “pure” sample of bulges by including only galaxiesfor which the data support a bulge+disk decompositioncompared to a single S´ersic fit (for example, by requir-ing P pS < .
32; see Simard et al. 2011). The size ofthis “pure” bulge sample depends mostly on the dataquality and so it can be highly incomplete. This sampleincludes roughly one-third of our reference catalog andonly three of our TDE host galaxies (numbers 5, 7, and8 in Table 1). Although the data quality cannot alwaysstatistically justify the bulge+disk decomposition, bulgemeasurements can be applied in a consistent way to ourentire sample, and we find intriguing differences betweenthe TDE host galaxies and our reference catalog (seeSection 3).The SDSS is biased in a few ways, and this leads tosome sample limitations. Most importantly, the SDSS is We note that Wevers et al. (2017) recently published the firsthomogeneously measured BH masses for a complete sample of12 optical-/UV-selected TDE host galaxies. We use the SDSSvelocity dispersions, even though they are less accurate, as ourgoal is to use consistent metrics in comparing between TDE hostsand our reference catalog. That being said, the TDE host galaxyBH masses we match on in this work are broadly consistent withthe range found by Wevers et al. (2017). inherently flux limited. This limits our sample to TDEsthat are fairly low z , but since most observed TDEs arefairly low z , this is not a major problem. Our approachin this paper—of not just using the entire SDSS for ourcomparison, but selecting matched control samples onseveral parameters, forcing the parameter space to bethe same—should mitigate most inherent biases in oursample. 2.2. TDE Host Galaxies
We use the Auchettl et al. (2017a) catalog of 71 candi-date TDEs as a parent sample of TDE host galaxies. Weremove candidates with only one observation and thosein the
Not a TDE and
Unknown categories, leaving uswith 42 candidate TDEs. We use the R.A., decl., and z of these host galaxies to find matches in our referencecatalog described above. Of the 42 candidate TDE hostgalaxies, 10 are in our reference catalog; the relativelylow number of matches is mainly due to the redshift andmagnitude limits of the various catalogs we draw from,as well as the fact that most galaxies require an SDSSspectrum for inclusion in these catalogs. These matchesand their numbering (1-10) used throughout the paperare listed in Table 1, along with the relevant primaryreferences.We will use TDE host galaxies numbered 1-5 in ourmatching analysis of the extent of selection effects on theoverrepresentation of TDEs in quiescent Balmer-stronggalaxies (Section 4). TDE candidates 6-9 were not iden-tified photometrically (i.e., by their light curves) butwere instead proposed as TDE candidates due to theirunique spectra: they are “extreme coronal line emit-ters,” and are difficult to explain as standard AGNs.Numbers 9 and 10 are not in the MPA-JHU catalog,and so do not have velocity dispersion, H α EW, LickH δ A , or SFR measurements from this catalog. Addi-tionally, No. 8 does not have a reliable velocity dis-persion measurement from the MPA-JHU catalog (it isflagged with a negative error). We use the M bh - M (cid:63), bulge relation from Kormendy & Ho (2013), using M (cid:63), bulge measurements from the Mendel et al. (2014) catalog, toestimate BH masses for Nos. 8, 9, and 10—these BHmasses do not enter into the analysis and are only usedto place these TDE host galaxies on our 2D plots ver-sus BH mass. TDE hosts 1-5 are shown with red pointsand histograms throughout the paper, and TDE hosts6-10 are shown with orange points. In our 1D stackeddistributions, we will show the histograms for all TDEcandidate with matches in our reference catalog (1-10)with dotted black lines. Finally, we will show H α EWand Lick H δ A measurements for five TDE host galaxies(labeled a-e, see Table 1) not in our reference catalog in idal Disruption Event Host Galaxies Table 1.
TDE Host Galaxies Used in This Work a Host Name Host R.A. Host Decl. Redshift Reference1 ASASSN-14ae Veiled SDSS J110840.11+340552.2 11:08:40.116 34:05:52.23 0.0436 Holoien et al. (2014)2 ASASSN-14li X-ray SDSS J124815.23+174626.4 12:48:15.230 17:46:26.45 0.0206 Holoien et al. (2016)3 PTF-09ge Veiled SDSS J145703.17+493640.9 14:57:03.18 49:36:40.97 0.064 Arcavi et al. (2014)4 RBS 1032 Possible X-ray SDSS J114726.69+494257.8 11:47:26.80 49:42:59.00 0.026 Maksym et al. (2014)5 SDSS J1323 Likely X-ray SDSS J132341.97+482701.3 13:23:41.973 48:27:01.26 0.08754 Esquej et al. (2007, 2008)6 SDSS J0748 Veiled SDSS J074820.67+471214.3 07:48:20.667 47:12:14.23 0.0615 Wang et al. (2012)7 SDSS J1342 Veiled SDSS J134244.41+053056.1 13:42:44.416 05:30:56.14 0.0366 Wang et al. (2012)8 SDSS J1350 Veiled SDSS J135001.49+291609.7 13:50:01.507 29:16:09.71 0.0777 Wang et al. (2012)9 SDSS J0952 Veiled SDSS J095209.56+214313.3 09:52:09.555 21:43:13.24 0.0789 Komossa et al. (2008)10 SDSS J1201 Likely X-ray SDSS J120136.02+300305.5 12:01:36.028 30:03:05.52 0.146 Saxton et al. (2012)a PTF-09axc Veiled SDSS J145313.07+221432.2 14:53:13.08 22:14:32.27 0.1146 Arcavi et al. (2014)b PTF-09djl Veiled SDSS J163355.97+301416.6 16:33:55.97 30:14:16.65 0.184 Arcavi et al. (2014)c PS1-10jh Veiled SDSS J160928.27+534023.9 16:09:28.28 53:40:23.99 0.1696 Gezari et al. (2012)d Swift J1644 X-ray Swift J164449.3+573451 16:44:49.30 57:34:51.00 0.3543 Bloom et al. (2011)e PTF-15af b NA SDSS J084828.13+220333.4 08:48:28.13 22:03:33.4 0.0790 French et al. (2016) a From Auchettl et al. (2017a). b The discovery article for PTF-15af has not yet been published in the literature, but we include it here as it is included in the French et al.(2016) sample and defines the boundary of their wF16 selection (see Figure 1).
Note —TDE host galaxies 1-10 are in our reference catalog, and we use hosts 1-5 in our matching analysis (see text). Host galaxies a-eare not used in our analysis, but have published H α EW and Lick H δ A measurements and are shown in Figure 1. Figure 1. These last five are not used in our analysis,yet they provide additional evidence of an overrepre-sentation of TDEs in E+A/post-starburst galaxies. Al-though the small number of TDE host galaxies precludesperforming detailed statistics, we are nonetheless ableto draw compelling conclusions about the uniqueness ofthese galaxies.The robustness of our conclusions may suffer fromsmall numbers. Additionally, it is possible that the 10TDE host galaxies (and ultimately the five used in ourmain analysis) are a special subset of TDE host galaxiesand are not representative of the parent sample of 42host galaxies. The 10 TDE host galaxies in our refer-ence catalog are relatively low z and are not particularlyfaint, so that they are included in SDSS, and are there-fore the most well-characterized in terms of their hostproperties. This is a potential source of bias in the TDEhost sample.Auchettl et al. (2017a) divide their candidate eventsinto the categories X-ray TDE , Likely X-ray TDE , Pos-sible X-ray TDE , and
Veiled TDE . We provide this clas-sification in Table 1, but the number of matches in each category is too small to make robust conclusions aboutdifferences in TDE host galaxies between categories.The categorization is explained in detail in Auchettlet al. (2017a) but we summarize it here. Events in the
X-ray TDE category have a well-defined and trustworthyX-ray light curve. Events in the
Likely X-ray TDE cate-gory have very similar properties, yet with more limiteddata coverage. Events in the
Possible X-ray TDE cate-gory have even more limited X-ray observations. Eventsin the
Veiled TDE category have a well-defined opti-cal/UV light curve but no X-ray emission near the peak. UNIQUENESS OF TDE HOSTSFollowing French et al. (2016), we define the follow-ing selections in order to isolate quiescent Balmer-stronggalaxies. We define the strong F16 (sF16) selection asH δ A − σ (H δ A ) > α EW < σ (H δ A )is the error in the Lick H δ A index. H α EW emission isan indicator of current star formation, and so this se-lects for galaxies with little ongoing star formation (i.e.,with specific SF rates well below the main sequence ofstar-forming galaxies). H δ A absorption, from A stars, Law-Smith et al.
Lick H δ A Absorption [ Å ] H α E W E m i ss i o n [ Å ] abc de 1 23 45 67 8strong F16 (sF16)weak F16 (wF16) Figure 1. H α equivalent width emission vs. Lick H δ A absorption, following French et al. (2016), for TDE hostgalaxies (filled circles) and our reference catalog (contours).Galaxies numbered 1-5 are used in our matching analysis (seetext). The solid-line selection (including errors on Lick H δ A ;sF16, see text) contains 0.2% of the galaxies in our referencecatalog and the dotted-line region (containing sF16; wF16)contains 2.3%. Contours are spaced by 0 . σ , with the darkestshading containing 0 . σ and the lightest shading containing2 σ . Median errors in the TDE host galaxy measurementsare shown in the top right. indicates star formation within the past ∼ Gyr. So, sF16galaxies have had a strong starburst in the last ∼ Gyr.We define the weak F16 (wF16) selection as H δ A > . α EW < δ A means thatwF16 galaxies could have several possible star forma-tion histories. Not accounting for selection effects, 0.2%of our reference catalog falls in the sF16 selection and2.3% falls in the wF16 selection. Throughout this paper,we will define galaxies in the sF16 selection as “E+A”galaxies—we note that it is also common to define E+Agalaxies with a stricter cut on H δ A (Goto 2007)—andgalaxies in either the wF16 or sF16 selections more gen-erally as “quiescent Balmer-strong” galaxies.Figure 1 shows H α EW emission versus Lick H δ A ab-sorption, following French et al. (2016), for TDE hostgalaxies and our reference catalog . TDE hosts are num-bered following Table 1. Excluding TDE candidates 6, In this and other 2D plots that follow, we use contours to showour reference catalog galaxies. Note that for a 2D distribution, σ levels are defined differently from those for a 1D distribution. Intwo dimensions, the cumulative density function of a Gaussian is F ( x ) = 1 − e − ( x/σ ) / , meaning that “1 σ ” contains 39.3% of thevolume and “2 σ ” contains 86.5% of the volume. Table 2.
Fraction of Reference Catalog Galaxies in theStrong and Weak F16 Selections According to AGN/SF Clas-sificationCategory Number % in sF16 % in wF16TDE hosts (1-5) 5 20 60TDE hosts (1-5, a-e) 10 30 80Full reference catalog 500,707 0.20 2.29AGN 52,613 0.09 0.60Low-S/N AGN 93,304 0.36 4.02SF 110,133 0.0 0.01Low-S/N SF 42,616 0.01 0.20Unclassified 202,041 0.30 3.61
7, and 8 (not identified photometrically), and includingcandidates a-e, 3 /
10 = 30% of the TDE host galaxies fallin the sF16 selection and 6 /
10 = 60% fall in the wF16selection. TDEs thus remain significantly overrepre-sented in quiescent Balmer-strong galaxies. A straight-forward comparison to all of the galaxies in our refer-ence catalog suggests that TDEs are overrepresented inthe sF16 selection by a factor of ∼
150 (or ∼ ∼
35. Of the five TDE hosts weuse in our matching analysis, 1/5=20% are in the sF16selection and 3/5=60% are in the wF16 selection. Re-stricting ourselves to these five TDE host galaxies doesnot allow us to claim as robust an overrepresentationin sF16 galaxies, but that is not the direct aim of thiswork. Our aim is to compare a wide range TDE ofhost galaxy properties to a larger reference catalog us-ing consistent metrics. Two (of four; two do not havemeasurements available) of the events in the Auchettlet al. (2017a) X-ray TDE category, ASASSN-14li andSwift J1644, appear to be in quiescent Balmer-stronggalaxies, suggesting that X-ray TDEs share the samepreference for these galaxies as do optical/UV TDEs.Next, we show where AGN and SF galaxies fall in thisH α EW and Lick H δ A parameter space. This is impor-tant, as there may be selection effects against detecting Note that the sF16 selection includes the error on Lick H δ A ,which excludes Though the binomial false-positive percentage here for 1/5 ofthe galaxies in a sample being sF16 is still relatively small, ∼ idal Disruption Event Host Galaxies Figure 2. H α EW vs. Lick H δ A as in Figure 1, for our reference catalog, but split according to AGN/SF classification followingKauffmann et al. (2003a). Low-S/N is taken as S/N < . σ , with the darkestshading containing 0 . σ and the lightest shading containing 2 σ . The distributions of each subsample are normalized separately,so the relative number in each of the categories is not represented (see Table 2 for this), only their relative distributions. TDEs in some of these galaxies (in particular in galax-ies hosting a strong AGN and in strongly SF galaxies),as discussed in Sections 4 and 6. Figure 2 shows ourreference catalog split into AGN, low-S/N AGN, SF,low-S/N SF, and unclassified subsamples (see Section 2for definitions). The low-S/N AGN scatter into the F16selections more than the AGN. The unclassified galaxiesalso scatter into the F16 selections.Table 2 shows the fraction of our reference cataloggalaxies in the strong and weak F16 selections accord-ing to AGN/SF classification. Somewhat by construc-tion, almost no SF or low-S/N SF galaxies are in theF16 selections. Importantly, however, if we restrict our-selves to only low-S/N AGN or unclassified galaxies,E+A/post-starburst galaxies are still rare, and TDEhost galaxies remain overrepresented.As a related metric of the uniqueness of TDE hostgalaxies, we show total star formation rate versus to-tal stellar mass for TDE host galaxies and our referencecatalog in the top panel of Figure 3. The solid blue linedescribes the star-forming main sequence (SFMS; Penget al. 2010). We assume a 1 σ scatter of 0.5 dex for theSFMS—this is the median scatter of the SFR measure-ments, shown by the dashed blue lines. We conserva-tively define the “green valley” or “transition region”in this diagram as being 1 σ -3 σ below the SFMS nor-malization (e.g., see Pandya et al. 2016, and referencestherein)—this is between the lower blue dashed line andthe orange dashed line. It is immediately apparent fromFigure 3 that none of our TDE host galaxies lie above theSFMS normalization . Instead, all of our TDE hosts lie Though note that the errors on some of the TDE host galaxiesextend above the SFMS. below the SFMS normalization, with some being withinour assumed 1 σ SFMS scatter and some inhabiting thegreen valley. The location of the TDE host galaxies inthis diagram suggests that they could be making a tran-sition from the SFMS toward quiescence, but additionalconstraints on their stellar populations and star forma-tion histories are needed to test this hypothesis (also seeFrench et al. 2017).The distributions of sF16 and wF16 galaxies, sep-arately normalized to the reference catalog’s distribu-tion, are shown in orange and light blue in the bottompanel of Figure 3. We calculate the fraction of quiescentBalmer-strong galaxies in our three bands of increasingdegrees of quiescence below the SFMS (each spaced by0.5 dex). Recall that the nominal percentage of sF16(wF16) galaxies in our reference catalog is 0.2% (2.3%).Between the solid blue line and the dashed blue line,the percentage of sF16 (wF16) galaxies is 0.5% (0.9%).Between the dashed blue line and the dashed greenline, the percentage of sF16 (wF16) galaxies is 1.1%(7.3%). Between the dashed green line and the dashedorange line, the percentage of sF16 (wF16) galaxies is0.1% (7.8%). Outside of these three bands, the percent-age of sF16 and wF16 galaxies drops well below nomi-nal. If we restrict our reference catalog to galaxies withlog( M (cid:63), tot /M (cid:12) ) < .
5, to match the M (cid:63), tot values ofour TDE hosts, the fractions of quiescent Balmer-stronggalaxies quoted above inside the three bands increaseonly slightly. Thus, quiescent Balmer-strong galaxiesreside preferentially in the green valley, and the relativefraction of E+A/post-starburst galaxies in the band in-habited by TDE hosts 1, 3, and 4 is a factor of fivegreater than nominal. Law-Smith et al.
Figure 3.
Top panel: total star formation rate vs. totalstellar mass for TDE host galaxies (numbered points) andour reference catalog (contours). Galaxies 1-5 are used inour matching analysis. Median errors in the TDE host galaxymeasurements are shown in the top left. The blue solid linedescribes the main sequence of SF galaxies (Peng et al. 2010),with dashed lines spaced by 0.5 dex (the median scatter ofour SFR measurements) above and below. The green andorange dashed lines are also spaced by 0.5 dex, and indicatedegrees of quiescence from the SFMS. Bottom panel: thedistribution for galaxies in the sF16 selection (E+A galaxies)is shown in orange and for galaxies in the wF16 selection inlight blue. sF16 galaxies account for 0.2% of our referencecatalog and wF16 galaxies for 2.3%; their distributions arenormalized separately. Contours are spaced by 0 . σ , withthe darkest shading containing 0 . σ and the lightest shadingcontaining 2 σ .4. TDE SELECTION EFFECTSAs seen above, TDEs appear to show a distinct prefer-ence for quiescent Balmer-strong (and more restrictivelyE+A) galaxies. Much of this preference may in fact be due to physical and observational selection effects. Inthis section, we explore their extent.4.1.
Matching
Our strategy is to create matched comparison samplesdrawn from our SDSS reference catalog that are controlsfor the TDE host galaxies in various observables relatedto possible selection effects, and then to calculate thefraction of quiescent Balmer-strong galaxies in the con-trols. We use TDE host galaxies numbered 1-5 for thisanalysis, as these TDE candidates were photometricallyidentified and their hosts have measurements in our ref-erence catalog of the properties we match. That themeasurements are determined consistently between theTDE host galaxy sample and the reference catalog al-lows us to match and compare properties in an unbiasedway. Our results are similar if we include TDE hostgalaxies 6-8 (which have H α EW and Lick H δ A mea-surements in our reference catalog), and if we include 9and 10 (which do not have H α EW and Lick H δ A mea-surements in our reference catalog).As we match only on five TDE host galaxies, we im-plement our matching as simple tolerances in each pa-rameter. We match on BH mass, redshift, bulge col-ors, and half-light surface brightness (motivated and dis-cussed below). The baseline tolerance used for match-ing is 1% of the “spread” (=97 . th − . th percentile)in each parameter, which corresponds to roughly 0.0018in z , 0.037 dex in BH mass, 0.028 mag in bulge color,and 0.074 mag/arcsec in half-light surface brightness.We limit our control sample to a maximum of 10,000matches per TDE host galaxy; if this is not reached, weincrease the tolerance in intervals of 1%, up to a maxi-mum of 5% of the “spread” in each parameter. We re-quire the same number of matches per TDE host galaxy,limited by the TDE host with the fewest matches. Wethen calculate the fraction of quiescent Balmer-stronggalaxies in the control. We do this matching for one pa-rameter at a time and for several simultaneously. Thisallows us to control for possible selection effects in dif-ferent observables without needing to understand the(likely complicated) exact form of the selection effect.Our results are relatively insensitive to the matchingtechnique and absolute or fractional tolerances used.4.2. Overview of Selection Effect Matching Results
Table 3 lists the fraction of quiescent Balmer-stronggalaxies in our control samples for both individual andsimultaneous matches. This table also includes re-sults from matching on galaxy S´ersic index ( n g ) andbulge-to-total-light ratio ( B/T ), discussed in Section 5.We find that matching individually on redshift or BH idal Disruption Event Host Galaxies Table 3.
Fraction of quiescent Balmer-strong galaxies in control samples matched toTDE hosts 1-5. We tested all combinations of these properties, but only list combina-tions that (1) result in enough controls to compute a reliable fraction of sF16 or wF16galaxies, (2) lead to an increase in these fractions, and (3) are interesting in comparisonwith similar combinations.Properties matched z M bh g − r hl , g n g B/T ) g z , bulge g − r z , Σ hl , g M bh , bulge g − r M bh , n g M bh , ( B/T ) g g − r , Σ hl , g g − r , n g g − r , ( B/T ) g Σ hl , g , ( B/T ) g z , M bh , bulge g − r z , M bh , n g z , M bh , ( B/T ) g z , bulge g − r , n g
279 6.05 z , bulge g − r , ( B/T ) g z , Σ hl , g , n g
185 2 1.08 9 4.86 M bh , bulge g − r , Σ hl , g M bh , bulge g − r , n g M bh , bulge g − r , ( B/T ) g M bh , Σ hl , g , n g
215 0 0.0 19 M bh , Σ hl , g , ( B/T ) g
580 3 0.52 48 8.28 M bh , n g , ( B/T ) g g − r , Σ hl , g , n g g − r , Σ hl , g , ( B/T ) g
161 7.2bulge g − r , n g , ( B/T ) g z , M bh , bulge g − r , Σ hl , g
285 1 0.35 17 z , M bh , bulge g − r , n g z , M bh , n g , ( B/T ) g
440 0 0.0 40 z , bulge g − r , Σ hl , g , n g
110 2 1.82 7 6.36 M bh , bulge g − r , n g , ( B/T ) g
945 3 0.32 79 8.36
Note —Σ hl , g is the g -band half-light surface brightness, n g is the galaxy S´ersic index,and ( B/T ) g is the g -band bulge-to-total-light ratio. Bold numbers highlight particu-larly large enhancements in the fraction of sF16 or wF16 galaxies. Law-Smith et al.
Table 4.
Fraction of quiescent Balmer-strong galaxies in samples created with simple cutson the reference catalog; simultaneous matching (as in Table 3) on these combinations ofparameters returns few controls. These cuts are chosen to include TDE host galaxies 1, 2,and 4 (all quiescent Balmer-strong).Sample a :5 . < log( M bh /M (cid:12) ) < . z < . g − r < .
51, Σ hl , g > .
05 4301 33 0.77 118 2.74Cut B: cut A plus no S/N > n g > .
24 1662 28 1.68 85 5.11Cut D: cut B plus (
B/T ) g > .
55 1807 25 1.38 78 4.32
Note —Σ hl , g is the g -band half-light surface brightness, n g is the galaxy S´ersic index, and( B/T ) g is the g -band bulge-to-total-light ratio. a Of the three TDE hosts (in 1-5) that pass cut A, one of three are sF16 and three of threeare wF16. All of these also pass cuts B, C, and D.
Table 5.
Medians of 1D Distributions in Samples Matched in BH Mass to TDE Host Galaxies 1-5Parameter Control TDE hosts (1-5) TDE hosts (1-10) sF16 wF16Bulge g − r [mag] 0 . +1 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . Σ hl , g [mag/arcsec ] 0 . +1 . − . . +2 . − . . +2 . − . . +8 . − . . +2 . − . Galaxy S´ersic index 1 . +1 . − . . +0 . − . . +1 . − . . +2 . − . . +2 . − . ( B/T ) g . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . M g, galaxy [mag] − . +1 . − . − . +0 . − . − . +1 . − . − . +1 . − . − . +1 . − . M g, bulge [mag] − . +2 . − . − . +0 . − . − . +0 . − . − . +1 . − . − . +1 . − . Galaxy g − r [mag] 0 . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . Inclination 49 . +21 . − . . +26 . − . . +19 . − . . +16 . − . . +22 . − . Note — ± values indicate 84th and 16th percentiles. Σ hl , g is the g -band half-light surface brightness,( B/T ) g is the g -band bulge-to-total-light ratio, and M g is the g -band absolute magnitude. Results for r -band are similar. idal Disruption Event Host Galaxies ∼
10 (from 0.2%to 2%). Matching on half-light surface brightness in-creases the fraction of quiescent Balmer-strong galaxiesonly slightly. Matching on several parameters simulta-neously can increase the fraction of quiescent Balmer-strong galaxies by a factor of ∼
4. The number of con-trols within our tolerances is often too few to computea reliable sF16 fraction when matching on more thanthree parameters—we address this with a simple cut-based approach later in this section, and the results ofthese cuts are listed in Table 4.Figure 4 shows 2D distributions of redshift, bulge col-ors, and half-light surface brightness versus BH massas well as their 1D distributions matched on the BHmasses of TDE hosts 1-5 and split into different sub-samples. This figure includes measurements for TDEhosts numbered 6-10, though the matching analysis (re-sulting in Table 3) uses only TDE host galaxies 1-5.The 1D distributions are all smoothed and normalizedto equal area. These are shown for presentation pur-poses only—the smoothing does not enter into or affectour analysis. Red distributions correspond to TDE hostgalaxies 1-5. We show the unsmoothed histograms forTDE hosts 1-5 in solid red and those for hosts 1-10 indotted black. The different subsamples shown on theright-hand side of Figure 4 are, from top to bottom,the TDE host galaxies, quiescent Balmer-strong galax-ies (sF16 and wF16 selections), and AGNs and low-S/NAGNs. We compare to AGNs both to better under-stand the connection between TDEs and AGNs (see,e.g., Auchettl et al. 2017b), and because there may be abias against detecting TDEs in galaxies hosting a strongAGN, as discussed later in this section. Table 5 lists themedians and spreads of these 1D distributions matchedon BH mass, as well as for some properties consideredin Section 5 and Appendix A.4.3.
BH Mass
The first selection effect we consider is BH mass. Aspresented in the tidal disruption menu of Law-Smithet al. (2017), most main sequence (MS) stars cannot bedisrupted outside the innermost bound circular orbit ofBHs of M bh (cid:38) . M (cid:12) . Giant stars can be disruptedby higher-mass BHs, but their relative disruption rate islower and their flares last on the order of years (MacLeodet al. 2012) and may not be seen in current surveys. BHs Though combinations of rapid BH spin and favorable orbitalorientation can permit MS disruption by BHs with masses of upto a few × M (cid:12) (Kesden 2012). with M bh (cid:46) M (cid:12) , on the other hand, may inefficientlycircularize the debris from the majority of MS star dis-ruptions (Guillochon & Ramirez-Ruiz 2015), leading toa delayed flare that is more difficult to detect due to alower luminosity and longer timescale (Hayasaki et al.2013; Dai et al. 2013; Cheng & Bogdanovi´c 2014; Bon-nerot et al. 2016). These lower-mass BHs can disruptdenser objects such as white dwarfs, but we do not ex-pect this to be a significant contribution to the currentsample of observed TDEs (MacLeod et al. 2016).BH mass is thus a primary physical constraint inwhether a TDE can be observed in a given galaxy, andwe control for it by creating matched samples with sim-ilar distributions in BH mass to the observed TDE hostgalaxies. We obtain BH masses from velocity disper-sions, though our conclusions are insensitive to the exactmethod used to derive BH masses (see Section 2). Aslong as BH mass is determined homogeneously betweenthe different samples we consider, the effect of our un-certainty on BH mass is largely mitigated; i.e., we donot rely on accurate determinations of BH mass for ourconclusions. In fact, if we simply replace BH mass with M (cid:63), tot throughout our analysis, our conclusions remainthe same. Matching on only BH mass to TDE hosts 1-5(see Table 3) slightly decreases the fraction of quiescentBalmer-strong galaxies in our reference catalog.We expect (and find; see Figure 4, or Wevers et al.(2017) for TDE hosts not in our reference catalog) nearlyall currently observed TDEs to occur in the BH massrange of 10 . < M bh /M (cid:12) < . . TDE host galaxiesthus have significantly lower BH masses than the bulkof our reference catalog. We plot each of the propertiesdiscussed below versus BH mass on the left-hand sideof Figure 4, and we control for BH mass (matching onTDE hosts 1-5) in the stacked 1D distributions on theright-hand side.4.4. Redshift Completeness
The second selection effect we consider is redshift com-pleteness. The luminosity of a tidal disruption flaredepends on the stellar mass, stellar radius, BH mass,impact parameter, and circularization efficiency of thedebris. For a typical TDE, however, the maximum peakluminosity does not vary by more than an order of mag-nitude, and most observed TDEs have peak luminositiesof 10 to 10 erg s − . Additionally, most TDEs appearto be sub-Eddington or Eddington limited for their BHs(Hung et al. 2017).A typical TDE can thus only be observed out to aredshift that depends on the detection limits of the tele-scope and, especially if the flare is Eddington limited,the mass of the BH. Strubbe & Quataert (2009) and2 Law-Smith et al.
Kochanek (2016) studied the dependence of TDE rateson BH mass and redshift in detail; generally, they foundthat detection rates decrease rapidly with redshift, butthat future surveys such as LSST could be sensitive toa sizable number of TDEs at z >
1. TDE detectabilityis a strong function of redshift, and we control for thisby creating samples matched on redshift to the TDEhost galaxies in our sample. Matching on only redshift(see Table 3) slightly decreases the fraction of quiescentBalmer-strong galaxies in our reference catalog.We show redshift versus BH mass for TDE host galax-ies and our reference catalog in the top-left panel of Fig-ure 4. All but one of the TDE hosts in our sample have z < .
1. The top-right panel of Figure 4 shows that,after controlling for BH mass, the redshift distributionsof the TDE host galaxies, sF16 and wF16 galaxies, andboth low- and high-S/N AGNs are similar to that of ourreference catalog. We note that the “sF16” classifica-tion of E+A galaxies is likely to change with redshiftif it is based on a single slit width. Although this is asmall effect for galaxies with z < .
1, it is sizable whencomparing to galaxies at z > .
2. The fraction of sF16galaxies in our reference catalog as a function of redshiftbin (we took ∆ z = 0 . z (cid:46) .
1, is relatively con-stant at the nominal (cid:39) z (cid:38) .
25, it can be > Bulge Colors
The third selection effect we consider is the color of thegalaxy bulge, as dusty, red bulges might obscure TDEs.Indeed (see below), TDE hosts have bluer bulge colorsthan most galaxies. We thus control for this possible se-lection effect by creating samples matched in bulge g − r to TDE host galaxies 1-5. Controlling for only bulge g − r (see Table 3) results in a large increase in the fraction ofE+A galaxies in our reference catalog: compared to thenominal percentage of 0.2% sF16 galaxies, the matchedsample has 2% sF16 galaxies, a factor of ∼
10 increase.The fraction of wF16 galaxies increases by a factor oftwo, from 2 .
3% to 5 . / /
10 = 30% including hosts not in our reference cata-log) are in the sF16 selection and 3/5 (or 8/10) are inthe wF16 selection.The middle-left panel of Figure 4 shows bulge g − r color versus BH mass for the TDE host galaxies and ourreference catalog. In the middle-right panel (where wehave controlled for BH mass), we see that TDE hosts The color of the core/nucleus may be more relevant for TDEdetectability but this measurement is not available for nearly asmany galaxies as in our catalog drawn from SDSS. have bluer bulge g − r colors than the reference catalog,suggesting a preference against observing TDEs in red-der bulges. This is also seen clearly in the sF16 samplebut only very weakly in the wF16 sample. The AGN andlow-S/N AGN samples appear similar to the referencecatalog, with the low-S/N AGN sample showing slightlybluer bulge colors than the AGN sample. TDE hosts 1-5have a median bulge g − r of 0.46 mag, and TDE hosts1-10 of 0.49 mag. After matching in BH mass to TDEhosts 1-5, the reference catalog has a median bulge g − r of 0.78 mag, and sF16 galaxies of 0.42 mag. This is alsolisted in Table 5. So both TDE hosts and E+A/post-starburst galaxies have median bulge colors ∼ Half-light Surface Brightness
The fourth selection effect we consider is on surfacebrightness, as image subtraction for transients mightbe more challenging for high surface brightness galax-ies. We define the half-light surface brightness, Σ hl ,as half the galaxy apparent magnitude divided by thegalaxy half-light size in square arcseconds. Using theSimard et al. (2011) measurements, this is Σ hl,g =( g g d + 0 . /πθ , where g g d is the g -band appar-ent magnitude of the GIM2D output pure S´ersic model, g g d +0 . θ hl = R chl,g / Scale. R chl,g is the circular half-light radius in the g band, andScale is the physical scale in kpc/arcsec at redshift z .We use only g band, as results for r band, or a combi-nation of both g and r bands, are similar. Controllingfor Σ hl,g (see Table 3) increases the fraction of quiescentBalmer-strong galaxies in our reference catalog slightly,from 2.3% to 3.1% for wF16 galaxies.The bottom-left panel of Figure 4 shows the half-lightsurface brightness versus BH mass for the TDE hostgalaxies and our reference catalog. It is evident thatTDEs are found preferentially in galaxies with lowerhalf-light surface brightnesses. In the right panel (con-trolled for BH mass), we see that sF16 and wF16 galax-ies have slightly lower half-light surface brightnessesthan the reference catalog, and that AGN and SF galaxies have higher half-light surface brightnesses thaneither TDE hosts or quiescent Balmer-strong galaxies.TDE host galaxies 1-5 have a median half-light surfacebrightness of 2.06 mag/arcsec , and TDE hosts 1-10 of1.95 mag/arcsec . After matching in BH mass to TDEhosts 1-5, the reference catalog has a median half-lightsurface brightness of 0.98 mag/arcsec , and sF16 galax-ies of 1.58 mag/arcsec ; this is also shown in Table 5. So, Low-S/N AGNs have a distribution in Σ hl,g that is indistin-guishable from that of AGNs here, so we show SF galaxies instead. idal Disruption Event Host Galaxies log M bh /M fl z T D E ho s t s c on t r o l E + A sF16wF16 z AGN
AGNlow S/N AGN log M bh /M fl bu l g e g − r T D E ho s t s c on t r o l E + A sF16wF16 bulge g − r AGN
AGNlow S/N AGN log M bh /M fl Σ h l , g [ m ag / s q . ] T D E ho s t s c on t r o l E + A sF16wF16 Σ hl , g [mag / sq . ] AGN
AGNSF
Figure 4.
Left panels, top to bottom: redshift, bulge g − r , and half-light surface brightness vs. BH mass for TDE hostgalaxies (numbered points) and our reference catalog (contours). Galaxies 1-5 are used in our matching analysis. BH massesfor 8, 9, and 10 are determined via M (cid:63), bulge . Contours are spaced by 0 . σ , with the darkest shading containing 0 . σ and thelightest shading containing 2 σ . Median errors in the TDE host galaxy measurements are shown in the top left. Right panels:1D distributions in these properties in different subsamples, matched on BH mass of TDE hosts 1-5. From top to bottom ineach panel, the subsamples are: TDE host galaxies (1-5 in red, showing both smoothed and actual distributions, and 1-10 indotted black), our reference catalog (black), the strong F16 selection (orange), weak F16 selection (light blue), AGN (green),and low-S/N AGN (purple). In the bottom-right panel, we show SF galaxies, rather than low-S/N AGN, in purple, as thesehave a very similar distribution to the AGN. All 1D histograms are smoothed and normalized to equal area. Law-Smith et al.
TDE hosts and sF16 galaxies have median half-light sur-face brightnesses ∼ and ∼ fainter than the control sample, respectively.4.7. Galaxies Hosting a Strong AGN
We also consider a possible selection effect based onthe presence of a strong AGN. Observational identifi-cation of TDEs is biased against galaxies with strongAGN, as it is difficult to distinguish a TDE signal fromregular variability in a strong AGN. Galaxies with astrong AGN are often not considered for spectral follow-up on potential TDEs. As a first-order exploration ofthis selection effect, we performed a cut on all AGNswith S/N > >
3) from our referencecatalog. Applying this cut—both individually and incombination with controlling for other parameters—hasa relatively small effect, but does slightly increase thefraction of quiescent Balmer-strong galaxies in our con-trol samples. We do not show this high-S/N AGN cutin Table 3 for clarity, as its effect is generally small, butwe show its effect in combination with other simple cuts(described below) in Table 4.4.8.
Cumulative Effect
Individually, controlling for BH mass or redshiftslightly decreases the fraction of quiescent Balmer-strong galaxies in our control sample, controlling forhalf-light surface brightness slightly increases this frac-tion, and controlling for bulge g − r increases the fractionof sF16 galaxies by a factor of ∼
10 and of wF16 galax-ies by a factor of ∼
2. Table 3 also lists the effect ofcontrolling for these parameters simultaneously. Notethat this table also includes results from matching ontwo indicators of central light concentration: the galaxyS´ersic index ( n g ) and bulge-to-total-light ratio ( B/T );we discuss these parameters in Section 5. We summarizeour major findings with respect to matching on M bh , z , bulge g − r , and Σ hl , g below. Of these four observ-ables, bulge g − r is the most important in increasingthe fraction of quiescent Balmer-strong galaxies in ourcontrol samples. However, its effect is largest on thefraction of sF16 galaxies (and remains similar for wF16galaxies) when matched on individually. Matching si-multaneously on z , M bh , and bulge g − r increases thepercentage of sF16 galaxies by a factor of 2.5, to 0.5%.Matching simultaneously on all four parameters resultsin too few controls to calculate the fraction of sF16galaxies but increases the fraction of wF16 galaxies bya factor of 2.6, to 6%. Recall that 1/5=20% of theTDE host galaxies used in our matching analysis (or3/10=30% including galaxies not in our reference cat-alog) are sF16 galaxies, and 3/5=60% (or 8/10=80%)are wF16 galaxies. As matching simultaneously on z , M bh , bulge g − r ,and Σ hl,g results in too few controls to calculate thefraction of sF16 galaxies, we perform simple cuts onour full reference catalog as a cruder probe of the ex-tent of these selection effects; this is shown in Table 4.We chose cuts that are consistent with the proper-ties of our quiescent Balmer-strong TDE host galaxies:5 . < log( M bh /M (cid:12) ) < . z < .
09, bulge g − r < . hl , g > .
05. This results in a sample with 0.77%sF16 galaxies (a factor of ∼ > ∼ ∼ ∼
25 (from ∼ ∼
10 (from ∼ ∼
38 (from ∼ ∼
13 (from ∼ Small Sample Size
As mentioned, only 1/5 of the TDE hosts we use in ourmatching analysis is in the sF16 selection. This is toosensitive to small number statistics to claim a true over-representation in sF16 galaxies using this sample alone.However, the sF16 overrepresentation increases with thelarger sample of 10 photometrically identified TDE hostswith available H α EW and Lick H δ A measurements, andso it appears robust. Restricting ourselves to the fiveTDE hosts we use in the matching, however, we can testthe probability of having 1/5 sF16 galaxies (i.e., the ex-tent to which this is due to small number statistics) ina Monte Carlo approach. We create 10,000 samples offive galaxies drawn from our full catalog and matchedin M bh and z to the five TDE hosts. For each of these10,000 samples of five galaxies, we count the number (ifany) that fall in the sF16 selection. The percentage ofthe samples that have 1/5 or more sF16 galaxies is 1.0%;this is the chance likelihood of having 1/5 sF16 galaxiesin our sample. PHYSICAL ENHANCEMENTS TO THE TDERATE idal Disruption Event Host Galaxies T D E ho s t s c on t r o l E + A sF16wF16 galaxy S´ersic index AGN
AGNlow S/N AGN log M bh /M fl ( B / T ) g T D E ho s t s c on t r o l E + A sF16wF16 ( B/T ) g AGN
AGNlow S/N AGN
Figure 5.
Top-left panel: galaxy S´ersic index vs. BH mass for TDE host galaxies and our reference catalog. We use TDEhosts 1-5 in our matching analysis. BH masses for TDE hosts 8, 9, and 10 are determined via M (cid:63), bulge . Contours are spaced by0 . σ , with the darkest shading containing 0 . σ and the lightest shading containing 2 σ . Average errors in the TDE host galaxymeasurements are shown in the top left. The region above the light green line contains ∼
2% of our reference catalog galaxiesbut 5/5 (or 6/10) of our TDE host galaxies. Top-right panel: galaxy S´ersic index distribution in different subsamples, matchedon BH mass of TDE hosts 1-5. 1D histograms are smoothed and normalized to equal area. Unsmoothed 1D histograms are alsoshown for TDE hosts 1-5 in solid red and for TDE hosts 1-10 in dotted black. Bottom panels: g -band bulge-to-total-light ratio(bulge fraction); similar description to that above. Results are similar for r band. In this section, we consider two possible alternative(physical) explanations for the enhanced frequency ofTDEs in quiescent Balmer-strong galaxies: (1) highercentral stellar densities and (2) a recent merger. Inexploring the first, we find a new unique property ofall TDE host galaxies, regardless of E+A or quiescentBalmer-strong classification.5.1.
Higher Central Stellar Densities
S´ersic Index
It is expected that a higher stellar density in the nu-clear star cluster surrounding an SMBH leads to a highertidal disruption rate, as there are more dynamical en-counters between stars and therefore more scatteringsinto the loss cone (e.g., Magorrian & Tremaine 1999).The galaxy S´ersic index is a broad indicator of the steep-ness of a galaxy’s light profile and thus (to a certain ex-tent) its stellar density profile. A S´ersic profile has theform ln I ( R ) = ln I − kR /n , (4)6 Law-Smith et al. where I is the intensity, I is the intensity at R = 0, k is a constant, and n is the S´ersic index. The higherthe S´ersic index, the more centrally concentrated thegalaxy’s light profile.The top panel of Figure 5 shows the galaxy S´ersicindex versus BH mass for TDE hosts and our referencecatalog. Note that the S´ersic index fits are allowed tovary from 0.5 to 8 (see Simard et al. 2011). The regionabove the light green line contains ∼
2% of our referencecatalog galaxies but 5/5 of TDE host galaxies 1-5 (or6/10 including the “extreme coronal line emitters” 6-9and 10 with a BH mass determined via M (cid:63), bulge ) . Wecompute the fraction of reference catalog galaxies thathave a higher S´ersic index than each TDE host galaxyat its BH mass (in a bin of width 0.02 dex); we find thatall of the TDE host galaxies have high S´ersic indicesfor their BH masses. On average, TDE hosts 1-5 havegalaxy S´ersic indices in the top 10% of those of referencecatalog galaxies at their BH masses. Including TDEhost galaxies 6-10 results in an average S´ersic index inthe top 15%.In the top-right panel of Figure 5, we create a samplematched on the BH masses of TDE hosts 1-5 and com-pare distributions between different subsamples. TDEhosts have a much broader (toward higher values) dis-tribution of S´ersic indices than the control sample. Thedistribution for sF16 galaxies is similarly weighted to-ward high S´ersic indices, and wF16 galaxies show asimilar but weaker effect. AGNs and low-S/N AGNsshow a fairly similar distribution to the reference cat-alog, though with a slight preference for higher S´ersicindices. TDE host galaxies 1-5 have a median galaxyS´ersic index of 4.03, and TDE hosts 1-10 of 4.30. Af-ter controlling for BH mass, our reference catalog hasa median galaxy S´ersic index of 1.21, and sF16 galax-ies of 4.33. This is also listed in Table 5. So both TDEhost galaxies and E+A/post-starburst galaxies have rel-atively high galaxy S´ersic indices, especially after con-trolling for BH mass.Importantly, all of our TDE host galaxies, regardlessof E+A or quiescent Balmer-strong classification, havehigh galaxy S´ersic indices for their BH masses. We havethus identified a photometric criterion (S´ersic index) This is for a single S´ersic fit, also referred to as n g throughoutthe paper. We show galaxy rather than bulge S´ersic index here, asonly a fraction of our TDE host galaxies and the reference cataloghave high enough resolution data to justify (see Section 2) free- n S´ersic fits to their bulges. We note, however, that in this justifiedfree- n b sample, Simard et al. (2011) find that galaxies with lowand high n b values also have low and high n g values. Note that this region is drawn to include these TDE hostgalaxies. that may predict an enhanced TDE rate more broadlythan a spectroscopic criterion (E+A classification).5.1.2.
Bulge-to-total-light Ratio
The relatively high central concentration of light inTDE host galaxies is also apparent in their bulge-to-total-light ratios. We show the g -band bulge fraction,( B/T ) g , in the bottom panel of Figure 5. Results aresimilar for r band. Similarly to above, we computethe fraction of reference catalog galaxies that have ahigher ( B/T ) g than each TDE host galaxy at its BHmass. Again, we find that all of the TDE host galax-ies have high bulge fractions for their BH masses: onaverage in the top 10% for TDE host galaxies 1-5, andin the top 20% for TDE hosts 1-10. Controlling for BHmass (right panel), we see that both TDE host galax-ies and quiescent Balmer-strong galaxies have signifi-cantly higher bulge fractions than our reference cata-log. This is related to our galaxy S´ersic index result:observed TDEs show a preference for centrally concen-trated, bulge-dominated galaxies (and/or tend to avoiddisk-dominated galaxies). The bulge fraction and theS´ersic index are correlated—see Figure 14 of Simardet al. (2011) or our Figure 11. TDE host galaxies 1-5 have a median ( B/T ) g of 0.56 (0.54 for TDE hosts1-10). Matched on BH mass to TDE hosts 1-5, the ref-erence catalog has a median ( B/T ) g of 0.06, and sF16galaxies of 0.60. This is also listed in Table 5.5.1.3. Can S´ersic Index or B/T Explain the QuiescentBalmer-strong Overrepresentation?
We perform matches on galaxy S´ersic index and bulgefraction, as we did for the selection effects considered inSection 4, and compute the fraction of quiescent Balmer-strong galaxies in the matched samples. The results ofthis matching (using TDE hosts 1-5 as before) are listedin Table 3; we will summarize below. We tested all com-binations of the properties we studied, but only list com-binations that (1) result in enough controls to computea reliable fraction of sF16 or wF16 galaxies, (2) lead toan increase in these fractions, and (3) are interesting incomparison with similar combinations. Matching onlyon S´ersic index slightly decreases the fraction of sF16galaxies in our control sample, while matching only on(
B/T ) g increases the fraction of sF16 and wF16 galaxiesby a factor of ∼ n g or B/T in combination with the selection effects we consid-ered in Section 4 can increase the fraction of quiescentBalmer-strong galaxies in our reference catalog by a fac-tor of 3-10 depending on the combination. The highestincrease in the fraction of sF16 galaxies—from 0.2% to2.45%—is given by matching on bulge g − r and ( B/T ) g .The highest increase in the fraction of wF16 galaxies— idal Disruption Event Host Galaxies z , M bh , n g ,and ( B/T ) g .Simultaneously matching on either S´ersic index or B/T and on all of the four parameters we consideredearlier returns too few controls to compute the fractionof quiescent Balmer-strong galaxies. We perform simplecuts on our full reference catalog as a cruder probe ofthe effect of controlling for five of these properties; theseare listed in Table 4. We chose cuts that are consistentwith the properties of our quiescent Balmer-strong TDEhost galaxies: 5 . < log( M bh /M (cid:12) ) < . z < . g − r < .
51, Σ hl , g > .
05, and n g > .
24 and/or(
B/T ) g > .
55. These result in samples with ∼ ∼ ∼ Merger Indicators
Galaxy mergers might also enhance the TDE rate ifthey trigger binary BH inspiraling. To study this, weuse the RA g and r bands. For recent major mergers, the asymmetryindicators can be (cid:38) .
04 (e.g., Patton et al. 2016). Theasymmetry indicators are shown plotted against eachother as well as against BH mass for TDE host galaxiesand our reference catalog in Figure 6. In the right twopanels, we show only g -band measurements; results aresimilar for r band. Interestingly, our TDE host galaxieshave small asymmetry indicators, suggesting that theyare not the products of recent major mergers. It is im-portant to note that a small SDSS asymmetry indicatordoes not necessarily correspond to the lack of a merger.The major limitation here is the SDSS resolution, alongwith the fact that asymmetries in mergers tend only tobe high for major mergers and gas-rich galaxies (Lotzet al. 2010; Ji et al. 2014). In the right panel (con-trolled for BH mass), we see that TDE host galaxiesand quiescent Balmer-strong galaxies both have a nar-rower distribution in asymmetry indicator compared tothe reference catalog. They do not share the referencecatalog’s tail toward high asymmetry indicators.5.3. Summary of TDE Host Galaxy Properties; SDSSImages
In summary, TDE host galaxies tend to have bluerbulges, lower half-light surface brightnesses, and morecentrally concentrated light profiles (in S´ersic index and bulge fraction) than “typical” galaxies at their BHmasses. As an illustrative example of this, in Figure 7,we show SDSS images of TDE host galaxies 1-5 aswell as, for each TDE host galaxy, a randomly selectedgalaxy matched in BH mass and redshift to the TDEhost galaxy, but with a galaxy S´ersic index, bulge g − r ,and half-light surface brightness very close to the medianvalues of our reference catalog at that BH mass. Thegalaxies in the bottom panels are thus “typical” galax-ies in a few of the parameters we considered above, butmatched in BH mass and redshift to the TDE host galax-ies. The higher central concentration of these TDE hostgalaxies is visually apparent. The galaxy S´ersic index ofeach galaxy is listed in the top right of each image. DISCUSSIONIn this section, we discuss (1) whether there is a selec-tion effect against detecting TDEs in SF galaxies, (2) ifthe observed time delay between SF and AGN activitycan help us understand the nature of the post-starburstTDE delay, (3) if the quiescent Balmer-strong enhance-ment can be understood in terms of S´ersic indices in thegreen valley, and (4) the implications of higher centrallight concentrations on the TDE rate. Before discussingthese issues, we list a summary of our key findings.6.1.
Summary
We studied TDE host galaxies in the context of a cat-alog of ∼ ∼ ∼× ∼× g − r colors (by ∼ ∼ ) than galaxies in our referencecatalog. TDE host galaxies have low galaxy asym-metry indicators, suggesting that they are not theresult of a recent major merger.3. TDE hosts and E+A galaxies have high galaxyS´ersic indices and high B/T for their BH masses, Though higher resolution imaging is available for several ofthese galaxies, we show SDSS photometry here as this is what wasused to derive the photometric galaxy properties in our referencecatalog. Law-Smith et al. A s y m (r)
12 345 6 78910 log M bh /M fl A s y m ( g ) T D E ho s t s c on t r o l E + A sF16wF16 AGN
AGNlow S/N AGN
Figure 6.
Left panel: galaxy asymmetry indicators in the g and r bands for TDE host galaxies and our reference catalog.Middle panel: asymmetry indicator in the g band vs. BH mass. The r band is similar. BH masses for TDE hosts 8, 9, and 10are determined via M (cid:63), bulge . Contours are spaced by 0 . σ , with the darkest shading containing 0 . σ and the lightest shadingcontaining 2 σ . Average errors in the TDE host galaxy measurements are shown in the top left (asymmetry indicators do nothave associated errors in our catalog). Right panel: asymmetry indicator distribution in the g band in different subsamples,matched on BH mass of TDE hosts 1-5. The r band is similar. 1D histograms are smoothed (we also show the true histogramfor the TDE hosts 1-5 in red and 1-10 in dotted black) and normalized to equal area. n g = 2 . n g = 4 . n g = 4 . n g = 2 . n g = 5 . n g = 1 . n g = 1 . n g = 1 . n g = 1 . n g = 1 . Figure 7.
Top panels: SDSS gri images of TDE host galaxies 1-5. Bottom panels: for each TDE host galaxy, a randomlyselected galaxy matched in BH mass and redshift to the TDE host galaxy, but with a galaxy S´ersic index, bulge g − r , andhalf-light surface brightness very close to the median values of our reference catalog at that BH mass. Images are 20 (cid:48)(cid:48) × (cid:48)(cid:48) .The galaxy S´ersic index of each galaxy is listed in the top right of each image. suggesting a higher stellar density in their cores.On average, our TDE host galaxies have galaxyS´ersic indices and bulge fractions in the top 10%of those of reference catalog galaxies at their BHmasses. We identify a region in galaxy S´ersic indexversus BH mass space that contains ∼
2% of ourreference catalog galaxies but 5/5 (or 6/10) of ourTDE host galaxies.We also note that Graur et al. (2017) appeared onarXiv after submission of this work and is an indepen-dent and complementary analysis of TDE host galaxyproperties. They study a smaller set of host galaxy prop- erties, but, importantly, find that TDE host galaxieshave high stellar surface mass densities. This is simi-lar to our finding that TDE host galaxies appear morecentrally concentrated, with higher galaxy S´ersic indicesand
B/T . They control for the type of galaxy in whichthe TDE is found (quiescent or star-forming) and findthat this result is driven particularly by the star-forminghosts.6.2.
Is There a Selection Effect Against DetectingTDEs in SF Galaxies?
In order to understand the TDE rate enhancementin quiescent Balmer-strong galaxies, it is important to idal Disruption Event Host Galaxies
Chandra
X-ray data, indicating that they might be heav-ily obscured. As a case study, the Seyfert 2 galaxy NGC4968 is found to have heavy obscuration and circumnu-clear SF, as well as SF-associated gas that may increasethe covering factor of the enshrouding gas and play arole in obscuring the AGN (LaMassa et al. 2017). If aTDE occurred in NGC 4968, the large column density( N H > . × cm − ) would prevent X-ray—andoptical/UV, for dust-to-gas ratios similar to the MilkyWay—identification.Although difficult to detect, AGNs are often found inSF galaxies (Bongiorno et al. 2012; Ellison et al. 2016).Nuclear activity in SF galaxies might be fairly common,and it is possible that TDEs are missed in these galaxiesprimarily due to selection effects. We note that Tad-hunter et al. (2017) discovered a TDE in a nearby ultra-luminous infrared galaxy. The galaxy features suggestthat there is an unusually clear view of the nuclear star-forming region, whose obscuration is known to have acomplex structure (e.g., Buchner & Bauer 2017). Twogalaxies in our TDE host galaxy sample, SDSS J0748and SDSS J1342 (Wang et al. 2012), are in SF galax-ies. We note, however, that both events were classifiedas TDEs in a search for extreme coronal line emitters,and in both cases were not identified based on their lightcurve properties.Second, we recognize that there could be a bias againstTDE identification in SF galaxies, as TDE characteriza-tion for nuclear transients is not done systematically. Inaddition, current TDE host galaxies have relatively lowhalf-light surface brightnesses, while SF galaxies haverelatively high half-light surface brightnesses (see Fig-ure 4). If image subtraction is less robust for thesegalaxies, this could help explain the lack of TDEs inSF galaxies. Though we caution that this strongly depends on how theAGN is selected (see e.g., Ellison et al. 2016). This galaxy is not in our sample as it is not in the SDSScatalogs we draw from.
Time Delay between SF and AGN/TDEs
If TDEs occur preferentially in post-starburst galax-ies, then the TDE rate is time dependent and, in par-ticular, depends on the recent SF history of the galaxy.If the starburst was caused by a galaxy merger, the sec-ond inspiraling BH could certainly enhance the TDErate; we discuss this further below, but first we considerthe relationship between the SF and the TDE. After astarburst occurs, the gas is transported inwards on sometimescale and may drive an enhanced TDE rate (possi-bly through contributing to disk instabilities; Madiganet al. 2017).This transport timescale is also seen in AGN activity,and there is an intriguing connection between AGN ac-tivity and SF episodes. AGN activity appears to be trig-gered by the same gas that drives SF episodes (Trumpet al. 2015), but with a time delay, presumably due tothe transport timescale of gas to the BH. The post-starburst timescale in TDE hosts appears tantalizinglysimilar to this observed AGN activity–SF episode delay.Wild et al. (2010) study a sample of 400 galaxies withBH masses of 10 . -10 . M (cid:12) that have experienced astarburst in the past 600 Myr. They find that the aver-age rate of accretion of matter onto the BH rises steeply ≈
250 Myr after the starburst begins. Similarly, Davieset al. (2007) study the nuclei of nine AGNs at spatialscales of ≈
10 pc and find a hint of a delay of 50-100Myr between the onset of star formation and accretiononto the BH. This delay is strikingly reminiscent of thepost-starburst timescale inferred for TDE hosts, whichhave post-starburst ages of 10-1000 Myr (French et al.2017) .LaMassa et al. (2013) study the connection betweenAGN activity and star formation with a sample of ≈ Though note that only 4/8 of their sample have post-starburst ages of <
250 Myr. Law-Smith et al.
The Green Valley, S´ersic Index, andE+A/Post-Starburst Galaxies
In Figure 3, we saw that our TDE host galaxies lie be-low the SFMS, but not by more than 1.0 dex (some arein the green valley and some appear near it), and thatE+A/post-starburst galaxies inhabit a similar region.In the left panel of Figure 8, we show SFR versus M ∗ , tot for our TDE host galaxies and the reference catalog,similarly to Figure 3, but color-coded by galaxy S´ersicindex; this shows the evolution of galaxy surface den-sity profiles in this parameter space. TDE host galaxiesclearly inhabit a transition region in S´ersic index.In the right panel of Figure 8, we show normalized his-tograms of S´ersic index for different subsamples of thereference catalog: the band from the SFMS to 0.5 dexbelow the SFMS (band 1), the band 0.5 dex below this,from the dashed to the dotted line (band 2), and galaxiesin the sF16 selection. Though their population is small,E+A galaxies have much higher S´ersic indices relativeto the larger population of galaxies in the green val-ley. If we restrict our reference catalog to galaxies withlog( M (cid:63), tot /M (cid:12) ) < .
5, to match the TDE hosts, as wellas to galaxy S´ersic indices of > .
0, the percentage ofsF16 (wF16) galaxies in band 2 is 1.9% (16%). Furtherrestricting to galaxies with S´ersic indices > . On the TDE Rate Enhancement
The increased TDE rates in E+A/post-starburstgalaxies might be explained by mergers and/or by highernuclear stellar densities.A merger is thought to increase the disruption rateby many orders of magnitude (e.g., Ivanov et al. 2005).But these rate enhancements are short lived (less than 1Myr), so that the fraction of TDEs resulting from merg-ing BHs is expected to be low (Wegg & Nate Bode 2011).The asymmetry indicators for TDE hosts are compara-tively small, meaning that they show no obvious signsof recent major mergers. However, TDE hosts could bethe product of high mass ratio mergers. Indeed, unequalmass ratio mergers are more effective at enhancing theTDE rate as stars are scattered into the loss cone of dis-ruptive orbits more efficiently (Chen et al. 2009). Frenchet al. (2017) find that the post-starburst ages of TDEhosts, if the starburst arose from a galaxy merger, are consistent with mergers of mass ratios more equal than12:1 for most hosts, which is still consistent with ourfindings.Additionally, higher resolution observations could re-veal signs of mergers that the asymmetry indicatorsmiss. Using MUSE integral field unit (IFU) spec-troscopy observations, Prieto et al. (2016) find that thehost galaxy of ASASSN-14li shows asymmetric and fila-mentary structures—signs of a recent merger—yet thisgalaxy has a small RA g and r bands, respectively).Higher S´ersic indices and bulge-to-total-light ratios forboth TDE host galaxies and E+A/post-starburst galax-ies (see Figure 5) provide a natural explanation for theenhanced disruption rates in these galaxies. Stone &Metzger (2016) were the first to predict that the en-hanced rate in E+A/post-starburst galaxies might bedue to their large central stellar densities, as per-galaxyTDE rates scale roughly as ˙ N TDE ∝ ρ (cid:63) . We caution thatour galaxy S´ersic indices were derived using measure-ments that typically do not resolve the nuclear regionsof the galaxy and, as such, the density of the sphere ofinfluence of the BH cannot be directly constrained. Wenote, however, that Simard et al. (2011) isolate a sub-sample of ∼ n S´ersicfits to their bulges, and they find that galaxies with lowand high n bulge values also have low and high n g values.There is also some direct evidence that E+A galaxieshave higher central stellar densities. Stone & van Velzen(2016) find that the E+A galaxy NGC 3156 is centrallyoverdense, leading to an estimated TDE rate via two-body relaxation of ∼ − yr − , an order of magnitudehigher than for other galaxies with similar BH masses.Pracy et al. (2012) study a sample of seven local E+Agalaxies with IFU spectroscopy and find that they havecompact young cores and stellar population gradientsthat are predicted from models of mergers and tidal in-teractions that funnel gas into the galaxy core. Thissuggests that these galaxies are being seen in the latestage of a merger where the nuclei have already coa-lesced.Importantly—and separate from understanding theE+A galaxy preference—we are able to identify a pho-tometric criterion (light concentration, given by eitherthe S´ersic index or bulge fraction) that may predict aTDE overabundance more broadly than a spectroscopiccriterion (E+A classification). For upcoming transientsurveys where there may be up to ∼ transients dis-covered each year, a photometric host galaxy selectioncriterion could be extremely useful for focusing limitedfollow-up resources. For instance, choosing nuclear tran- idal Disruption Event Host Galaxies b a n d b a n d galaxy S´ersic index N ( no r m a li ze d ) log( M ∗ , tot /M fl ) < . band 1band 2sF16 galaxiesreference catalog Figure 8.
Left panel: total star formation rate vs. total stellar mass for our reference catalog and TDE host galaxies 1-8.Color corresponds to galaxy S´ersic index, ranging from 0.5 (blue) to 8 (red); for the reference catalog galaxies, this is the meanwithin each hexagonal bin. Right panel: normalized histograms of galaxy S´ersic index for our reference catalog (dashed black),galaxies between the SFMS and 0.5 dex below the SFMS (band 1; blue), galaxies between the SFMS − − M (cid:63), tot /M (cid:12) ) < . sients in high-S´ersic galaxies could significantly increasethe success of confirming TDEs.We thank Trevor Mendel for useful conversations, aswell as for performing bulge+disk decompositions ontwo TDE host galaxies (numbers 9 and 10) that are inthe Simard et al. (2011) catalog but not in the Mendelet al. (2014) catalog. We thank Jarle Brinchmann,Viraj Pandya, Peter Behroozi, Kevin Bundy, XavierProchaska, and Andrea Merloni for useful conversations. J.L.-S. and E.R.-R. are grateful for support from thePackard Foundation, NASA ATP grant NNX14AH37Gand NSF grant AST-1615881. S.L.E. acknowledges thereceipt of an NSERC Discovery Grant. The UCSC tran-sients group is supported in part by NSF grant AST-1518052 and from fellowships from the Alfred P. SloanFoundation and the David and Lucile Packard Founda-tion to R.J.F. Software: corner (Foreman-Mackey 2017), Mat-plotlib ( ? ), seaborn ( ? ).APPENDIX A. OTHER TDE HOST GALAXY PROPERTIESHere, we discuss a few other properties of the host galaxies of TDEs. The top-left panel of Figure 9 shows D n (4000)versus BH mass for TDE host galaxies and our reference catalog. D n (4000) corresponds to the strength of the4000˚A break and is an indicator of the age of the galaxy stellar population: no break means a young galaxy and astrong break means an old galaxy. Following Kauffmann et al. (2003b,c) and Brinchmann et al. (2004), the peak inD n (4000) at ∼ . r -band weighted mean stellar ages of ∼ ∼ n (4000) at ∼ .
85 corresponds to older elliptical galaxies with meanstellar ages of ∼
10 Gyr. The TDE host galaxies lie roughly in between these two peaks, with a peak in D n (4000)at ∼ .
5, indicating that they have mean stellar ages in between those of these two populations. Note that thismeasurement is only sensitive to the dominant stellar population and does not reveal multiple stellar populations. Inthe right panel, where we match on BH mass to TDE hosts 1-5, we see that sF16 galaxies exhibit a younger meanstellar age than wF16 galaxies. AGNs exhibit a younger mean stellar age than low-S/N AGNs. We note that D n (4000)is not as effective a metric as the H α EW and Lick H δ A metric (Figure 1) in isolating quiescent Balmer-strong galaxies.The middle-left panel of Figure 9 shows g -band galaxy absolute magnitude versus BH mass for TDE host galaxiesand our reference catalog. Results are similar for r band. Controlling for BH mass (right panel), TDE hosts are slightlyfainter than the catalog galaxies, and sF16 galaxies are slightly brighter than wF16 galaxies. The AGN and SF sampleshave very similar distributions. The bottom panel of Figure 9 shows bulge g -band absolute magnitude. Results aresimilar for r band. Here, controlling for BH mass, both TDE host galaxies and (to a somewhat greater extent) sF162 Law-Smith et al. log M bh /M fl D n ( ) T D E ho s t s c on t r o l E + A sF16wF16 D n (4000) AGN
AGNlow S/N AGN log M bh /M fl M g , ga l a xy T D E ho s t s c on t r o l E + A sF16wF16
22 21 20 19 18 17 M g, galaxy AGN
AGNSF log M bh /M fl M g , bu l g e T D E ho s t s c on t r o l E + A sF16wF16
21 20 19 18 17 16 M g, bulge AGN
AGNSF
Figure 9.
Left panels, top to bottom: D n (4000), g -band galaxy absolute magnitude, and g -band bulge absolute magnitudevs. BH mass for TDE host galaxies and our reference catalog. For galaxy and bulge magnitudes, the results are similar for the r -band. BH masses for TDE hosts 8, 9, and 10 are determined via M (cid:63), bulge . Contours are spaced by 0 . σ , with the darkestshading containing 0 . σ and the lightest shading containing 2 σ . Average errors in the TDE host galaxy measurements are shownin the top or bottom left. Right panels: 1D distributions of these properties in different subsamples, matched on BH mass ofTDE hosts 1-5. All 1D distributions are smoothed and normalized to equal area. Unsmoothed 1D histogram for TDE hosts 1-5is shown in solid red, and for TDE hosts 1-10 in dotted black. idal Disruption Event Host Galaxies log M bh /M fl g g2 d − r g2 d T D E ho s t s c on t r o l E + A sF16wF16 g g2d − r g2d AGN
AGNlow S/N AGN log M bh /M fl i n c li n a ti on T D E ho s t s c on t r o l E + A sF16wF16 AGN
AGNlow S/N AGN
Figure 10.
Same description as in Figure 9, but galaxy g − r in the top panel and inclination (face-on is 0 ◦ , maximum of 85 ◦ )in the bottom panel. galaxies have brighter bulge magnitudes than the reference catalog. Medians and spreads on the distributions of thegalaxy and bulge absolute magnitudes are given in Table 5.Lastly, we show galaxy g − r in the top panel of Figure 10, and galaxy inclination (face-on is 0 ◦ ) in the bottom panel.Controlling for BH mass, the galaxy g − r colors for TDE host galaxies and our reference catalog appear fairly similar.sF16 galaxies have bluer colors than wF16 galaxies. The inclinations of TDE hosts appear fairly uniform, with a hintof a preference for lower inclinations. Interestingly, the inclination distribution of sF16 galaxies appears different fromthat of the reference catalog, with a preference for 20 (cid:46) i (cid:46)
65 in contrast to the catalog’s preference for 40 (cid:46) i (cid:46) B. CORRELATIONSWe show correlations between many of the properties explored in this paper in Figure 11. We show BH mass, totalstellar mass, redshift, half-light surface brightness, bulge g − r , galaxy S´ersic index, and bulge-to-total-light ratio. Thetotal stellar mass behaves very similarly to BH mass in this diagram, which, in addition to it being a more physicallyrelevant parameter for tidal disruptions, is why we use BH mass in our analysis. However, as mentioned in the text,if we simply replace BH mass with M (cid:63), tot in our analysis, our conclusions remain the same.4 Law-Smith et al. . . . l og M ∗ / M fl . . . . l og S F R . . . . z Σ h l , g . . . bu l g e g − r n g log M bh /M fl . . . . ( B / T ) g . . . log M ∗ /M fl . . . . log SFR .
05 0 .
10 0 .
15 0 . z Σ hl , g . . . bulge g − r n g . . . . ( B/T ) g Figure 11.
Correlations between many of the properties explored in this paper for our reference catalog of ∼ M (cid:12) yr − ),redshift, g -band half-light surface brightness (in mag/arcsec ), bulge g − r , galaxy S´ersic index ( n g ), and g -band bulge-to-total-light ratio ( B/T ) g . Each panel contains 95% of the points. REFERENCES