Circumstellar Dust, PAHs, and Stellar Populations in Early-Type Galaxies: Insights from GALEX and WISE
aa r X i v : . [ a s t r o - ph . GA ] N ov MNRAS , 000–000 (0000) Preprint 13 August 2018 Compiled using MNRAS L A TEX style file v3.0
Circumstellar Dust, PAHs, and Stellar Populations inEarly-Type Galaxies: Insights from
GALEX and
WISE
Gregory V. Simonian ⋆ and Paul Martini , Department of Astronomy, The Ohio State University, 140 West 18th Avenue, Columbus, OH 43210, USA Center for Cosmology and AstroParticle Physics (CCAPP), The Ohio State University, 191 W. Woodruff Ave., Columbus, OH 43210, USA
13 August 2018
ABSTRACT
A majority of early-type galaxies contain interstellar dust, yet the origin of this dust,and why the dust sometimes exhibits unusual polycyclic aromatic hydrocarbon (PAH)ratios, remains a mystery. If the dust is internally produced, it likely originates fromthe large number of asymptotic giant branch stars associated with the old stellarpopulation. We present
GALEX and
WISE elliptical aperture photometry of ∼ early-type galaxies with Spitzer mid-infrared spectroscopy and/or ancillary data fromATLAS , to characterize their circumstellar dust and the shape of the radiation fieldthat illuminates the interstellar PAHs. We find that circumstellar dust is ubiquitousin early-type galaxies, which indicates some tension between stellar population ageestimates and models for circumstellar dust production in very old stellar popula-tions. We also use dynamical masses from ATLAS to show that WISE
W1 (3.4 µ m)mass-to-light ratios are consistent with the initial mass function variation found byprevious work. While the stellar population differences in early-type galaxies corre-spond to a range of radiation field shapes incident upon the diffuse dust, the ratio of theionization-sensitive . µ m to . µ m PAH feature does not correlate with the shape ofthe radiation field, nor to variations with the size-sensitive . µ m to µ m ratio. The . µ m to . µ m PAH ratio does tend to be smaller in galaxies with proportionallygreater H emission, which is evidence that processing of primarily smaller grains byshocks is responsible for the unusual ratios, rather than substantial differences in theoverall PAH size or ionization distribution. Key words: galaxies: elliptical and lenticular, cD – infrared: galaxies – ultraviolet:galaxies – galaxies: stellar content
While early-type galaxies (ETGs) were historically as-sociated with uniformly old stellar populations and nocold interstellar medium (ISM), forty years of multiwave-length observations have demonstrated that view is toosimplistic. Instead, many ETGs have a complex, multi-phase ISM, often with a mixture of cold (Knapp et al.1985; Wardle & Knapp 1986), warm (Caldwell 1984;Phillips et al. 1986; Sadler 1987), and hot (Forman et al.1985; Canizares et al. 1987) gas. Improved angular resolu-tion has also led to the detection of dust lanes in many ETGs(Sadler & Gerhard 1985; Sparks et al. 1985; Ebneter et al.1988; Veron-Cetty & Veron 1988; Goudfrooij et al. 1994),and observations in the far-infrared (FIR) indicated thatmany ETGs contain a diffuse, cold dust component(Jura et al. 1987; Knapp et al. 1989; Goudfrooij & de Jong ⋆ E-mail: [email protected]
Infrared Space Observatory ( ISO ) satel-lite found flux in excess of expectations for the stellarpopulation of many ETGs, which was postulated to arisefrom either Polycyclic Aromatic Hydrocarbon (PAH) orVery Small Grain (VSG) emission (Madden et al. 1999;Ferrari et al. 2002; Xilouris et al. 2004; Pahre et al. 2004).The existence of these small grains was surprising because oftheir short lifetimes in hot plasma (Draine & Salpeter 1979;Dwek & Arendt 1992), and so the origin of these grains ishotly debated.MIR spectroscopy of individual PAH features in ETGsindicated that the relative strengths of short-wavelengthand long-wavelength PAH features were often vastly re-duced compared to the same features in star-forming galax-ies (Kaneda et al. 2005). Proposed physical conditions whichcould lead to the relatively weaker short-wavelength featuresinclude a grain population dominated by neutral rather than c (cid:13) Simonian & Martini ionized PAHs, and a larger grain size distribution comparedto star-forming galaxies (Draine & Li 2007).An explanation for both the existence of PAHs andtheir anomalous line ratios in ETGs has proven elusive.Kaneda et al. (2008) found that PAH emission is uncorre-lated with stellar emission, and suggested that larger neutralPAHs were externally accreted through mergers. Alterna-tively, Vega et al. (2010) posited that PAHs are producedby carbon stars formed in a minor star-forming event, andthat shocks with the ambient ISM then preferentially de-stroyed the small grains.For galaxies without interstellar dust, we get a di-rect view of the MIR emission of the stellar population,which makes these galaxies well-suited to test Stellar Pop-ulation Synthesis (SPS) models. They are representativeof old stellar populations with little ongoing star forma-tion (Yi et al. 2005; Kaviraj et al. 2007; Temi et al. 2009;Shapiro et al. 2010), and negligible extinction due to dust.We will use them to test stellar models for the evolved gi-ant stars that dominate their light, especially the substan-tial progress on the Thermally Pulsating Asymptotic GiantBranch (TP-AGB) phase over the last decade (Marigo et al.2008; Girardi et al. 2010; Cassarà et al. 2013; Marigo et al.2013; Rosenfield et al. 2014). This phase is critical as thestars can contribute a significant amount of the integratedflux in the infrared (Maraston 2005; Kelson & Holden 2010;Melbourne et al. 2012; Conroy 2013; Melbourne & Boyer2013).Previous studies of AGB stars have shown that circum-stellar dust is necessary to adequately describe their spectrain the MIR (Bedijn 1987; van Loon et al. 1999; Trams et al.1999). However, the inclusion of circumstellar dust withinSPS models has been difficult, and only a few modelsincorporate dusty AGB stars into their spectral libraries(Bressan et al. 1998; Silva et al. 1998; Marigo et al. 2008;Villaume et al. 2015). The expansion of MIR observationsprovide new opportunities to compare these circumstellardust models to real stellar populations (e.g. Norris et al.2014; Villaume et al. 2015).It is difficult to compare circumstellar dust models todata in active star-forming galaxies because the circumstel-lar dust emission is often dwarfed by emission from dust inthe diffuse ISM. For our sample of passive early-type galax-ies with much less interstellar dust, the circumstellar com-ponent can be observed in the MIR (beyond about 10 µ m).Excess flux associated with circumstellar dust has been iden-tified in a number of studies of galaxies without evidencefor interstellar dust (Bressan et al. 1998; Athey et al. 2002;Martini et al. 2013).The MIR region from 3–5 µ m dominated by photo-spheric emission is valuable for stellar mass measurements.Since low-mass stars contain most of the stellar mass of agalaxy, observations at these wavelengths are more robust tovariations in metallicity, star formation history (SFH), andstar formation rate (SFR; Meidt et al. 2014). Results fromSPS models have traditionally been the only way to mea-sure the stellar masses of large numbers of galaxies. Alter-native mass estimates were recently released by ATLAS for a volume-limited sample of 260 ETGs closer than 42Mpc with M ∗ > ∼ × M ⊙ . Cappellari et al. (2013) deriveddynamical masses for these galaxies using r -band photome-try, SAURON integral-field unit (IFU) spectroscopy, and dy- namical models based on the Jeans equations. They then ob-tained stellar masses by subtracting a Navarro-Frenk-White(NFW) halo; which yielded stellar masses with assumptionsindependent from SPS models.We use MIR data from the WISE satellite, which ob-served the entire sky in four MIR bands: W1 (3.4 µ m), W2(4.6 µ m), W3 (12 µ m), and W4 (22 µ m) (see Wright et al.(2010) for further details). The first two bands are similarto the Spitzer
Infrared Array Camera (IRAC) [3.6] and [4.5]bands, and the W4 band is similar to the Multiband Imag-ing Photometer on
Spitzer (MIPS) 24 µ m band. In typicalgalaxies, the W1 and W2 bands are expected to trace theevolved stellar population; the W3 band will contain signif-icant PAH features; and the W4 band will be dominated bycontinuum emission from hot dust grains (e.g. Jarrett et al.2013). Due to the all-sky coverage of WISE , all sufficientlybright ETGs can be studied in the MIR, a substantial in-crease over previous targeted surveys. This wide coveragewill yield valuable demographic data about the stellar pop-ulations, circumstellar dust, and PAHs in ETGs, as well asidentify promising targets for future study with targetedmissions.We also include UV photometry from the
GALEX satellite to measure the shape of the radiation field inci-dent on any PAHs that may be present.
GALEX observed63% of the sky to a depth of at least m AB = 20 mag in theFUV ( Å) and NUV (
Å) bands, with a resolutionof about 4.25 ′′ and 5.25 ′′ respectively (see Morrissey et al.(2005) and Martin et al. (2005) for further details).In this paper, we distinguish between “interstellar” and“circumstellar” dust as follows: “circumstellar” dust resideswithin the stellar winds of AGB stars while “interstellar”dust resides within the diffuse ISM. Circumstellar dust is in-trinsically connected to the stellar population, which makesit a valuable extension to Stellar Population Synthesis mod-els. Meanwhile, interstellar dust is often uncorrelated withthe stellar population and often dominates the IR emissionwhen it is present.Our paper is organized as follows: the next section ofthis paper contains a description of our samples of ETGs.Section 3 describes how we performed aperture photome-try on both WISE and
GALEX images and how we distin-guish between galaxies with and without diffuse dust. In Sec-tion 4, we calculate stellar mass-to-light ratios for W1 fromATLAS dynamical masses and compare them to mass-to-light ratios predicted by SPS models. We also comparemodels of circumstellar dust to our data, and use these tojointly constrain stellar ages and the masses of stars thatproduce circumstellar dust, as well as investigate the extentto which circumstellar dust can contaminate measurementsof the SFR as measured from MIR indicators. Section 5combines GALEX and
WISE photometry with
Spitzer
In-frared Spectrograph (
Spitzer -IRS) spectroscopy from previ-ous works to investigate the properties of PAHs and their en-vironments in ETGs. We also comment on the use of
WISE photometry to determine the MIR properties of ETGs. Wesummarize our results in Section 6.
MNRAS , 000–000 (0000)
TGs in GALEX and WISE We use MIR and UV photometry to study circumstellarand interstellar dust in ETGs. Our sample is drawn fromtwo recent, comprehensive studies of ETGs: the characteri-zations of interstellar dust in ETGs with
Spitzer -IRS spec-tra (Rampazzo et al. 2013), and the stellar population anddynamical study from the ATLAS survey. Spitzer -IRS Atlas
The Revised Shapley-Ames (RSA) catalog is a canon-ical collection of bright, well-studied, nearby galaxies.Rampazzo et al. (2013) constructed their RSA
Spitzer -IRSatlas by cross-matching the E–S0 galaxies from the RSAcatalog with
Spitzer -IRS observations available in the
Spitzer
Heritage Archive (SHA). This sample consists of91 ETGs, including 56 E-type, 27 S0-type, and 8 mixedE/S0+S0/E-type galaxies; their properties are given in Ta-ble 1. Rampazzo et al. (2013) uniformly reprocessed and an-alyzed all of these spectra and measured line intensities foreach of the detected emission lines.Rampazzo et al. (2013) classified galaxies according tothe Panuzzo et al. (2011) classification scheme for MIR spec-tra. A summary of the classification scheme is as follows:Class-0 galaxies are completely passive, that is apart from afew broad circumstellar dust features (Bressan et al. 1998),they have no emission lines in the MIR. These galaxies havespectra consistent with only an old stellar population. Class-1 galaxies have emission features in their spectra, but noPAH features. Class-2 and Class-3 galaxies are those withanomalous and normal PAH features, respectively; they willbe the primary focus of this work. Finally, Class-4 galaxieshave a hot dust continuum. Sample
The ATLAS project surveyed a volume-limited samplewithin 42 Mpc of morphologically-selected early-type galax-ies with M K < − . mag (Cappellari et al. 2011). Thissample contains 260 ETGs: 68 E galaxies and 192 S0 galaxies(see Table 1). The extensive data available for the ATLAS sample includes Sloan Digital Sky Survey (SDSS) ugriz pho-tometry (Abazajian et al. 2009; Scott et al. 2013), observa-tions with the SAURON IFU spectrograph (Cappellari et al.2011), 21 cm emission observations (Serra et al. 2012), and CO J=1–0 and J=2–1 observations (Alatalo et al. 2013).These data are available from the ATLAS website .The data collected for the ATLAS sample, combinedwith extensive dynamical and stellar population modeling,has resulted in a wealth of valuable measurements. Somerelevant observational results include the presence of opti-cal dust features (Krajnović et al. 2011), surface brightnessprofiles (Scott et al. 2013), and luminosities in the r-band(Cappellari et al. 2013). We use their stellar population pa-rameters derived from both SSP models and reconstructedSFH models (McDermid et al. 2015) to evaluate circumstel-lar dust models.Finally, the dynamical analysis and modeling by Cappellari et al. (2013) include dynamical mass-to-light ra-tios derived by fitting model h v los i / to V RMS measure-ments. These were derived from Jeans Anisotropic Multi-gaussian expansion (JAM) modeling (Cappellari 2008). Theoutput from these fits include stellar mass-to-light ratios de-rived by simultaneously fitting an NFW halo (Navarro et al.1996) and a separate stellar distribution constrained by theobserved surface brightness profile (Cappellari et al. 2013).
We chose to perform all aperture photometry with the stan-dard aperture used by the Two Micron All-Sky Survey(2MASS) Extended Source Catalog (XSC) : the K s =20 magarcsec isophote (hereafter K20). This isophote correspondsroughly to σ of the typical background noise in the K s images. Despite the fact that the K20 isophote underrepre-sents the “total” flux by ∼ % , it provides the most re-producible measurement of a galaxy’s flux . There are also2MASS JHK s measurements of the entire sample in thissame aperture. The sizes of apertures for selected galaxiesare shown in Fig. 1.Due to PSF differences between the 2MASS and WISE surveys, the effective shape of the aperture needs to be cor-rected in order for the isophotes in the different beams tomatch . Because the WISE
W4 beam is larger and morecircular than the beams for the other three
WISE bands,the W4 aperture is a different size and shape from the rest.For all but three objects, this adjustment was performedby the
WISE pipeline, which generates a corrected apertureany time a 2MASS XSC object is centered within 2 ′′ of the WISE source. For the three objects whose centers in thetwo surveys are separated by more than 2 ′′ , we generated amatched aperture manually. We calculated these aperturesby binning the ATLAS sample by 2MASS axis ratio, andinterpolated the adjusted WISE semimajor axis and axisratio to the object.
GALEX apertures were not adjustedbecause the PSF differences are less important. The pho-tometry parameters for all of the galaxies in our sample aregiven in Table 1.
WISE
We measured aperture photometry on the
WISE
Atlas im-ages. These are coadded images available as high-level dataproducts. The details of the image construction and calibra-tion are described in the
WISE explanatory supplement .There are presently two recommended data releases:All-Sky and AllWISE. The AllWISE release is more preciseand sensitive because it incorporates observations from thepost-cryo mission into the standard mission observations,among other improvements. However, inclusion of post-cryoobservations decreases the saturation limit of the images,and hence decreases accuracy for bright objects. We there-fore used All-Sky images for objects with saturated pixels, http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/sec4_4c.html http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/index.html MNRAS000
WISE explanatory supplement .There are presently two recommended data releases:All-Sky and AllWISE. The AllWISE release is more preciseand sensitive because it incorporates observations from thepost-cryo mission into the standard mission observations,among other improvements. However, inclusion of post-cryoobservations decreases the saturation limit of the images,and hence decreases accuracy for bright objects. We there-fore used All-Sky images for objects with saturated pixels, http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/sec4_4c.html http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/index.html MNRAS000 , 000–000 (0000)
Simonian & Martini
Table 1.
Aperture photometry parameters for galaxies in the ATLAS and Rampazzo et al. (2013) samplesGalaxy Morph D S a W a W b/a W b/a W PA Survey NUV Tile FUV Tile(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)IC0560 S0-a 27.2 A 31.1 35.23 0.47 0.56 10 AllWISE MISDR1_24342_0266 MISDR1_24342_0266IC0598 S0-a 35.3 A 31.4 37.39 0.43 0.55 10 AllWISE – –IC0676 S0-a 24.6 A 37.9 40.93 0.62 0.62 -20 AllWISE GI4_042019_J111315p093030 GI4_042019_J111315p093030IC0719 S0 29.4 A 42.73 48.45 0.35 0.43 50 AllWISE AIS_232_sg80 AIS_232_sg80IC0782 SBab 36.3 A 18.53 22.42 0.65 0.75 60 AllWISE GI6_001033_GUVICS033 GI3_079021_NGC4261IC1024 S0 24.2 A 33.1 38.67 0.4 0.51 30 AllWISE MISDR1_33738_0535 MISDR1_33738_0535IC1459 E1 29.0 R 97.03 100.97 0.84 0.86 40 AllWISE GI1_093001_IC1459 GI1_093001_IC1459IC2006 E1 20.0 R 46.86 48.75 0.88 0.89 50 AllWISE AIS_423_sg76 AIS_423_sg76IC3370 E2 pec 27.0 R 66.94 69.65 0.85 0.87 55 AllWISE – –IC3631 S0-a 42.0 A 24.4 28.31 0.6 0.7 90 AllWISE GI6_001074_GUVICS074 AIS_223_sg22
Notes:
Galaxy information and aperture photometry parameters. Columns: (1) galaxy name; (2) morphological type; (3) distance inMpc; (4) Originating Sample (A: ATLAS R: Rampazzo et al. (2013) B: Both); (5) W1 aperture semimajor axis in arcseconds;(6) W4 aperture semimajor axis in arcseconds; (7) W1 aperture axis ratio; (8) W4 aperture axis ratio; (9) Position Angle (10)
WISE data release catalog; (11)
GALEX
NUV tilename; (12)
GALEX
FUV tilename. References for morphological type and distance comefrom Cappellari et al. (2011) and Rampazzo et al. (2013). The full version of this table with the full galaxy sample is included in theonline journal article. and otherwise used AllWISE. The data release of each imageis included in Table 1.We measured elliptical aperture photometry with the ellipse package in IRAF using the adjusted K20 parametersgiven in Table 1. Sky values were estimated via the fitsky package in IRAF. The inner radius of the annulus was cho-sen to be 1.5 times the semimajor axis of the photometricaperture in order to exclude galaxy flux. The thickness ofthe annulus was chosen to be 30 pixels in order to providesufficient sky pixels to adequately measure the background.We used SExtractor (Bertin & Arnouts 1996) to detectand mask foreground sources in the
WISE
W1 Atlas images,and then applied the masks to all four
WISE bands. Themost reliable method of foreground removal is PSF subtrac-tion, as done in Jarrett et al. (2013). However, this methodis not feasible for our large sample of galaxies due to thespatially variable PSF in the Atlas images. Since the stel-lar emission in ETGs is morphologically smooth, foregroundstars bright enough to significantly affect the galaxy’s fluxare easily identified by SExtractor.We also analyzed the 17 galaxies (including 3 ellipticals)from Jarrett et al. (2013) in order to test our pipeline. Weexcluded the M51 pair because none of our targets requiresimilarly complex deblending. After comparing the flux mea-sured from our pipeline to those reported by Jarrett et al.(2013), we encounter RMS differences of W1: 0.05 mag,W2: 0.06 mag, W3: 0.12 mag, and W4: 0.07 mag; we adoptthese values as estimates of our photometric uncertainties.Since the Jarrett et al. (2013) sample is more morphologi-cally complex than our measurements, we expect these dif-ferences are upper limits to the true photometric uncertain-ties. These differences cannot be explained by color correc-tions, which only correspond to 1% differences. Our mea-surements for the full sample are provided in Table 2, andhave not been corrected for extinction. The uncertaintiesin Table 2 are formal uncertainties, and include zero-pointuncertainties of 0.006, 0.007, 0.015, and 0.012 mag for W1,W2, W3, and W4, respectively. For our analysis, we applyextinction corrections for 2MASS and
WISE bands fromIndebetouw et al. (2005).
GALEX
The
GALEX
GR6/7 data release has images from six differ-ent observing programs with varying breadths and depths.The deepest is the targeted GII program, followed by severalscience surveys. The surveys vary in both sky coverage andexposure time, reaching 29,000 s for fields with well-studiedgalaxies, down to ∼ s for the shallowest survey, whichcovers 63% of the sky. For galaxies observed in multiple sur-veys, we chose the highest exposure-time image which con-tained the entire photometric aperture in the field-of-view.Although the PSF is distorted near the edge of the field(Morrissey et al. 2007), we found that this did not affectthe photometry. The tilename for each galaxy identifies theexposure and is given in Table 1.We similarly used ellipse in IRAF to measure the GALEX images, although we estimated the sky differently.While the
GALEX pipeline provides background images,we noticed galaxy flux in the background images for someobjects. We therefore opted for traditional background es-timation from an annulus. Unfortunately, fitsky in IRAFdoes not perform well when the background counts arevery small, as it expects sky values to be normally dis-tributed. We therefore implemented the method describedin Gil de Paz et al. (2007), which divides two elliptical an-nuli into radial segments and averages over those segments.Similar to Gil de Paz et al. (2007), we divided the annuliinto 90 segments total, with a typical segment area of 4000pixels. We set the semimajor axis of the inner annulus at1.5 times the semimajor axis of the photometric aperture.The uncertainty of the background comes from the standarddeviation of the segment mean values.We tested our approach with 20 morphologically diversegalaxies from the Gil de Paz et al. (2007) atlas which over-lapped our sample, and used the same D25 aperture. De-spite using the same technique, we systematically measuredless flux than Gil de Paz et al. (2007) by 0.1 to 0.4 mag.We did successfully reproduce the Gil de Paz et al. (2007)results with the original cutouts from the NASA/IPAC Ex-
MNRAS , 000–000 (0000)
TGs in GALEX and WISE Table 2.
Galaxy magnitudes for the ATLAS and Rampazzo et al. (2013) samples.Galaxy FUV σ F UV
NUV σ NUV W1 σ W W2 σ W W3 σ W W4 σ W (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)IC0560 21.49 0.117 19.40 0.035 13.27 0.006 13.93 0.009 14.35 0.056 13.67 0.154IC0598 – – – – 12.93 0.006 13.59 0.008 14.59 0.062 14.35 0.296IC0676 18.11 0.054 16.92 0.031 12.46 0.006 13.01 0.008 11.06 0.016 9.681 0.013IC0719 17.29 0.072 16.68 0.037 12.34 0.006 12.96 0.008 12.07 0.017 11.99 0.045IC0782 21.57 0.108 19.21 0.035 13.93 0.007 14.62 0.012 15.78 0.142 > > Notes:
Calculated magnitudes for galaxies in our sample reported in the AB system. They have not been corrected for extinction.Columns: (1) Galaxy name; (2) FUV magnitude; (3) FUV magnitude uncertainty; (4) NUV magnitude; (5) NUV magnitudeuncertainty; (6) W1 magnitude; (7) W1 magnitude uncertainty; (8) W2 magnitude; (9) W2 magnitude uncertainty; (10) W3magnitude; (11) W3 magnitude uncertainty; (12) W4 magnitude; (13) W4 magnitude uncertainty. Uncertainties contain statistical andcalibration uncertainties. All upper limits are σ upper limits. Blank UV entries mean that no GALEX observations were taken. Thefull version of this table with the full galaxy sample is included in the online journal article tragalactic Database (NED) , so we conclude that the dis-crepancy is due to changes in the GALEX pipeline. We alsocompare our measurements with Bai et al. (2015), who re-measured the Gil de Paz et al. (2007) atlas using data pro-cessed by the current GR6/7 pipeline. Our results agree withBai et al. (2015) to an RMS difference of 0.14 mag withouta systematic trend in both the NUV and FUV bands, whichwe attribute to different background estimation methods.We conclude that our UV measurements are accurate andhave an RMS precision of 0.14 mag.Our
GALEX measurements are also in Table 2. Theyhave not been corrected for extinction, although for all of ouranalysis we used the prescription in Gil de Paz et al. (2007)to apply extinction corrections to these measurements. Val-ues for E(B-V) are from the Schlegel et al. (1998) maps ob-tained from the IRSA dust map service.
Fig. 2 shows a color-absolute magnitude diagram that illus-trates the diversity in the shape of the SED for the ATLAS and Rampazzo et al. (2013) galaxies. The majority of ETGsare extremely deficient in UV photons, as is expected fromtheir generally old stellar populations. There appears tobe a trend in the SED, where more UV-rich galaxies tendto be less luminous. This trend appears unrelated to theMIR classes of Rampazzo et al. (2013), although the RSA Spitzer -IRS atlas contains few low-luminosity ETGs. Theregion with blue ETGs is also sparely populated comparedto the area with NUV-W1 > . The Class-4 object which isan outlier at the bottom-right is NGC 1275, a Seyfert 1.5galaxy in the Rampazzo et al. (2013) sample. There doesnot appear to be foreground contamination for this galaxy,so we believe the extreme NUV-W1 color and luminosity aredue to the AGN component.As indicated by Rampazzo et al. (2013), about half ofETGs contain observable traces of interstellar dust, whichcan dominate the MIR signal. We therefore attempted to https://ned.ipac.caltech.edu separate the passive from the non-passive galaxies. For theRampazzo et al. (2013) sample, non-passive galaxies areclassified from Spitzer -IRS spectra as Class-1–4 fairly re-liably. The ATLAS sample does not have MIR spectra, sowe attempt to distinguish between passive and non-passivegalaxies with the extensive ancillary data.Martini et al. (2013) demonstrated a one-to-one cor-respondence between optical dust lanes observed with the Hubble Space Telescope (HST) and emission from cold dustdetected by
Spitzer
MIPS. We attempted to remove obviouscontaminants with observed dust features in r-band obser-vations (Krajnović et al. 2011). However, since the photo-metric resolution of SDSS is significantly lower than HST,we also searched for false negatives in our sample by cross-matching the dust detections in Martini et al. (2013) tothe non-detections in Krajnović et al. (2011). This revealedmany cases where the ATLAS images did not reveal dustlanes that were clearly visible with HST and via FIR de-tections, so we also excluded galaxies with CO detections(Young et al. 2011).Fig. 3 shows the W1-W3 vs. W1-W4 colors for galax-ies without evidence for diffuse, interstellar dust. We alsocompare this sample to SPS tracks with metallicities thatbracket the ATLAS sample. We used SPS models with anexponential SFH timescale of 100 Myr. We also included aminimum sSFR of − yr − to account for constant, verylow levels of star formation (Ford & Bregman 2013). Fig. 3shows that the Flexible Stellar Population Synthesis (FSPS)high and median metallicity model tracks follow the shapeof the data quite well, albeit with an offset. We explore thesource of this offset in Section 4.2.When compared to the passive Class-0 galaxies fromRampazzo et al. (2013), there still appears to be a tail ex-tending redward of the clump of Class-0 galaxies. The sin-gle Class-0 object in the tail is NGC 4377, which has po-tential foreground contamination. Therefore, we performedan external check on this sample by cross-matching with Herschel detections at FIR wavelengths (Smith et al. 2012;di Serego Alighieri et al. 2013; Amblard et al. 2014). For ob-jects in Amblard et al. (2014) we defined a “dust detection”
MNRAS000
MNRAS000 , 000–000 (0000)
Simonian & Martini
NGC6278NGC4570NGC4660PGC051753IC3631
Figure 1.
Cutouts of selected galaxies in (Left to right:) GALEX
FUV and NUV, and
WISE
W1, W2, W3, and W4. The white regionsare masked pixels, the photometric aperture is in red, and the sky annulus is in blue. Galaxies are shown in order of decreasing ATLAS stellar mass from top to bottom. Scale is preserved from left to right, and a 30 ′′ scalebar is shown in the bottom right of each FUVimage. IC 3631 is not detected in W4. PGC 051753 is not detected in either FUV or W4. as a 5 σ detection in at least one of the 250 µ m, 350 µ m,and 500 µ m bands. For galaxies from the other two stud-ies, we used their internal criteria to indicate a dust detec-tion. We note that two of the Class-0 objects had Herschel
FIR detections, along with 31% of the overlapping ATLAS sample, which suggests there is still substantial contamina-tion. The region of the color-color diagram not populatedby the galaxies with FIR detections is W − W < . and W − W < . (AB: W − W < − . and W − W < − . ). NGC 4486A, a tidally disrupted satel- lite of M87, is the only galaxy in this region with a marginal Herschel detection at 250 µ m. Because of its tidal interac-tions, we classify it as a peculiar case, and assume that therest of the galaxies in this region do not contain interestellardust. We denote the galaxies in this region as the “color-cut dustless” sample, compared to the subset with only COand/or dust exclusions, which we just term “dustless”. MNRAS , 000–000 (0000)
TGs in GALEX and WISE −24−23−22−21−20−19−18 M W1 (AB) N U V - W ( A B ) ATLAS3DClass 0Class 1Class 2Class 3Class 4
Figure 2.
NUV-W1 vs W1 luminosity for the Rampazzo et al.(2013) and ATLAS galaxies. These quantities are reasonableproxies for the shape of the radiation field and stellar mass,respectively. There is no clear trend between Rampazzo et al.(2013) class and NUV-W1, even for the Class-2 galaxies withanomalous PAH ratios. The luminous, blue Class-4 object is NGC1275, which is a Seyfert 1.5 and thus may not be representativeof ETGs. −2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.5 W1-W3 (AB) −4.0−3.5−3.0−2.5−2.0−1.5−1.0−0.50.00.5 W - W ( A B ) [Z/H] = 0.2[Z/H] = -0.2[Z/H] = -0.89ATLAS3D (FIR)ATLAS3DClass 0 W1-W3 (V ga) W - W ( V g a ) Figure 3.
W1-W3 vs. W1-W4 colors of ATLAS galaxies with-out CO or dust detections. Circles are data and triangles areupper limits due to W4 non-detections. The dashed black linesrepresent the “color-cut” criteria. Magenta and green points rep-resent galaxies inside and outside the “color-cut” dustless re-gion, respectively. Also shown is the passive Class-0 sample fromRampazzo et al. (2013). These data are compared to circumstel-lar tracks from Flexible Stellar Population Synthesis models. Thered track corresponds to the highest metallicity galaxy in theATLAS sample with [Z/H]=0.2. The black track correspondsto the median metallicity in the ATLAS sample of [Z/H]=-0.2. The blue track corresponds to the lowest metallicity galaxywith [Z/H]=-0.89. The beginning of the track at 1 Gyr is markedwith a star, and continues until 14 Gyr. Solid tracks represent theFSPS model with the circumstellar dust model of Villaume et al.(2015); the very short dotted tracks in the lower left corner arethe same models without circumstellar dust. -1 Wavelength (µm) -3 -2 -1 N o r m a li z e d ν f ν Figure 4.
Spectral Energy Distributions of galaxies withinthe K20 aperture from broadband FUV, NUV, J, H, Ks, W1,W2, W3, W4 photometry, normalized to the K s band. Modelspectra are high-metallicity FSPS models (Conroy et al. 2009;Conroy & Gunn 2010) with exponentially decreasing SFH with τ SF H = 100
Myr and [Z/H] = 0.2. The different spectra rep-resent t=0.1, 1, 5, and 10 Gyr ages. Data points are the com-bined ATLAS and Rampazzo et al. (2013) galaxies. ATLAS galaxies without CO detections and dust lane morphology,or Rampazzo et al. (2013) Class-0 galaxies make up the pas-sive group, while those with either detection, or classified asRampazzo et al. (2013) Class-1–4, are in the non-passive group. We used the
GALEX , 2MASS, and
WISE photometry toconstruct broadband SEDs for our sample. These SEDs areshown in Fig. 4, normalized to the K s band. NGC 2974 isa very large outlier in W1 and W2, most likely due to verysevere foreground contamination from a nearby bright star.While its photometry is reported in Table 2, we exclude itfrom our plots due to its suspect color.We compare our data to the FSPS v2.5 models(Conroy et al. 2009; Conroy & Gunn 2010), which recentlyincorporated circumstellar dust (Villaume et al. 2015).These models allow for very extensive customization ofisochrones and stellar atmosphere libraries. The FSPS mod-els match the data well, particularly in the UV. In W3 andW4 many of the points extend above the model predictions.However, galaxies in the dustless sample described in Sec-tion 3.3 fall well within the range of colors predicted by themodels. One valuable use for SPS models has been to predict stel-lar mass-to-light ratios (M/L) for galaxies from broadbandcolors (Bell & de Jong 2001; Bell et al. 2003; Zibetti et al.2009). Bell & de Jong (2001) illustrated the power of thisapproach with their relation between
M/L B and B − R color, which was largely robust to metallicity, extinction, andbursty star formation history (SFH). More recent improve-ments to M/L determinations include the use of multiplecolors and models of dust extinction (Zibetti et al. 2009). MNRAS000
M/L B and B − R color, which was largely robust to metallicity, extinction, andbursty star formation history (SFH). More recent improve-ments to M/L determinations include the use of multiplecolors and models of dust extinction (Zibetti et al. 2009). MNRAS000 , 000–000 (0000)
Simonian & Martini
Yet despite these improvements, the presence of young stel-lar populations and internal extinction remains a majorsource of uncertainty for these models (Bell & de Jong 2001;MacArthur et al. 2004; Zibetti et al. 2009). These uncer-tainties are less severe at longer wavelengths, where theeffects of young stellar populations and extinction on themass-to-light ratio are diminished (Meidt et al. 2014). How-ever, this reduction in uncertainty comes at a cost of in-creased uncertainties in modeling the Asymptotic GiantBranch (AGB) phase in the stellar models, which are es-pecially poorly studied at low metallicity (Maraston et al.2006). SPS models predict that the impact of AGBstars peak around 1 Gyr and decrease at greater ages(Melbourne et al. 2012), so should be less important formost ETGs.Many different prescriptions exist to determine stellarmasses of galaxies (e.g. Meidt et al. 2014; Cluver et al. 2014)that are either directly based on or calibrated from SPSmodels (Bruzual & Charlot 2003; Salim et al. 2007). How-ever, the accuracy of these prescriptions is limited by theuncertainties associated with models for evolved stars, thedust geometry, IMF, and SFH.Stellar mass-to-light ratios determined by dynamicalmeans from ATLAS are extremely valuable because theyare not dependent on the same assumptions made by SPSmodels, such as an IMF. Results from ATLAS indi-cate a tension between stellar masses generated from SPSmodels and from dynamics, which is correlated with ob-served velocity dispersions. This disagreement has beenput forward as evidence that the IMF may be vari-able (van Dokkum & Conroy 2012; Cappellari et al. 2012),which would imply greater and systematic uncertainties inSPS-derived stellar masses that assume a universal IMF.In order to determine if the tension is also seen in theMIR, we transformed the ATLAS mass-to-light ratio fromr-band to W1, and corrected for the difference between thelargest Multi-Gaussian Expansion (MGE) aperture used inATLAS and the modified K20 aperture used here. Theband was transformed according to the formula: log (cid:18) ML W (cid:19) K = log (cid:18) ML r (cid:19) MGE + log (cid:18) L r L r, ⊙ (cid:19) + 0 . (cid:18) m W | MGE − (cid:18) d pc (cid:19) − M W , ⊙ (cid:19) (1)where ( M/L W ) K is the mass-to-light ratio within the K20aperture in W1, ( M/L r ) MGE is the ATLAS mass-to-lightratio within the region where the surface brightness is accu-rately modeled by the MGE, L r /L r, ⊙ is the r -band luminos-ity of the galaxy in units of the in-band solar luminosity with M r, ⊙ = 4 . in AB magnitudes (Blanton & Roweis 2007), m W is the AB magnitude of the galaxy as measured in thiswork, d is the distance to the galaxy, and M W , ⊙ = 5 . mag is the absolute magnitude of the Sun in AB magnitudes(transformed from the Vega value in Jarrett et al. (2013)).This transformation assumes that the stellar mass-to-lightratio is spatially constant, which is consistent with assump-tions made in JAM modeling (Cappellari et al. 2013). Inorder to find an aperture representative of the total fluxof the galaxy measured by ATLAS , we numerically inte-grated the MGE intensity model in an elliptical aperturesuch that at least 90% of the flux would be in that aperture.This cutoff was chosen because the accuracy of the inte- grated flux of the MGE models compared to SDSS fluxesis 10% (Scott et al. 2013). The transformed mass-to-lightratios are shown in Fig. 5. The uncertainty of the mass-to-light ratio for each data point includes the photometricand zero-point uncertainties in W1, the photometric uncer-tainty in r-band (Scott et al. 2013), JAM modeling uncer-tainties (Cappellari et al. 2013), and distance uncertainties(Cappellari et al. 2011).Although extinction is greatly reduced in the MIR, non-stellar emission—dust in particular—can be a major con-taminant, introducing uncertainties of up to 30% when inte-grated over the entire galaxy (Meidt et al. 2012). We there-fore use the color-cut dustless sample, described in Sec-tion 3.3, to minimize interstellar dust contamination.The color-cut dustless galaxies in Fig. 5 do not appearaffected by diffuse, interstellar dust, nor is there a notice-able trend in W1-W2. We calculate that log M/L W =0 . ± . for old stellar populations without diffuse dust.This value contrasts with the result of Meidt et al. (2014) of log M/L W = 0 . ± . . The offset is consistent with theIMF transition from Chabrier (2003) to Salpeter (1955) withincreasing velocity dispersion found by Cappellari et al.(2012) and van Dokkum & Conroy (2012). To illustrate thecontinuum of IMFs, we include FSPS tracks along metallic-ities [ Z/H ] = − . , [ Z/H ] = 0 . , and [ Z/H ] = 0 . whichinclude the bounds of the ATLAS sample and solar metal-licity, with both Chabrier (2003) and Salpeter (1955) IMFs.The majority of objects, as well as the average M/L W , liebetween these two bounds. This explanation fits our obser-vations better than the Chabrier (2003) IMF assumed byMeidt et al. (2014).The data may have a smaller range of W1-W2 thanpredicted by the FSPS models because the FSPS models donot include the metallicity-dependent CO absorption line inW2 (Peletier et al. 2012; Norris et al. 2014). We applied theGLIMPSE-calibrated correction described in Meidt et al.(2014) to the W1-W2 colors, which resulted in a narrowerand slightly bluer range of W1-W2 than observed in thecolor-cut dustless sample.One factor which may affect our interpretation of theM/L difference is that the color-cut used to separate dustlessgalaxies could preferentially select against galaxies with cer-tain velocity dispersions if there is a trend with stellar popu-lations. We found that the color-cut did preferentially selecthigh velocity dispersion galaxies, so we fit a second time andonly excluded galaxies with CO/dust detections. The newcriteria sampled a broader range of velocity dispersion andresulted in a revised value of log M/L W = − . ± . .This is closer to the Meidt et al. (2014) value, but still dis-crepant at the σ level.We also compare our results to an empirically-derivedM/L ratio from GAMA using WISE colors (Cluver et al.2014). This relation differs from that in Meidt et al. (2014)in two respects. First, the Meidt et al. (2014) relation wasfit to a grid of Bruzual & Charlot (2003) models, rectilinearin age and metallicity. In contrast, the Cluver et al. (2014)relation was fit to the resolved portion of the GAMA sample,with stellar masses derived from optical colors (Taylor et al.2011). This difference in how the fits are populated likelyleads to the difference in the slope. Secondly, the Meidt et al.(2014) relation was fit to a pure stellar population, while theCluver et al. (2014) relation was fit to very dusty galaxies.
MNRAS , 000–000 (0000)
TGs in GALEX and WISE −0.75 −0.70 −0.65 −0.60 −0.55 −0.50 W1-W2 (AB) −1.0−0.8−0.6−0.4−0.20.00.20.4 l og M ∗ L W (cid:1) M ⊙ L ⊙ (cid:0) T is workFSPS (s)FSPS (c)Meidt M/LCl()er M/LATLAS3D (d(sty)ATLAS3D (d(stless) −0.10 −0.05 0.00 0.05 0.10
W1-W2 (Vega)
Figure 5.
Stellar mass-to-light in the W1 band for the ATLAS sample along with several SPS predictions and empirical measure-ments. The data points are converted from the ATLAS r-bandstellar mass-to-light ratio, corrected to match the 2MASS K20aperture. Large blue points meet the color-cut dustless criteriadefined in Section 3.3, while small grey points do not. A repre-sentative error bar is given in the lower left. The red line is therelation of Cluver et al. (2014) from the GAMA survey. The greensolid and dashed lines indicate the M/L vs. W1-W2 relation anddispersion from Meidt et al. (2014) for old, stellar populations,derived from the Bruzual & Charlot (2003) models. The horizon-tal dotted green line represents the color-independent M/L sug-gested by Meidt et al. (2014) using an age-metallicity relation forETGs. The dashed and dotted lines connect square points fromFSPS models with Salpeter (1955) and Chabrier (2003) IMFs, re-spectively, evaluated at three different metallicities (from left toright): [Z/H] = -0.89, 0.0, 0.20 for a 10 Gyr population. Closedsquares use W1-W2 directly from FSPS. Open squares use W1-W2 colors as corrected by Meidt et al. (2014). This likely explains why the relation seems to trace the dustypopulation, but is very discrepant with the dust-free ETGs.
All low- to intermediate-mass stars (between 0.5–8 M ⊙ )pass through the Asymptotic Giant Branch (AGB) when un-dergoing hydrogen and helium shell burning (Marigo et al.2008). These stars have extremely cool and tenuous at-mospheres where dust grains condense and drive mass-loss(Salpeter 1974a,b; Goldreich & Scoville 1976; Bedijn 1987).As a result, we expect to observe this stage of stellar evo-lution in populations with ages ranging from 100 Myr togreater than the age of the Universe. However, SSP modelsindicate that the effects of circumstellar dust on integratedMIR flux peaks at 1 Gyr after star formation, and then de-creases significantly by 10 Gyr (Villaume et al. 2015).Circumstellar dust emission has been proposed as anage tracer of old stellar populations because broadband opti-cal colors have an age-metallicity degeneracy (Bressan et al.1998). Despite the lack of specific MIR age indicators for the WISE bands, SPS modeling can potentially yield an evo-lutionary track for a given color due to circumstellar dust(Villaume et al. 2015). Bregman et al. (2006) showed thatthere is some tension between ages determined from opticalline indices and circumstellar dust fits. We compare these −3.0−2.5−2.0−1.5−1.0−0.5 W - W ( A B ) SSP Age (G r) −4.0−3.5−3.0−2.5−2.0−1.5−1.0−0.50.0 W - W ( A B ) [Z/H]=-0.10[Z/H]=-0.30[Z/H] >= -0.2[Z/H] < -0.2 −0.50.00.51.01.5 W - W ( V e g a ) W - W ( V e g a ) Figure 6.
Colors of ATLAS galaxies vs. SSP age comparedwith color evolution from FSPS for W1-W3 (Top) and W1-W4 (Bottom) . Red and blue lines represent SSP models 0.1 dex aboveand below the median metallicity of the ATLAS sample, whichshould bound 60% of the ATLAS sample in metallicity. Thesolid lines represent the color evolution of the FSPS model withthe circumstellar dust prescription of Villaume et al. (2015), whilethe dashed lines represent the same population without circum-stellar dust. The ATLAS points are similarly color-coded andrepresent galaxies which lie above and below the median metallic-ity value of [ Z/H ] = − . . Galaxies with SSP ages greater than14 Gyr are reset to 14 Gyr. For objects without a robust W4detection, upper limits to the color are shown instead. two age determinations for the ATLAS sample using theFSPS models. We also investigate the effect of circumstel-lar dust on star formation rate indicators in the MIR forgalaxies with low specific SFRs (sSFR = SFR / M ∗ ).We used MIR colors to obtain a distance-independentmeasurement of circumstellar dust. Flux in the short-wavelength W1 and W2 bands is largely dominated bythe stellar continuum and is not sensitive to dust (e.g.Villaume et al. 2015). In contrast, the longer-wavelengthW3 and W4 bands should be increasingly sensitive to dust,since they lie in the region of the spectrum where circumstel-lar dust emission dominates over photospheric emission. Weavoid the W2 band in comparing data to models becausethe effect of CO absorption alluded to in Section 4.1 hasnot been included in all SPS models (Peletier et al. 2012;Norris et al. 2014).It is apparent from Fig. 3 that circumstellar dust is nec-essary in order to explain MIR colors; however, there is anoffset between the tracks and the data. We explore the sourceof this offset in more detail in Figs. 6 and 7. In Fig. 6 we usedSSP ages for the ATLAS sample from McDermid et al. MNRAS000
Colors of ATLAS galaxies vs. SSP age comparedwith color evolution from FSPS for W1-W3 (Top) and W1-W4 (Bottom) . Red and blue lines represent SSP models 0.1 dex aboveand below the median metallicity of the ATLAS sample, whichshould bound 60% of the ATLAS sample in metallicity. Thesolid lines represent the color evolution of the FSPS model withthe circumstellar dust prescription of Villaume et al. (2015), whilethe dashed lines represent the same population without circum-stellar dust. The ATLAS points are similarly color-coded andrepresent galaxies which lie above and below the median metallic-ity value of [ Z/H ] = − . . Galaxies with SSP ages greater than14 Gyr are reset to 14 Gyr. For objects without a robust W4detection, upper limits to the color are shown instead. two age determinations for the ATLAS sample using theFSPS models. We also investigate the effect of circumstel-lar dust on star formation rate indicators in the MIR forgalaxies with low specific SFRs (sSFR = SFR / M ∗ ).We used MIR colors to obtain a distance-independentmeasurement of circumstellar dust. Flux in the short-wavelength W1 and W2 bands is largely dominated bythe stellar continuum and is not sensitive to dust (e.g.Villaume et al. 2015). In contrast, the longer-wavelengthW3 and W4 bands should be increasingly sensitive to dust,since they lie in the region of the spectrum where circumstel-lar dust emission dominates over photospheric emission. Weavoid the W2 band in comparing data to models becausethe effect of CO absorption alluded to in Section 4.1 hasnot been included in all SPS models (Peletier et al. 2012;Norris et al. 2014).It is apparent from Fig. 3 that circumstellar dust is nec-essary in order to explain MIR colors; however, there is anoffset between the tracks and the data. We explore the sourceof this offset in more detail in Figs. 6 and 7. In Fig. 6 we usedSSP ages for the ATLAS sample from McDermid et al. MNRAS000 , 000–000 (0000) Simonian & Martini −2.5−2.0−1.5−1.0−0.5 W - W ( A B ) FSPSMarigo08PARSEC
SFH Age (G r) −4.0−3.5−3.0−2.5−2.0−1.5−1.0−0.50.0 W - W ( A B ) W - W ( V e g a ) W - W ( V e g a ) Figure 7.
Comparison between models and data similar to Fig. 6,except with SFH ages instead of SSP ages. Red and blue col-ors have the same meaning as in Fig. 6. Line styles correspondto different SPS models: solid lines correspond to FSPS models,dotted lines correspond to Marigo et al. (2008) SSP models, anddot-dashed lines correspond to the PARSEC SSP models. (2015) derived by fitting optical line indices. Since the SSPages in McDermid et al. (2015) were not constrained to beconsistent with the age of the Universe, we set the galaxieswith SSP ages greater than 14 Gyr to have ages of 14 Gyr,in order to be consistent with current cosmological models(Planck Collaboration 2014); this was 19% of the sample.While the discrepancies in W1-W3 do not seem be too large,the models appear to underpredict the flux in W4 by a fac-tor of 2.5, which is comparable to the offset seen in Fig. 3.As seen in Fig. 7, mass-weighted ages estimate galaxies tobe older, making the offset even more apparent. The po-tential causes of this discrepancy could include a ubiquitousintermediate-age population whose influence is seen in theMIR, but not detectable in the optical, and that a widermass range of AGB stars produce circumstellar dust at sub-solar metallicities than predicted by the models.Fig. 7 compares the FSPS models to the PAR-SEC v1.2S + COLIBRI PR16 SSP models (Bressan et al.2012; Marigo et al. 2013; Rosenfield et al. 2016) and theMarigo et al. (2008) models. The PARSEC models seem topredict no circumstellar dust at late ages, exacerbating thetension seen with the FSPS models. On the other hand, theMarigo et al. (2008) models predict more dust at late timescompared to both the FSPS and PARSEC models. However,the predicted colors of the dusty stellar population are stillbluer than indicated by the data.While some of the extremely red galaxies are likely con-taminated by interstellar dust, it is significant that very few galaxies are consistent with having no circumstellardust. Since the number of galaxies redder than the no-circumstellar-dust models is significantly greater than ourestimated contamination rate, we conclude that circumstel-lar dust is the source of the MIR excess for galaxies withoutinterstellar dust. These results are robust to differences inthe IMF.
Accurate Star Formation Rates (SFRs) provide valuableinsights into galaxy formation. The most direct methodto determine SFR measures Balmer emission from HII re-gions, which is related to the number of ionizing pho-tons from young stars; however, the extinction correctionscan be substantial and uncertain (Kennicutt 1998b). SFRsderived from MIR to FIR observations are more indi-rect as they measure the SFR through radiation repro-cessed by dust grains. This method is quite insensitiveto extinction uncertainties, but at low SFRs the old stel-lar population may also significantly heat the dust (Helou1986; Lonsdale Persson & Helou 1987; Kennicutt 1998a;Groves et al. 2012).We quantify the important contribution of circumstellardust to the total MIR luminosity, and its impact on SFRestimates with
WISE
W1, W3 and W4. W1 traces the oldstellar population that dominates stellar mass, and W3 andW4 trace the PAHs and warm dust heated by young stars,respectively, so sSFR can be calculated without additionalobservations. Although W4 is considered a superior tracerof SFR over W3, fainter galaxies are more readily detectedin W3. Since SFRs can be measured in W3 for galaxies toofaint for a W4 detection, we include it in our analysis of SFRmeasurements.Davis et al. (2014) investigated the impact of circum-stellar dust on SFR for the ATLAS sample using WISE catalog photometry, with prescriptions calibrated for
Spitzer bands. We reinvestigate this prescription with our updatedphotometry, more stringent criteria for dustless galaxies, andwith the more recent SFR relations of Cluver et al. (2014).We plot the W1 vs. W3 and W4 luminosity for theATLAS sample in Fig. 8 (see Figure 1 of Davis et al.(2014) for a similar plot with K s ). The color-cut dustlesssample described in Section 3.3 is shown in red. The color-cut sample was chosen because it should have less interstellardust contamination compared to the regular dustless sam-ple, and should represent an accurate relation between W1and W3 (or W4) in the absence of star formation. Becausewe may have excluded some truly dust-free galaxies whosecolors do not match our exclusion criteria, the scatter inour relation may be underestimated. Follow-up observationsof the ATLAS sample with FIR/submillimeter measure-ments or high-resolution visible-wavelength images to searchfor dust lanes would result in a more representative sampleof dustless galaxies.In order to characterize the minimum SFR which canbe usefully measured with W4 observations, we calculatesSFR limits with the relations in Cluver et al. (2014) anduse the mean W1-W2 value of our sample to get M/L W .The results shown in Fig. 8 indicate that the contributionof circumstellar dust to W3 and W4 will mimic an sSFR MNRAS , 000–000 (0000)
TGs in GALEX and WISE l og ν L W ( e r g s − ) log L W1 (L ⊙ ) l og ν L W ( e r g s − ) log sSFR=-10log sSFR=-11log sSFR=-12Dusty colorDustless color Figure 8. (Top)
W3 vs. W1 and (Bottom)
W4 vs. W1 luminosi-ties for the ATLAS sample. Circles are detections and trianglesare W4 upper limits. Red points are in the dustless color-cut sam-ple. Error bars are smaller than the points. The blue, green, andred lines represent where specific star formation rate is − yr − , − yr − , − yr − . The black line represents a fit tothe color-cut dustless galaxy detections. Note that W3 is signifi-cantly more sensitive than W4. of × − yr − . Lower values of sSFR cannot be reliablyinferred with integrated WISE photometry alone.Even at larger sSFR values, it is still necessary to re-move the effect of circumstellar dust. In order to quantifythe effect, we fit a line to the dustless detections and obtain: log νL W = 1 .
03 log L W + 30 . , (2) log νL W = 0 .
97 log L W + 30 . (3)with an RMS scatter of 0.03 and 0.04 dex, where L W is the“in-band” luminosity in solar luminosities, and νL ν is thespectral luminosity of the band in erg s − . These relationsmay be used to subtract the circumstellar dust contributionto W3 or W4 for relatively quiescent galaxies. The W4 rela-tion is similar to the relation derived by Davis et al. (2014)between Ks and W4. These corrections will likely be negli-gible for typical star-forming galaxies as the sSFR of local L ∗ galaxies is around − yr − (Cluver et al. 2014). GALEX
AND
WISE
Mid-infrared spectra of early-type galaxies have shownthat many exhibit much weaker short-wavelength PAH fea-tures (6.2, 7.7, and 8.6 µ m) relative to those at longerwavelengths (e.g. 11.3 and 12.7 µ m) (Kaneda et al. 2005,2008; Vega et al. 2010; Rampazzo et al. 2013) compared tostar forming galaxies (Helou et al. 2000; Brandl et al. 2004;Smith et al. 2007). Observations of later-type galaxies havealso shown that low values of these band ratios are com-monly found in low-luminosity AGN (Smith et al. 2007; . µ m / . µ m SeyfertsSINGS LINERSSINGS HIIETGs
Figure 9.
Plot of 11.3 µ m/7.7 µ m vs. 6.2 µ m/7.7 µ m line ra-tios for different samples of galaxies. Seyfert line ratios weretaken from Diamond-Stanic & Rieke (2010); SINGS ratios werefrom Smith et al. (2007); the ETG sample contains galaxies fromSmith et al. (2007) and Rampazzo et al. (2013). Stars indicatethat the ETG is Class-4 according to Rampazzo et al. (2013). O’Dowd et al. 2009). Many early-type galaxies contain ev-idence for LINERs (Ho et al. 1997), although as most maynot be dominated by AGN (Sarzi et al. 2010), it is not clearif there is a direct connection between AGN and the propor-tionally weaker short-wavelength PAH bands in early-typegalaxies.
We have used data from three significant studies of PAHemission from nearby galaxies to probe the relationship be-tween PAH emission, AGN, and host galaxy morphologyto further explore the different PAH band ratios seen inearly-type galaxies. The largest study of early-type galaxiesis the RSA
Spitzer -IRS Atlas by Rampazzo et al. (2013),and about half of their sample of ETGs exhibit PAH emis-sion. Smith et al. (2007) performed a detailed analysis ofPAH emission from galaxies in the
Spitzer
Infrared NearbyGalaxy Survey (SINGS: Kennicutt et al. 2003). This sam-ple spans a wide range of luminosity and infrared to visi-ble wavelength flux ratio, and includes early and late-typegalaxies. They have also classified the galaxies as either “HII”galaxies, which are dominated by star formation, LINERs,or Seyferts. Finally, Diamond-Stanic & Rieke (2010) studied35 Seyfert galaxies in the Revised Shapley-Ames catalog. Allof these Seyferts have late-type host galaxies (later than S0),and generally exhibit weaker short-wavelength PAH emis-sion compared to HII galaxies. This study included 21 galax-ies with PAH emission in spectra not centered on the nuclearregion, and these off-nuclear spectra exhibit PAH ratios sim-ilar to HII galaxies.Diamond-Stanic & Rieke (2010) investigated if theweaker 6.2, 7.7, and 8.6 µ m bands in Seyferts could bedue to radiative or mechanical processing. Previous work(Szczepanski & Vala 1993; Hudgins & Allamandola 1995;DeFrees et al. 1993; Langhoff 1996) has shown that the C − C stretching modes that give rise to the 6.2 and 7.7 µ m MNRAS000
Infrared NearbyGalaxy Survey (SINGS: Kennicutt et al. 2003). This sam-ple spans a wide range of luminosity and infrared to visi-ble wavelength flux ratio, and includes early and late-typegalaxies. They have also classified the galaxies as either “HII”galaxies, which are dominated by star formation, LINERs,or Seyferts. Finally, Diamond-Stanic & Rieke (2010) studied35 Seyfert galaxies in the Revised Shapley-Ames catalog. Allof these Seyferts have late-type host galaxies (later than S0),and generally exhibit weaker short-wavelength PAH emis-sion compared to HII galaxies. This study included 21 galax-ies with PAH emission in spectra not centered on the nuclearregion, and these off-nuclear spectra exhibit PAH ratios sim-ilar to HII galaxies.Diamond-Stanic & Rieke (2010) investigated if theweaker 6.2, 7.7, and 8.6 µ m bands in Seyferts could bedue to radiative or mechanical processing. Previous work(Szczepanski & Vala 1993; Hudgins & Allamandola 1995;DeFrees et al. 1993; Langhoff 1996) has shown that the C − C stretching modes that give rise to the 6.2 and 7.7 µ m MNRAS000 , 000–000 (0000) Simonian & Martini features, and the C − H in-plane bending modes that giverise to the 8.6 µ m feature, are more readily produced inionized PAHs. The ratio of these features to the C − H out-of-plane bending mode that gives rise to the 11.3 µ mfeature (Duley & Williams 1981; Allamandola et al. 1989)will be lower for more neutral PAHs. Fig. 9 shows theratio of 11.3 µ m/7.7 µ m vs. 6.2 µ m/7.7 µ m for measure-ments from these three studies. The early-type galaxies in-clude both the Rampazzo et al. (2013) sample and early-type galaxies in Smith et al. (2007), the Seyfert sample in-cludes late-type Seyferts from both Smith et al. (2007) andDiamond-Stanic & Rieke (2010), and the LINERs and HIIgalaxies are only galaxies with late-type morphology fromSmith et al. (2007). This diagram clearly shows that theearly-type galaxies have larger 11.3 µ m/7.7 µ m than the HIIgalaxies, but also that the ratio is larger than for the Seyfertsand perhaps the LINER sample. Only galaxies with detec-tions in both ratios are shown, as the literature sources gen-erally do not quote upper limits for non-detections. Verydifferent values of 11.3 µ m/7.7 µ m vs. 6.2 µ m/7.7 µ m areexpected for neutral vs. ionized PAHs. Allamandola et al.(1999) and Draine & Li (2001) showed that neutral PAHslead to the intensity ratio 11.3 µ m/7.7 µ m > . , whereas11.3 µ m/7.7 µ m < . is more characteristic of ionizedPAHs, although the value of this ratio also depends onthe PAH size distribution, as does the 6.2 µ m/7.7 µ m ratio.Only about half of the ETGs have 11.3 µ m/7.7 µ m > . asexpected for neutral PAHs.Fig. 10 plots the cumulative distribution functions ofthese four classes of galaxies. We used a Kolmogorov-Smirnov two-sample test to measure the significance of thedifferences. We found that the 7.7 µ m/11.3 µ m differencebetween the early-type galaxies and the Seyferts, as well asbetween the early-type galaxies and the HII galaxies, is verysignificant ( p ≪ . ). The difference between the early-typegalaxies and LINERs is marginally significant ( p ∼ . )as well. Note that the LINER sample is somewhat smallerthan the other samples, and that the cumulative distribu-tions contain all galaxies with detections in the single ratio,and thus may have more points than Fig. 9.Galliano et al. (2008) noted that the typical ionizationstate of PAHs is set by G T / /n e (Bakes & Tielens 1994),where G is the intensity of the UV radiation field, T isthe gas temperature, and n e is the electron density, and weinvestigated if the distribution of Class-2 and Class-3 ob-jects followed trends in any of these physical quantities. Thegas temperature can be determined from the H S(3)/H S(1)line ratio, and the electron density can be determined from[S iii ] 18.7 µ m/[S iii ] 33.5 µ m. These lines have been mea-sured by Rampazzo et al. (2013), and any trends shouldcause the two classes to separate in one of the dimensions.As seen in Fig. 11, Class-2 and Class-3 objects have simi-lar values of these ratios, so we conclude that trends withgas temperature or electron density are not the origin of thedifference between the two classes.Since the temperatures and densities for the Class-2and Class-3 galaxies are similar, we expect the degree ofionization for PAHs will depend on the shape of the inter-stellar radiation field (ISRF) (Weingartner & Draine 2001).The ISRF in early and late-type galaxies varies dramati-cally, both in normalization and shape. The dramatic lackof young stars, and therefore UV emission, can even lead op- tical photons to become the primary heating mechanism forinterstellar dust (Groves et al. 2012). While we do not di-rectly measure the intensity of the ISRF, we use the NUV-Jcolor as a proxy for the slope. We show the trend in theslope of the radiation field with the 7.7 µ m/11.3 µ m ra-tio in Fig. 12. Even though Class-4 objects also have PAHemission, they were excluded as likely AGN because theirhigh 1.4 GHz and nuclear X-ray emission may distort theISRF (Rampazzo et al. 2013). There does not appear tobe any correlation between the shape of the ISRF and the7.7 µ m/11.3 µ m ratio, which corresponds directly with MIRclass.While the relation between ionization state and UVintensity works well for many environments (Joblin et al.1996; Bregman & Temi 2005), Diamond-Stanic & Rieke(2010) pointed out that many Seyferts have ratios consistentwith completely neutral PAHs, or even more extreme ratiosthan the models allow, and in any case the more intenseradiation field associated with AGN should not lead to lessionization. We find that the early-type galaxies often haveeven smaller values of the 7.7 µ m/11.3 µ m band ratio thanthe Seyferts. Draine & Li (2001) noted that smaller valuesof 7.7 µ m/11.3 µ m could be produced in the charging condi-tions characteristic of the Cold Neutral Medium (CNM) orWarm Ionized Medium (WIM), which have lower densitiesand radiation field intensities than typical of the Photodis-sociation Regions (PDRs) associated with star formation. Ifmost of the PAHs are associated with conditions more sim-ilar to the CNM or WIM, this could explain the dramaticdifference in their ratios relative to star-forming galaxies, aswell as why they are not similar to Seyferts, which typicallyalso have PDRs.As the three shorter-wavelength PAH features are alsoassociated with smaller PAHs, an alternative explanation oftheir proportionally lower fluxes is a different size distribu-tion (Schutte et al. 1993; Draine & Li 2001; Galliano et al.2008). Experimental work on PAHs has indicated that PAHsbelow 15-20 C atoms would be destroyed in most environ-ments, PAHs smaller than 20-30 C atoms may be strippedof their H atoms, and PAHs of 30-50 C atoms would mostlybe photoionized (Jochims et al. 1994; Allain et al. 1996). Ifsmaller PAHs have been destroyed, then this would alsolead to lower 6.2 µ m relative to 7.7 µ m and lower 7.7 µ mrelative to 8.6 µ m. Fig. 13 shows these two ratios for thesame galaxies shown in Fig. 9. Based on inspection of thisfigure, as well as Kolmogorov-Smirnov (KS) tests appliedto the cumulative distributions, we do not see significantdifferences between the early-type galaxies and other pop-ulations. Diamond-Stanic & Rieke (2010) similarly did notfind significant differences between Seyferts and either theiroff-nuclear positions or the HII galaxies from Smith et al.(2007). As illustrated in Draine & Li (2001), larger PAHscan also contribute to the larger 11.3 µ m/7.7 µ m ratios seenin Fig. 9, although they also have smaller 6.2 µ m/7.7 µ m ra-tios. Intriguingly, about half of the early-type galaxies have6.2 µ m/7.7 µ m < . , although the difference is not for-mally significant. Careful estimates of upper limits on theweak 6.2 µ m feature in the several early-type galaxies withonly 7.7 µ m detections, as well as more sensitive measure-ments with future facilities, would help to better character-ize the smaller PAHs in these galaxies that are most sensitiveto destruction processes. MNRAS , 000–000 (0000)
TGs in GALEX and WISE All ETGsSINGS LT HIISINGS LT LINERsAll Seyferts
Figure 10.
Cumulative distribution functions of the galaxy classes in Fig. 9 for (Top left) µ m/7.7 µ m, (Top right) µ m/11.3 µ m, (Bottom right) µ m/11.3 µ m, (Bottom left) µ m/11.3 µ m. There is a statistically significant difference between the ETGs andSINGS late-type (LT) HII galaxies, as well as the ETGs and Seyferts in 7.7 µ m/11.3 µ m. The distribution between ETGs and LINERsin 7.7 µ m/11.3 µ m was marginally significant. ETGs are also significantly different from Seyferts and LINERs in 17 µ m/11.3 µ m, butare similar to HII galaxies. There are no significant differences between the galaxy distributions in any of the other line ratios. Another explanation proposed for the variations in7.7 µ m/11.3 µ m is interstellar shocks. Shock velocities ofon order 100 km s − can reduce the relative number ofsmall grains (Jones et al. 1994; Micelotta et al. 2010), andthus lead to weaker emission in all three of the shorter-wavelength features. If shocks are present, then the galax-ies may also exhibit strong H emission. Many studieshave found a correlation between H emission and the7.7 µ m/11.3 µ m ratio (Roussel et al. 2007; Ogle et al. 2007;Kaneda et al. 2008; Vega et al. 2010). Fig. 14 (left) illus-trates the correlation shown by Diamond-Stanic & Rieke(2010) for the RSA Seyferts, along with the early-typegalaxies from Rampazzo et al. (2013). The early-type galax-ies mostly follow the same trend as the Seyferts, al-though they generally have more H emission and thereis more dispersion in 7.7 µ m/11.3 µ m for the early-type galaxies with the proportionally strongest H emis-sion. Diamond-Stanic & Rieke (2010) and Rampazzo et al.(2013) found that AGN power did not correlate with thisratio, and Fig. 14 (right) confirms this point with a plot of the ratio of H S(3)/(7.7 µ m+ 11.3 µ m) vs. nuclear X-rayluminosity from Pellegrini (2010). Processing by shocks ap-pears to be a more viable contributor to the 7.7 µ m/11.3 µ mratios in early-type galaxies, as it is for Seyferts.In addition to the preferential destruction of smallPAHs, shocks could also affect the chemistry of the PAHs,even those that are fully hydrogenated. For example,the 11.3 µ m band is produced by single C—H bonds,while the 12.7 µ m feature is produced by C—H multi-plets (Hony et al. 2001). Diamond-Stanic & Rieke (2010)found that Seyferts exhibit significantly smaller ratios of12.7 µ m/11.3 µ m than off-nuclear and HII galaxies, whichcould be due to different processing. Figs. 10 and 15 showsthat the early-type galaxies are not significantly differentfrom the Seyferts or the HII galaxies, so we see no evidencethat this is an important physical difference in the early-typegalaxy population.The 12.7 µ m/11.3 µ m band ratio shown in Fig. 15is plotted vs. the 17 µ m/11.3 µ m band ratio. Both the11.3 µ m band and the 17 µ m band originate from neutral MNRAS000
Cumulative distribution functions of the galaxy classes in Fig. 9 for (Top left) µ m/7.7 µ m, (Top right) µ m/11.3 µ m, (Bottom right) µ m/11.3 µ m, (Bottom left) µ m/11.3 µ m. There is a statistically significant difference between the ETGs andSINGS late-type (LT) HII galaxies, as well as the ETGs and Seyferts in 7.7 µ m/11.3 µ m. The distribution between ETGs and LINERsin 7.7 µ m/11.3 µ m was marginally significant. ETGs are also significantly different from Seyferts and LINERs in 17 µ m/11.3 µ m, butare similar to HII galaxies. There are no significant differences between the galaxy distributions in any of the other line ratios. Another explanation proposed for the variations in7.7 µ m/11.3 µ m is interstellar shocks. Shock velocities ofon order 100 km s − can reduce the relative number ofsmall grains (Jones et al. 1994; Micelotta et al. 2010), andthus lead to weaker emission in all three of the shorter-wavelength features. If shocks are present, then the galax-ies may also exhibit strong H emission. Many studieshave found a correlation between H emission and the7.7 µ m/11.3 µ m ratio (Roussel et al. 2007; Ogle et al. 2007;Kaneda et al. 2008; Vega et al. 2010). Fig. 14 (left) illus-trates the correlation shown by Diamond-Stanic & Rieke(2010) for the RSA Seyferts, along with the early-typegalaxies from Rampazzo et al. (2013). The early-type galax-ies mostly follow the same trend as the Seyferts, al-though they generally have more H emission and thereis more dispersion in 7.7 µ m/11.3 µ m for the early-type galaxies with the proportionally strongest H emis-sion. Diamond-Stanic & Rieke (2010) and Rampazzo et al.(2013) found that AGN power did not correlate with thisratio, and Fig. 14 (right) confirms this point with a plot of the ratio of H S(3)/(7.7 µ m+ 11.3 µ m) vs. nuclear X-rayluminosity from Pellegrini (2010). Processing by shocks ap-pears to be a more viable contributor to the 7.7 µ m/11.3 µ mratios in early-type galaxies, as it is for Seyferts.In addition to the preferential destruction of smallPAHs, shocks could also affect the chemistry of the PAHs,even those that are fully hydrogenated. For example,the 11.3 µ m band is produced by single C—H bonds,while the 12.7 µ m feature is produced by C—H multi-plets (Hony et al. 2001). Diamond-Stanic & Rieke (2010)found that Seyferts exhibit significantly smaller ratios of12.7 µ m/11.3 µ m than off-nuclear and HII galaxies, whichcould be due to different processing. Figs. 10 and 15 showsthat the early-type galaxies are not significantly differentfrom the Seyferts or the HII galaxies, so we see no evidencethat this is an important physical difference in the early-typegalaxy population.The 12.7 µ m/11.3 µ m band ratio shown in Fig. 15is plotted vs. the 17 µ m/11.3 µ m band ratio. Both the11.3 µ m band and the 17 µ m band originate from neutral MNRAS000 , 000–000 (0000) Simonian & Martini H S(3) / H S(1) [ S III ] . µ m / [ S III ] . µ m Class 2Class 3
Figure 11.
Class-2 and Class-3 sulfur and molecular hydrogenline ratios as measured in Rampazzo et al. (2013). We expectH S(3)/H S(1) to be a proxy for the gas temperature of the emit-ting region, and [S iii ] 18.7 µ m/[S iii ] 33.5 µ m to be a proxy forthe electron density. Objects without detections in any of the lineshave been excluded from the figure. There does not appear to bea significant distinction between these two classes in this plane. NUV-J (AB) . µ m / . µ m Class 2Class 3
Figure 12.
Distribution of Class-2 and Class-3 sources in NUV-J, which acts as a proxy for the slope of the ISRF, and7.7 µ m/11.3 µ m, which reflects the level of “anomaly” in the PAHratios. PAHs, and so their relative strength is primarily sensitive tothe size distribution. Smith et al. (2007) found that larger17 µ m/11.3 µ m intensity ratio correlates with metallicity,and suggest that it may be easier to form the larger grainsthat contribute to the 17 µ m band in higher metallicityenvironments. They find that 17 µ m/11.3 µ m is largest inSeyferts and suggest that these high ratios are due to a com-bination of the higher metallicities of the Seyfert hosts andthe relative destruction of the carriers of the shorter wave-length 11.3 µ m feature. As shown in Fig. 15, we find thatearly-type galaxies have significantly lower µ m/11.3 µ mratios than the Seyferts (KS p < . ) and LINERs (KS p < . ) from Smith et al. (2007), and are similar toHII galaxies (see also Fig. 10). Early-type galaxies may have . µ m / . µ m SeyfertsSINGS LT LINERSSINGS LT HIIETGs
Figure 13.
Plot of line ratios like Fig. 9 except for 8.6 µ m/7.7 µ mand 6.2 µ m/7.7 µ m line ratios. Larger grain size distributionstend to have suppressed 6.2 µ m/7.7 µ m emission. There are nostatistically significant differences in the ratios between classes,indicating that a systematic difference in grain size distributionmay not be correlated with anomalous PAH emission. lower 17 µ m/11.3 µ m band ratios because they lack the hardradiation field of Seyferts. Alternatively, this ratio could belower in some early-type galaxies because their PAHs havebeen externally accreted from lower-metallicity dwarf galax-ies. We cannot separate the effect of metallicity on the17 µ m/11.3 µ m ratio because of the sparse overlap betweenthe ATLAS and Rampazzo et al. (2013) samples. Only 13galaxies in our sample have both ATLAS metallicity and17 µ m/11.3 µ m measurements; their Pearson correlation co-efficient is -0.04, which indicates a very weak anticorrelation. Currently the only way to discriminate between MIR classesis through infrared spectra from targeted spectroscopic ob-servations in space. This is a significant investment for space-based missions which are already highly oversubscribed.Now that the
WISE mission has imaged the entire sky inthe MIR, a successful method of classifying galaxies intodifferent MIR classes based on
WISE data would allow de-mographic analysis of anomalous PAH emission.Because most of the PAH emission features lie in thevery broad W3 filter, we expect most of the discriminatingpower to occur with the W2-W3 and W3-W4 colors. We plotthe Rampazzo et al. (2013) sample in Fig. 16. Many of theClass-4 objects lie far away from the bulk of the distribution,likely due to the hot dust continuum in their spectra. Therealso seems to be good discrimination between Class-0 andClass-2/3 objects in W2-W3. The Class-0 objects are sys-tematically bluer than both the Class-2 (K-S p < − ) andClass-3 (K-S p < − ) objects. A cumulative histogram ofboth classes in W2-W3 is shown in Fig. 17. 85% of Class-0, 11% of Class-2, and no Class-3 objects are bluer than W − W . (Vega, or -1.2 AB), providing a conve-nient color-cut between dusty and passive ETGs. No other MNRAS , 000–000 (0000)
TGs in GALEX and WISE H S ( ) / ( . µ m + . µ m ) R13 ETGsDSR10 Seyferts
39 40 41log L
X,nuc
Figure 14.
Left:
Relative strength of the H S(3) transition to PAH emission, which traces shocks in cold gas, versus the 7.7 µ m/11.3 µ mratio. Black points are ETGs from Rampazzo et al. (2013) while blue points are Seyferts from Diamond-Stanic & Rieke (2010). Starsindicate Class-4 galaxies. Both the ETGs and Seyferts indicate a correlation between the shocks and anomalous PAH ratios. Right:
Molecular hydrogen strength vs. nuclear X-ray luminosity for Rampazzo et al. (2013) galaxies with nuclear X-ray detections. As notedin Diamond-Stanic & Rieke (2010), there does not appear to be a trend between shock strength and X-ray luminosity. major distinctions between classes are statistically signifi-cant.One reason why
WISE is unable to discriminate be-tween classes very well is because the bands are notwell-placed with respect to PAH features. The only bandsensitive to PAH emission is W3, which covers all ofthe 7.7 µ m, 8.6 µ m, 11.3 µ m, and 12.7 µ m PAH com-plexes (Wright et al. 2010). The W3 band accomplishes thisamount of spectral coverage with a wide effective bandwidthof 5.5 µ m (Jarrett et al. 2011). By comparison, the typicalequivalent width of a line is around 0.1 µ m (Rampazzo et al.2013). Therefore, the difference between the presence andabsence of all five PAH lines is around 0.1 mag, which issimilar to the photometric error of our measurements. Whilethis should suffice to separate Class-0 objects from Class-2/3objects, separation of Class-2 and Class-3 objects requiresmore precise photometry. Another complication is the largeintrinsic scatter in Class-2 and Class-3, which we attributeto varying levels of contribution by the dust continuum.Despite the fact that WISE colors on their own can-not discriminate between the MIR classes, they can be use-ful for informing follow-up targets. This will be particularlytrue with the Mid-Infrared Instrument (MIRI) on the
JamesWebb Space Telescope (JWST) . For galaxies which have W2-W3 colors consistent with some interstellar dust, better de-mographics on the prevalence of anomalous PAH ratios can be determined from follow-up with the MIRI Medium Res-olution Spectrograph (MRS).
We performed
WISE and
GALEX aperture photometry ona combined sample of 91 ETGs from Rampazzo et al. (2013)which had uniform MIR line flux measurements, and 260ETGs from the ATLAS survey, which have significant an-cillary data. The photometric accuracy is about 0.05 magin the MIR and 0.1 mag in the UV. We chose an identicalaperture to one used by the 2MASS XSC.The MIR is a useful wavelength region to calculate stel-lar masses since the effects of extinction and young stars arereduced compared to shorter wavelengths. We converted theATLAS ( M/L ) stars from r-band to W1 and corrected forthe change in aperture between the two studies. We foundthat the ATLAS M/L W measurements are significantlylarger than those predicted by SPS models. We did not finda trend between M/L W and W1-W2 for old stellar pop-ulations predicted by SPS models. We obtained an aver-age log M/L W = 0 . ± . , which is based on dynamical M/L W rather than fits to SPS models. Our color selectionof dustless galaxies biases our value toward higher M/L W ;but without the color-cut, we still observe a σ discrepancy.Our high M/L W compared to SPS models agrees with MNRAS , 000–000 (0000) Simonian & Martini . µ m / . µ m SINGS LT LINERSSINGS LT HIISINGS LT SeyfertsETGs
Figure 15.
Plot like Fig. 9, but with 12.7 µ m/11.3 µ m, whicharises from C—H multiplets, vs. 17 µ m/11.3 µ m, which arisesfrom single C—H bonds. While there does seem to be a distinctionbetween HII galaxies and Seyferts in the 12.7 µ m/11.3 µ m ratio,the ETG population is not statistically distinct from either ofthem. For the 17 µ m/11.3 µ m ratio, ETGs are similar to HIIgalaxies, but distinct from LINERs and Seyferts. recent research indicating that the IMF varies, and is moreSalpeter-like at high velocity dispersions. When the M/L W from our galaxies is compared to FSPS models with Chabrier(2003) and Salpeter (1955) IMFs, and we find that thedynamically-derived masses mostly fall in-between the twoIMFs, mirroring the effect seen from M/L W derived fromoptical spectra.We clearly identify circumstellar dust in the W3 and W4bands for many ETGs. Surprisingly, this circumstellar dustemission appears to be ubiquitous at much greater ages thanpredicted by SPS models, even after accounting for potentialcontamination in our dustless sample. We also find that W4emission is underpredicted by models by about a factor of2.5. This underprediction is also seen in other SPS models.With all-sky coverage, the WISE mission makes stellarmass and SFR estimates very accessible. The
WISE
W3 andW4 bands are sensitive to dust warmed by recent star forma-tion. For objects with low sSFR, circumstellar dust emissioncan contaminate these tracers, leading to an inflated result.We determined that circumstellar dust contributes approx-imately × − yr − to the sSFR signal in both W3 andW4 and we provide relations to correct for the circumstellardust contribution to W3 and W4.Lastly, we compared MIR spectra for spirals, LINERs,Seyferts, and ETGs in order to investigate the source ofanomalous PAH ratios. We found that ETGs and Seyfertshave significantly distinct 11.3 µ m/7.7 µ m ratios from spi-rals, as well as from each other. This strongly suggested thatthe cause of anomalous PAH ratios is more than just thepresence of an AGN. The ETGs also have significantly dis-tinct 17 µ m/11.3 µ m values from the LINERs and Seyferts.Using available [S iii ] 18.7 µ m/[S iii ] 33.5 µ m andH S(3)/H S(1) ratios, we noted that gas temperature andelectron density do not correlate with the presence ofanomalous PAH ratios. The incident UV flux on the PAHsalso does not appear to correlate with 7.7 µ m/11.3 µ m ra-tio. Based on these results, we conclude that the anomalous PAH ratios are unlikely to be due to global variation in theISRF.We did not see evidence for the preferential destruc-tion of smaller PAHs based on the 6.2 µ m/7.7 µ m and8.6 µ m/7.7 µ m ratios. However, due to the nonuniformavailability of upper limits in published line intensities, weexcluded galaxies without detections. A uniform treatmentof upper limits may reveal more information about smallPAHs in these galaxies. We did not find evidence of dehy-drogenation of small PAHs in the 12.7 µ m/11.3 µ m ratio.The strongest correlation with anomalous PAH emis-sion is the strength of shocks in H . We found that bothETGs and Seyferts follow this trend, although ETGs gener-ally have more H emission. Exactly how shocks bring aboutthe anomalous emission is still largely uncertain. However,we determined from X-ray luminosities that it is likely unre-lated to the activity of the central supermassive black hole.One particularly interesting result we found was thatthe ETGs have 17 µ m/11.3 µ m ratios significantly lowerthan Seyferts and LINERS, but similar to spirals. This ra-tio is sensitive to the presence of large grains, and has beenobserved to trend with metallicity. Our result can be ex-plained by the lack of the hard radiation environment foundin Seyferts, or if the PAHs were accreted from an external,low-metallicity galaxy.Finally, we attempted to use WISE to classify the PAHproperties of ETGs. We found that
WISE colors are onlyable to determine the presence of dust in ETGs, but notthe MIR class. We conclude that finer distinctions are notpossible because of the spectral placement of the
WISE fil-ters and our photometric precision. However, the W2-W3colors may be able to inform detailed follow-up spectra ofearly-type galaxies in order to more efficiently calculate theprevalence of anomalous PAH ratios.
ACKNOWLEDGMENTS
GVS and PM would like to thank C. Conroy for his valu-able insights into SPS models, A. Gil de Paz for helpfulinformation about
GALEX photometry, and J.D. Smith foruseful discussions about PAHs, and the anonymous referee.GVS would also like to thank T. Jarrett for providing addi-tional details regarding
WISE fluxes, luminosities, masses,and SFRs for nearby galaxies. This publication makes use ofdata products from the
Wide-field Infrared Survey Explorer ,which is a joint project of the University of California, LosAngeles, and the Jet Propulsion Laboratory/California In-stitute of Technology, funded by the National Aeronauticsand Space Administration. Some/all of the data presentedin this paper were obtained from the Mikulski Archive forSpace Telescopes (MAST). STScI is operated by the Associ-ation of Universities for Research in Astronomy, Inc., underNASA contract NAS5-26555. Support for MAST for non-
HST data is provided by the NASA Office of Space Sciencevia grant NNX13AC07G and by other grants and contracts.This research has made use of the NASA/IPAC Extragalac-tic Database (NED) which is operated by the Jet Propul-sion Laboratory, California Institute of Technology, undercontract with the National Aeronautics and Space Admin-istration.This research made use of Astropy, a community-
MNRAS , 000–000 (0000)
TGs in GALEX and WISE −1.4 −1.2 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 W2-W3 (AB) −0.74−0.72−0.70−0.68−0.66−0.64−0.62−0.60 W - W ( A B ) Class 0Class 1Class 2Class 3Class 4
W2-W3 (Vega) −0.10−0.08−0.06−0.04−0.020.000.020.04 W - W ( V e g a ) −1.4 −1.2 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 W3-W4 (AB) −1.4−1.2−1.0−0.8−0.6−0.4−0.20.0 W - W ( A B ) W3-W4 (Vega) W - W ( V e g a ) Figure 16.
Left:
W1-W2 vs. W2-W3 for the sample of galaxies in Rampazzo et al. (2013). W1-W2 is sensitive to the stellar population,so we do not see separation based on properties of the interstellar dust. W2-W3 measures the presence of PAH features compared to thestellar population. This causes the Class-0 objects separate from the Class-2-4 objects. Galaxies located off the plot are: IC 5063, NGC1052, NGC 1275, NGC 2685, NGC 4383, and NGC 5128.
Right:
W2-W3 vs W3-W4 for the sample of galaxies in Rampazzo et al. (2013).There does not appear to be strong discrimination in W3-W4 due to the hot dust population, as seen in W2-W3. Located off the plotto the top and to the right are: all Class-4 objects, NGC 2685, NGC 3245, NGC 4383, and NGC 4435. −1.5 −1.0 −0.5 0.0 0.5
W2-W3 (AB) N Class 3Class 2Class 0
W2-W3 (Vega)
Figure 17.
Cumulative histogram of the W2-W3 colors of theClass-0, Class-2, and Class-3 objects. The Class-0 objects arefrom a population statistically distinct from both the Class-2 andClass-3 objects. We propose a color-cut to distinguish betweenClass-0 and Class-2–3 objects at W2-W3=-1.2 (AB), which isdenoted by the red dashed line. Class-4 objects are from a pop-ulation statistically distinct from all other classes. There are noother separations which can be made using W2-W3 colors. developed core Python package for Astronomy(Astropy Collaboration 2013). This research made useof APLpy, an open-source plotting package for Pythonhosted at http://aplpy.github.com. This research made useof the IPython package (Pérez & Granger 2007). IRAF isdistributed by the National Optical Astronomy Observa-tory, which is operated by the Association of Universitiesfor Research in Astronomy (AURA) under a cooperativeagreement with the National Science Foundation. Thisresearch made use of matplotlib, a Python library forpublication quality graphics (Hunter 2007). PyRAF is aproduct of the Space Telescope Science Institute, which is operated by AURA for NASA. This research made use ofSciPy (Jones et al. 01 ).
REFERENCES
Abazajian K. N., et al., 2009, ApJS, 182, 543Alatalo K., et al., 2013, MNRAS, 432, 1796Allain T., Leach S., Sedlmayr E., 1996, A&A, 305, 602Allamandola L. J., Tielens A. G. G. M., Barker J. R., 1989, ApJS,71, 733Allamandola L. J., Hudgins D. M., Sandford S. A., 1999, ApJ,511, L115Amblard A., Riguccini L., Temi P., Im S., Fanelli M., Serra P.,2014, ApJ, 783, 135Astropy Collaboration et al., 2013, A&A, 558, A33Athey A., Bregman J., Bregman J., Temi P., Sauvage M., 2002,ApJ, 571, 272Bai Y., Zou H., Liu J., Wang S., 2015, ApJS, 220, 6Bakes E. L. O., Tielens A. G. G. M., 1994, ApJ, 427, 822Bedijn P. J., 1987, A&A, 186, 136Bell E. F., de Jong R. S., 2001, ApJ, 550, 212Bell E. F., McIntosh D. H., Katz N., Weinberg M. D., 2003, ApJS,149, 289Bertin E., Arnouts S., 1996, A&AS, 117, 393Blanton M. R., Roweis S., 2007, AJ, 133, 734Brandl B. R., et al., 2004, ApJS, 154, 188Bregman J., Temi P., 2005, ApJ, 621, 831Bregman J. N., Snider B. A., Grego L., Cox C. V., 1998, ApJ,499, 670Bregman J. N., Temi P., Bregman J. D., 2006, ApJ, 647, 265Bressan A., Granato G. L., Silva L., 1998, A&A, 332, 135Bressan A., Marigo P., Girardi L., Salasnich B., Dal Cero C.,Rubele S., Nanni A., 2012, MNRAS, 427, 127Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Caldwell N., 1984, PASP, 96, 287Canizares C. R., Fabbiano G., Trinchieri G., 1987, ApJ, 312, 503Cappellari M., 2008, MNRAS, 390, 71Cappellari M., et al., 2011, MNRAS, 413, 813Cappellari M., et al., 2012, Nature, 484, 485Cappellari M., et al., 2013, MNRAS, 432, 1709MNRAS000
Abazajian K. N., et al., 2009, ApJS, 182, 543Alatalo K., et al., 2013, MNRAS, 432, 1796Allain T., Leach S., Sedlmayr E., 1996, A&A, 305, 602Allamandola L. J., Tielens A. G. G. M., Barker J. R., 1989, ApJS,71, 733Allamandola L. J., Hudgins D. M., Sandford S. A., 1999, ApJ,511, L115Amblard A., Riguccini L., Temi P., Im S., Fanelli M., Serra P.,2014, ApJ, 783, 135Astropy Collaboration et al., 2013, A&A, 558, A33Athey A., Bregman J., Bregman J., Temi P., Sauvage M., 2002,ApJ, 571, 272Bai Y., Zou H., Liu J., Wang S., 2015, ApJS, 220, 6Bakes E. L. O., Tielens A. G. G. M., 1994, ApJ, 427, 822Bedijn P. J., 1987, A&A, 186, 136Bell E. F., de Jong R. S., 2001, ApJ, 550, 212Bell E. F., McIntosh D. H., Katz N., Weinberg M. D., 2003, ApJS,149, 289Bertin E., Arnouts S., 1996, A&AS, 117, 393Blanton M. R., Roweis S., 2007, AJ, 133, 734Brandl B. R., et al., 2004, ApJS, 154, 188Bregman J., Temi P., 2005, ApJ, 621, 831Bregman J. N., Snider B. A., Grego L., Cox C. V., 1998, ApJ,499, 670Bregman J. N., Temi P., Bregman J. D., 2006, ApJ, 647, 265Bressan A., Granato G. L., Silva L., 1998, A&A, 332, 135Bressan A., Marigo P., Girardi L., Salasnich B., Dal Cero C.,Rubele S., Nanni A., 2012, MNRAS, 427, 127Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Caldwell N., 1984, PASP, 96, 287Canizares C. R., Fabbiano G., Trinchieri G., 1987, ApJ, 312, 503Cappellari M., 2008, MNRAS, 390, 71Cappellari M., et al., 2011, MNRAS, 413, 813Cappellari M., et al., 2012, Nature, 484, 485Cappellari M., et al., 2013, MNRAS, 432, 1709MNRAS000 , 000–000 (0000) Simonian & Martini
Cassarà L. P., Piovan L., Weiss A., Salaris M., Chiosi C., 2013,MNRAS, 436, 2824Chabrier G., 2003, PASP, 115, 763Cluver M. E., et al., 2014, ApJ, 782, 90Conroy C., 2013, ARA&A, 51, 393Conroy C., Gunn J. E., 2010, ApJ, 712, 833Conroy C., Gunn J. E., White M., 2009, ApJ, 699, 486Davis T. A., et al., 2014, MNRAS, 444, 3427DeFrees D. J., Miller M. D., Talbi D., Pauzat F., Ellinger Y.,1993, ApJ, 408, 530Diamond-Stanic A. M., Rieke G. H., 2010, ApJ, 724, 140Draine B. T., Li A., 2001, ApJ, 551, 807Draine B. T., Li A., 2007, ApJ, 657, 810Draine B. T., Salpeter E. E., 1979, ApJ, 231, 438Duley W. W., Williams D. A., 1981, MNRAS, 196, 269Dwek E., Arendt R. G., 1992, ARA&A, 30, 11Ebneter K., Davis M., Djorgovski S., 1988, AJ, 95, 422Ferrari F., Pastoriza M. G., Macchetto F. D., Bonatto C., PanagiaN., Sparks W. B., 2002, A&A, 389, 355Ford H. A., Bregman J. N., 2013, ApJ, 770, 137Forman W., Jones C., Tucker W., 1985, ApJ, 293, 102Galliano F., Madden S. C., Tielens A. G. G. M., Peeters E., JonesA. P., 2008, ApJ, 679, 310Gil de Paz A., et al., 2007, ApJS, 173, 185Girardi L., et al., 2010, ApJ, 724, 1030Goldreich P., Scoville N., 1976, ApJ, 205, 144Goudfrooij P., de Jong T., 1995, A&A, 298, 784Goudfrooij P., de Jong T., Hansen L., Norgaard-Nielsen H. U.,1994, MNRAS, 271, 833Groves B., et al., 2012, MNRAS, 426, 892Helou G., 1986, ApJ, 311, L33Helou G., Lu N. Y., Werner M. W., Malhotra S., Silbermann N.,2000, ApJ, 532, L21Ho L. C., Filippenko A. V., Sargent W. L. W., 1997, ApJS,112, 315Hony S., Van Kerckhoven C., Peeters E., Tielens A. G. G. M.,Hudgins D. M., Allamandola L. J., 2001, A&A, 370, 1030Hudgins D. M., Allamandola L. J., 1995,The Journal of Physical Chemistry, 99, 3033Hunter J. D., 2007, Computing In Science & Engineering, 9, 90Indebetouw R., et al., 2005, ApJ, 619, 931Jarrett T. H., et al., 2011, ApJ, 735, 112Jarrett T. H., et al., 2013, AJ, 145, 6Joblin C., Tielens A. G. G. M., Geballe T. R., Wooden D. H.,1996, ApJ, 460, L119Jochims H. W., Ruhl E., Baumgartel H., Tobita S., Leach S.,1994, ApJ, 420, 307Jones A. P., Tielens A. G. G. M., Hollenbach D. J., McKee C. F.,1994, ApJ, 433, 797Jones E., Oliphant T., Peterson P., et al., 2001–, SciPy: Opensource scientific tools for Python,
Jura M., Kim D. W., Knapp G. R., Guhathakurta P., 1987, ApJ,312, L11Kaneda H., Onaka T., Sakon I., 2005, ApJ, 632, L83Kaneda H., Onaka T., Sakon I., Kitayama T., Okada Y., SuzukiT., 2008, ApJ, 684, 270Kaviraj S., et al., 2007, ApJS, 173, 619Kelson D. D., Holden B. P., 2010, ApJ, 713, L28Kennicutt Jr. R. C., 1998a, ARA&A, 36, 189Kennicutt Jr. R. C., 1998b, ApJ, 498, 541Kennicutt Jr. R. C., et al., 2003, PASP, 115, 928Knapp G. R., Turner E. L., Cunniffe P. E., 1985, AJ, 90, 454Knapp G. R., Guhathakurta P., Kim D.-W., Jura M. A., 1989,ApJS, 70, 329Krajnović D., et al., 2011, MNRAS, 414, 2923Langhoff S. R., 1996, The Journal of Physical Chemistry, 100,2819Lonsdale Persson C. J., Helou G., 1987, ApJ, 314, 513 MacArthur L. A., Courteau S., Bell E., Holtzman J. A., 2004,ApJS, 152, 175Madden S. C., Vigroux L., Sauvage M., 1999, in Cox P., KesslerM., eds, ESA Special Publication Vol. 427, The Universe asSeen by ISO. p. 933Maraston C., 2005, MNRAS, 362, 799Maraston C., Daddi E., Renzini A., Cimatti A., Dickinson M.,Papovich C., Pasquali A., Pirzkal N., 2006, ApJ, 652, 85Marigo P., Girardi L., Bressan A., Groenewegen M. A. T., SilvaL., Granato G. L., 2008, A&A, 482, 883Marigo P., Bressan A., Nanni A., Girardi L., Pumo M. L., 2013,MNRAS, 434, 488Martin D. C., et al., 2005, ApJ, 619, L1Martini P., Dicken D., Storchi-Bergmann T., 2013, ApJ, 766, 121McDermid R. M., et al., 2015, MNRAS, 448, 3484Meidt S. E., et al., 2012, ApJ, 744, 17Meidt S. E., et al., 2014, ApJ, 788, 144Melbourne J., Boyer M. L., 2013, ApJ, 764, 30Melbourne J., et al., 2012, ApJ, 748, 47Micelotta E. R., Jones A. P., Tielens A. G. G. M., 2010, A&A,510, A36Morrissey P., et al., 2005, ApJ, 619, L7Morrissey P., et al., 2007, ApJS, 173, 682Navarro J. F., Frenk C. S., White S. D. M., 1996, ApJ, 462, 563Norris M. A., Meidt S., Van de Ven G., Schinnerer E., Groves B.,Querejeta M., 2014, ApJ, 797, 55O’Dowd M. J., et al., 2009, ApJ, 705, 885Ogle P., Antonucci R., Appleton P. N., Whysong D., 2007, ApJ,668, 699Pahre M. A., Ashby M. L. N., Fazio G. G., Willner S. P., 2004,ApJS, 154, 229Panuzzo P., Rampazzo R., Bressan A., Vega O., Annibali F.,Buson L. M., Clemens M. S., Zeilinger W. W., 2011, A&A,528, A10Peletier R. F., et al., 2012, MNRAS, 419, 2031Pellegrini S., 2010, ApJ, 717, 640Pérez F., Granger B. E., 2007,Computing in Science and Engineering, 9, 21Phillips M. M., Jenkins C. R., Dopita M. A., Sadler E. M., BinetteL., 1986, AJ, 91, 1062Planck Collaboration et al., 2014, A&A, 571, A1Rampazzo R., Panuzzo P., Vega O., Marino A., Bressan A.,Clemens M. S., 2013, MNRAS, 432, 374Rosenfield P., et al., 2014, ApJ, 790, 22Rosenfield P., Marigo P., Girardi L., Dalcanton J. J., Bressan A.,Williams B. F., Dolphin A., 2016, ApJ, 822, 73Roussel H., et al., 2007, ApJ, 669, 959Sadler E. M., 1987, in de Zeeuw P. T., ed., IAU Symposium Vol.127, Structure and Dynamics of Elliptical Galaxies. pp 125–132Sadler E. M., Gerhard O. E., 1985, MNRAS, 214, 177Salim S., et al., 2007, ApJS, 173, 267Salpeter E. E., 1955, ApJ, 121, 161Salpeter E. E., 1974a, ApJ, 193, 579Salpeter E. E., 1974b, ApJ, 193, 585Sarzi M., et al., 2010, MNRAS, 402, 2187Schlegel D. J., Finkbeiner D. P., Davis M., 1998, ApJ, 500, 525Schutte W. A., Tielens A. G. G. M., Allamandola L. J., 1993,ApJ, 415, 397Scott N., et al., 2013, MNRAS, 432, 1894Serra P., et al., 2012, MNRAS, 422, 1835Shapiro K. L., et al., 2010, MNRAS, 402, 2140Silva L., Granato G. L., Bressan A., Danese L., 1998, ApJ,509, 103Smith B. J., Struck C., Hancock M., Appleton P. N., Charman-daris V., Reach W. T., 2007, AJ, 133, 791Smith M. W. L., et al., 2012, ApJ, 748, 123MNRAS , 000–000 (0000)
TGs in GALEX and WISE Sparks W. B., Wall J. V., Thorne D. J., Jorden P. R., van BredaI. G., Rudd P. J., Jorgensen H. E., 1985, MNRAS, 217, 87Szczepanski J., Vala M., 1993, ApJ, 414, 646Taylor E. N., et al., 2011, MNRAS, 418, 1587Temi P., Brighenti F., Mathews W. G., 2009, ApJ, 695, 1Trams N. R., et al., 1999, A&A, 346, 843Vega O., et al., 2010, ApJ, 721, 1090Veron-Cetty M.-P., Veron P., 1988, A&A, 204, 28Villaume A., Conroy C., Johnson B. D., 2015, ApJ, 806, 82Wardle M., Knapp G. R., 1986, AJ, 91, 23Weingartner J. C., Draine B. T., 2001, ApJS, 134, 263Wright E. L., et al., 2010, AJ, 140, 1868Xilouris E. M., Madden S. C., Galliano F., Vigroux L., SauvageM., 2004, A&A, 416, 41Yi S. K., et al., 2005, ApJ, 619, L111Young L. M., et al., 2011, MNRAS, 414, 940Zibetti S., Charlot S., Rix H.-W., 2009, MNRAS, 400, 1181di Serego Alighieri S., et al., 2013, A&A, 552, A8van Dokkum P. G., Conroy C., 2012, ApJ, 760, 70van Loon J. T., Groenewegen M. A. T., de Koter A., Trams N. R.,Waters L. B. F. M., Zijlstra A. A., Whitelock P. A., Loup C.,1999, A&A, 351, 559MNRAS000