Evidence for Quasar Activity Triggered by Galaxy Mergers in HST Observations of Dust-reddened Quasars
aa r X i v : . [ a s t r o - ph ] S e p To appear in ApJ
Preprint typeset using L A TEX style emulateapj v. 6/22/04
EVIDENCE FOR QUASAR ACTIVITY TRIGGERED BY GALAXY MERGERS IN HST OBSERVATIONS OFDUST-REDDENED QUASARS
Tanya Urrutia , Mark Lacy , Robert H. Becker To appear in ApJ
ABSTRACTWe present Hubble ACS images of thirteen dust reddened Type-1 quasars selected from theFIRST/2MASS Red Quasar Survey. These quasars have high intrinsic luminosities after correctionfor dust obscuration ( − ≥ M B ≥ − Subject headings: quasars: general — galaxies: active, evolution, interactions INTRODUCTION
The existence of a link between AGN and star-bursts/Ultraluminous Infrared Galaxies (ULIRGs) inthe local Universe has been discussed for a longtime (e.g. Heckman et al. (1989); Sanders & Mirabel(1996); Cid Fernandes et al. (2001)). Further evidencefor such a link comes from the tight correlation be-tween nuclear black hole mass and bulge mass (e.gMagorrian et al. (1998); Tremaine et al. (2002)) and theblack hole mass - host galaxy velocity dispersion (M- σ ) relation in local galaxies (e.g. Ferrarese & Merritt(2000); Gebhardt et al. (2000)), which suggest that star-formation and accretion onto black holes in galax-ies are intimately connected. However, HST imagingof the host galaxies of luminous quasars to moderatedepths showed that most of them have the morphol-ogy of elliptical galaxies, with little evidence for ongo-ing star formation (Dunlop et al. 2003). At low red-shift only 30% of the host galaxies show obvious signsof disturbance (Guyon et al. 2006). Although quasarswith infrared excesses are thought to host starbursts(Canalizo & Stockton 2001), these are a small minorityof the optically-selected quasar population.Recent surveys have shown that optically-selectedquasars comprise less than half of the total quasar popu-lation (Mart´ınez-Sansigre et al. 2005; Stern et al. 2005).Optical quasar surveys tend to miss dust-reddenedquasars which have been found either with infrared sur-veys (Cutri et al. 2001; Lacy et al. 2004), radio surveys Department of Physics, University of California, One ShieldsAvenue, Davis, CA 95616; [email protected] IGPP, L-413, Lawrence Livermore National Laboratory, Liv-ermore, CA 94550; [email protected] Spitzer Science Center, MS 314-6, California Institute ofTechnology, 1200 E. California Boulevard, Pasadena, CA 91125;[email protected] (White et al. 2003), hard X-ray surveys (Norman et al.2002) or surveys for high-ionization narrow-line objects(Zakamska et al. 2006). Therefore they represent a newand largely uninvestigated quasar population, which mayhave many members that are at an earlier stage in theirquasar activity in which dust and gas debris from themerger block the view of the central AGN. This fits withpopulation synthesis models of the X-ray backgroundwhich predict that a large fraction (close to 80%) ofthe accretion process in AGN is obscured and only themost energetic photons ( >
20 keV) can penetrate the ob-scuring dust barrier (Gilli et al. 2001; Ueda et al. 2003;Gilli, Comastri & Hasinger 2007)).Dusty quasars are generally divided into Type-2quasars, which show only narrow-line emission in therest-frame optical, and have inferred extinctions towardsthe nucleus of A V ∼ − A V ∼ − Fig. 1.—
Spectra of the FIRST/2MASS red quasars observed with HST. The black line in the quasar spectrum, the blue line the FBQSquasar composite redshifted to the quasar’s redshift and the red line after applying a SMC reddening law to the FBQS composite. from a merger and starburst which triggers the nuclearactivity, prior to quasar winds expelling the dust (e.g.Sanders et al. (1989)). Of particular interest are theclass of lightly reddened quasars, where the reddeningis smaller than that expected from a torus seen edge-on.These objects are good candidates for reddening by dustin the host galaxy, but may also correspond to objects inwhich the line of sight grazes the edge of the torus.Recent simulations of galaxy mergers tend to sup-port the evolution model, and suggest that the activenucleus is obscured for a long time before the feed-back from the accretion disperses the obscuring material(Hopkins et al. 2005). In this picture, the quasar is seenin the optical regime only at the end of its lifetime, whenit is the most luminous, radiating close to the Eddingtonlimit and therefore being capable of expelling the dust(Hopkins et al. 2006). A large population of these young,obscured, underluminous quasars would also account forthe mostly hard X-ray background; however there is stilldebate on the normalization factor of the hard X-ray background, that is the fraction of these sources is stillunknown. Observations of Type-2 quasars discovered inthe Sloan Digital Sky Survey (SDSS), however, tend tosupport the orientation hypothesis, with broad lines seenin polarized light, and host galaxies consistent with thoseof normal Type-1 quasars (Zakamska et al. 2006). Sim-ilarly, an HST study of lightly-reddened Type-1 quasarsselected from 2MASS found no significant difference inthe properties of host galaxies of infrared selected quasarsfrom quasars selected from other methods (Marble et al.2003), though ground-based studies of a similar sample-found more signs of interactions (Hutchings et al. 2003,2006).This paper focuses on HST observations of a sample oflightly dust-reddened Type-1 quasars with 0 . < z < H = 70 km s − Mpc − , Ω Λ =ST Observations of Dust-reddened Quasars 3 TABLE 1Source properties
Radio Flux E( B − V ) LuminositySource z K-mag 1.4GHz (mJy) (rest) Spectrum / K-magF2M0729+3336 0.954 14.5 3.3 0.83 ± ± ± ± ± ± ± ± − ± ± ± ± ± BACKGROUND AND SAMPLE SELECTION
The F2M sample
We have selected a sample of luminous, dust-reddenedquasars using a combination of 2MASS infrared sur-vey (Skrutskie et al. 1995) and the FIRST radio sur-vey (Becker et al. 1995) in order to answer some ofthe questions posed above. The bulk of these dust-reddened Type-1 quasars have reddenings around E ( B − V ) . J − K > . R − K >
4; or R − K > α or Paschen lines. So far, about 100red quasars have been found with this method, whichwould have been missed by traditional optical quasar sur-veys (Glikman et al. 2007). Following their conventions,objects from the FIRST-2MASS red quasars survey arenamed F2M.In contrast to the 2MASS quasar sample ofMarble et al. (2003), which were mostly low redshift,low luminosity and selected by their near-infrared col-ors alone, the red quasars in our FIRST/2MASS studyhave a median redshift of 0.7. We therefore expect theF2M sample to have have higher luminosity and perhapsmore star-formation in the host galaxies than other redquasar samples at lower redshift.While the selection of red quasars based on R − K colors could include objects which are red due to galaxystarlight in the infrared, the objects in this sample areintrinsic high luminosity quasars, in which the galaxylight should add a negligible contribution in the K-band.This becomes evident when we measure the reddening ofthe quasars. Sample spectra
From the F2M sample, we chose a subsample of 130.4 < z < F o ( λ ) = F c ( λ ) 10 E ( B − V ) k ( λ ) (1)with F c ( λ ) the composite (FBQS) and F o ( λ ) theobserved spectrum. Typically the SMC extinctioncurve k ( λ ) lacks a significant 2175 ˚A bump seenin Galactic dust. We follow the conventions fromFitzpatrick & Massa (1990) to obtain the extinctioncurve k ( λ ), with R V = 3 . E ( B − V ) areshown in Table 1; the values have such high errors in E ( B − V ) either because there was low signal to noisein the spectrum or the rms of the fit was not very good.Because there can be significant host galaxy contributionto the optical spectra, especially in the UV, our E ( B − V ) estimates can serve as a lower limit. Glikman et al.(2007) also includes the near-infrared SPEX spectra totheir fitting, so in some cases there is some discrepancybetween their quoted values and ours. Their deductionof the reddening is beyond the scope of this paper, so wewill only take the optical spectrum for an estimate of thereddening of the total system.For most of the quasars the reddening fits are con-sistent with dust-reddening and very little red starlightfrom the host galaxy is required. However, this isnot the case for F2M0825+4716, F2M0915+2418 andF2M1151+5359 in which the the fit for E ( B − V ) brokedown. These objects will be important later when weinspect the properties of their host galaxies. Note alsothat the reddenings are significant enough that the redcolors of the quasars could not solely have come from a“red” spectral slope. Even though the F2M sample is Urrutia et al. Fig. 2.—
Color composites of the 13 red quasars. We use the I-band data for the red part and the g-band for the blue and green parts ofthe composite. The images are 7 ′′ × ′′ . The g ′ -band cuts were set so that it the blue/green part of the image is somewhat more enhancedin comparison to the I c -band data; the actual details from the host galaxies can be seen. The true color images would appear redder thanshown here. For a quasar at redshift of z ∼ radio selected, the radio fluxes of the objects (Table 1)are also weak; the quasars are radio intermediate at best,so a strong synchrotron component showing up in the in-frared is unlikely to have much contribution to the redcolor.Once we have the reddening E ( B − V ), we can alsoderive the intrinsic luminosities of the F2M quasars fromthe spectrum (corrected for obscuration). We also cal-culated the luminosities of the quasars from the K-bandmagnitude assuming that in the K-band there is no host galaxy contribution. The range between the absolutemagnitude derived from the spectrum and from the K-band magnitude sometimes is quite large implying thatthere either were some large slit losses or that even in theK-band there is significant host galaxy contribution, seeTable 1. We are more inclined to use the derived luminos-ity from the K-band, since the slit will miss the obscuredAGN often and as mentioned before, the E ( B − V ) valuesare only a lower limit as there is host galaxy contribu-tion to the spectrum and we don’t have the full wave-ST Observations of Dust-reddened Quasars 5 Fig. 2.— cont. length coverage ranging into the Mid-Infrared to get thetrue continuum shape. The fact that the K-Band mag-nitude is so bright implies that the reddenings must belarger. What we can say is that the quasars are moreluminous than usual, having a luminosity range of − ≥ M B ≥ − M B ∼ − . M B ∼ − . z ∼ . M B we find for our quasars suggests that weare missing the majority of this type of objects. Thus,because of the shallowness of the 2MASS survey we areonly able to probe the tip of the “red quasar iceberg”. HUBBLE ACS OBSERVATIONS
We followed up this subsample of 13 F2M red quasarswith the ACS Wide Field Camera on HST (GO-10412).They were imaged with the F475W and the F814W fil-ters, which roughly correspond to g ′ and I c filters. Theredshift range was chosen such that all our objects hadtheir intrinsic luminosities well above the quasar/Seyfertdivide (so that the near-IR magnitudes used in their se-lection would not have much contribution from the hostgalaxies), and all were at low enough redshift that emis- sion above rest-frame 4000 ˚A fell within the bandpassof the F814W filter on ACS. The observing dates andcorresponding exposure times are presented in Table 2.Most of the objects were observed for one HST or-bit. However, as the sample of red quasars spanned asignificant range of redshift, two high redshift quasars( z > .
9) were observed for two orbits each to ensure amore uniform surface brightness limit for the sample as awhole. These images were drizzled with the python pro-gram multidrizzle in the usual manner. For the imageswith only one orbit we used the readily available driz-zled products from the archive, since they did not havesignificant cosmic ray contamination.Color composite images of the red quasars are shown inFigure 2. Inspecting the images by eye, one can see muchmore information of the host galaxy than one would froma blue quasar in which the quasar light dominates. How-ever, the images reveal much more interaction than theusual fraction of 30% (Marble et al. 2003; Guyon et al.2006) with tidal tails and many irregularities present inthe host galaxies.Some of the images show several compact knots or nu-clei. If the merger was driven by two galaxies with centralblack holes, it is possible that the two of them ignite asAGN, before merging into a massive active galaxy withonly one black hole in the center. This has been ob- Urrutia et al.
Fig. 2.— cont.
TABLE 2Journal of HST observations
Exposure TimeSource Obs. Date red (s) / blue (s)F2M0729+3336 Oct 24 2005 4650 / 4760F2M0825+4716 Oct 19 2005 1820 / 1924F2M0830+3759 Apr 18 2005 1720 / 1832F2M0834+3506 Apr 18 2005 1620 / 1720F2M0841+3604 Apr 24 2005 1600 / 1716F2M0915+2418 Jun 05 2005 1672 / 1776F2M1012+2825 Jun 13 2005 2190 / 2250F2M1113+1244 May 26 2005 1660 / 1760F2M1118 − served in NGC6240, an Ultraluminous Infrared Galaxyat very low redshift (Komossa et al. 2003). Table 4 sum-marizes the properties of the quasars that show morethan one candidate nucleus. F2M0841+3604, which hasthe largest angular separation, was observed with Chan-dra, but only had 7 (very hard) counts, which is notenough to resolve both components. Two of the objectsalso have double-peaked broad lines, F2M0825+4716 andF2M1507+3129. Such emission lines are not uncommon in broad-line radio galaxies, and are thought to be due toa disk-like BLR seen edge-on (e.g. Eracleous & Halpern(1994)), however, they could also be from two separateactive nuclei close to merger. Unfortunately, neither ofthese show multiple nuclei in their images.We then performed photometry measurements on theimages in the two bands. For this we used Sextrac-tor (Bertin & Arnouts 1996) with parameters set as de-scribed in Ben´ıtez et al. (2004). We used the AB mag-nitude zeropoints from the ACS webpage. We extractedtwo magnitudes this way: first the isophotal correctedmagnitude (assuming a symmetric Gaussian profile forthe object) and then a 3” aperture magnitude. A largedifference between the two may indicate that a lot of thelight comes from a galaxy component outside the nu-cleus. We also compared the magnitudes with the SDSSDR5 magnitudes. All quasars, except F2M0729+3336had SDSS imaging. Whenever the HST and SDSS mag-nitudes differ greatly, it means that the SDSS wasn’t ableto separate different components from the actual quasardue to its poor resolution. Table 3 presents the magni-tudes.The astrometry of ACS often has errors of up to 1.5arcseconds, so we shifted the WCS of the images us-ing SDSS stars in the field as references. This greatlyimproved our astrometry, to within a radial RMS er-ror of 0.1 arcseconds. We then compared these newlycalibrated images with the FIRST data to confirm thatST Observations of Dust-reddened Quasars 7 TABLE 3Comparison of total magnitudes mag I c mag g ′ mag I c mag g ′ mag i mag g Source Aperture (3”) Aperture (3”) Isophot Isophot SDSS (petro) SDSS (petro)F2M0729+3336 19.17 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± a ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± − ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± b ± ± ± ± ± ± ± ± ± ± ± ± Note . — a HST magnitude quoted for brightest component. The different components are not resolved in SDSS. b HST g ′ -band magnitute quoted for brightest component. The different components are not resolved in SDSS or HSTI-band. TABLE 4Sources with multiple components
Source Comp. Separation Comments0841+3604 2 1.70”/9.2 kpc Wide separation1012+2825 2 0.15”/1.2 kpc Resolved in g ′ -band1118 − the radio emission comes from the central nucleus or,in some cases, locate the likely position of the nucleus.The FIRST and SDSS surveys are fit to the same as-trometric system and both have position errors lessthan 0.2 arcseconds. In most cases the radio emissionis associated with the brightest central optical compo-nent. For F2M1656+3821, the radio emission comesfrom the brightest of the three components. Yet forF2M0841+3604 the radio emission emanates from thecenter between the two brightest points. The lower σ contours are skewed a little towards the brightest source.It is not clear where the radio emission is coming fromin this source. ANALYSIS AND RESULTS
Properties of the Quasars
We performed quasar point-source subtraction andhost galaxy fitting for the red quasars. We did per-form the subtraction for the most irregular merging sys-tems in our sample (F2M0841+3604, F2M1113+1244,F2M1532+2415, F2M1656+3821); however, thesesources have multiple components or are extremely ir-regular, so the host galaxy fitting will likely be wrong.For the fitting and subtracting we used our own IDL pro-gram fithost . It is based on the 2D fitting technique ofMcLure et al. (2000) and has been successfully appliedto ground-based AO data (Lacy et al. 2002). The pro-gram simultaneously fits and subtracts both PSF andhost galaxy; the host galaxy fitting is described in sec- tion 4.2.PSF stars were observed at the end of the HST orbits,immediately after the quasars. Stars were chosen to beclose to the quasars, so as to avoid extra overhead due toguide star acquisition, and to be bright enough to havea similar fluence to the quasar in short exposures. Ob-servations were made with the stars as close as possibleto the same position as the quasars on the detector, andwith the same dither offsets, The only drawback to thistechnique was that the ACS 1K subarray had to be usedfor both the quasars and the guide stars to avoid exces-sive readout overhead. Since we were mostly interestedin the quasars and their close companions this was not amajor problem.The fithost program was used in a mode where it au-tomatically tries to fit the PSF-function to the brightestpoint in the image and then scale and remove it, so thatthe host galaxy can be subsequently fit. The center ofthe quasar and the host galaxy need not necessarily co-incide. Starting parameters for the fithost program werederived using the IRAF program imexam and then re-fined with a second run. The first 2 postage stamp im-ages in Figure 3 show the red quasars before and afterPSF-subtraction. For the sake of brevity we chose toonly show the I c -band results, since they have most con-tribution from the quasar and most incident flux. In the I c -band the flux from the PSF is typically similar to thatof the host galaxies, but concentrated in the center of theimage, which is why in Figure 3 we don’t see glaring dif-ferences in the PSF-subtracted images from the original.In the g ′ -band, the quasar is even fainter. The magni-tude of the PSF/quasar after fitting is found in the firsttwo columns of Table 5The first result we can derive from the PSF fitting ofthe HST images is that the quasars are actually moreobscured than deduced from the spectrum. In the HST g ′ -band, which corresponds to rest-frame UV, we almostonly see host galaxy contribution with the quasar beingalmost entirely extinguished. The g ′ - I c colors the fittedquasar PSFs are also redder than for the total system.While the mean SDSS g’-i’ color for the total system is Urrutia et al. Fig. 3.— I c -band results of PSF and Host Galaxy fitting. The first column shows an original postage stamp of the red quasar image, thesecond column the PSF-subtracted image, the third the best-fit elliptical model, the fourth the residual after subtracting the model andthe fifth the radial surface brightness profile with the solid lines representing the elliptical fit. ST Observations of Dust-reddened Quasars 9
Fig. 3.— cont.
Fig. 3.— cont. g ′ - I c color for the quasar (just the PSF)is 0.61 magnitudes redder at 2.53.We then calculated the reddening for only the quasarby reddening the slope of the FBQS composite (account-ing for emission lines that might have fallen within thepassbands) to fit the g ′ - I c colors made by the PSF us-ing the reddening curve described in section 2.1. Mostof the objects which already had large E ( B − V ) red-denings around 1.0 increase their reddenings and colors,while the objects where the reddening of the total systemwere close to 0.5 on average don’t increase or decreasetheir reddening. Interestingly, the objects for which thefit of the reddened quasar composite broke down, mostlydecrease their reddenings, implying that the host galaxyitself has enough starlight that is responsible for some ofthe red colors of our objects. There could also be scat-tering of the nuclear light into the host galaxy itself, butwe assume that the deduced PSF magnitude is the totalquasar magnitude, since we can’t assess the amount ofscattering from the optical data alone.Figure 4 shows the shifts of the objects colors and red-denings. The total magnitudes and reddenings are rep-resented by the filled circles and the stars represent thequasar/PSF. Two of the quasars (F2M0841+3604 andF2M1656+3821) were not resolved in the SDSS, so the PSF magnitudes are much fainter than implied by SDSS.We mark this objects in Figure 4 in green color, theirshifts are likely to be wrong, but we include them in theFigure for completeness.From the g ′ - I c colors, magnitudes and reddenings, wecan then calculate the true luminosity for the quasar,since the absolute magnitudes calculated in section 2.1from the spectrum had contamination from the hostgalaxy, especially in the bluer passband. Table 5 givesthe new luminosities (corrected for reddening) deter-mined from the quasar colors and magnitudes. Thequasars are a bit more luminous on average, due to thefact that the quasar is more obscured than had beendeduced from the spectrum. The quasars on averageare still well above the quasar/Seyfert divide and muchmore luminous than the red quasar study conducted byMarble et al. (2003), so our claim that we are missing alarge population of obscured AGN is still valid. Properties of the Host Galaxies
Before modelling the full host/quasar system, we ob-tained a constraint on the host galaxy magnitude. Wedid this by constraining the scaled PSF subtraction sothat the residual was approximately flat in the centerST Observations of Dust-reddened Quasars 11
TABLE 5Quasar (PSF) properties
Source PSF magnitudes g-I colors E ( B − V ) Luminosities ( M B )Mag I c Mag g ′ Total PSF Total PSF Total PSF(SDSS) (HST) (spectrum) (HST) spectrum HSTF2M0729+3336 20.15 ± ± a ± ± − − ± ± ± ± − − ± ± ± ± − − ± ± ± ± − − ± ± ± ± − − b F2M0915+2418 19.88 ± ± ± ± − − ± ± ± ± − − b F2M1113+1244 19.11 ± ± ± ± − − − ± ± ± ± − − ± ± ± ± − − ± ± ± ± − − ± ± ± ± − − b F2M1656+3821 23.46 ± ± ± ± − − b Note . — a No SDSS photometry, we use HST values. b Luminosity is only for one nucleus.
Fig. 4.— g ′ - I c color vs. E ( B − V ) reddening. The filled dotsrepresent the magnitudes and reddenings for the total system. Themagnitudes are from SDSS data and don’t differ much from HSTphotometry and the E ( B − V ) reddening is from spectral fitting.The open stars represent quasar (PSF) color and reddenings de-rived from HST PSF photometry. We included lines to show howthe values shifted. On average the quasar has a 0.61 mag reddercolor and higher E ( B − V ) of 0.085 mag. of the quasar host within a 3 pixel radius and declinedmonotonically outside that radius. We also obtained thelower limit of the host galaxy magnitude by subtractingthe PSF until the central pixel was zero. The “mono”and “zero” magnitude are found in Table 6.The host galaxies were modeled by fitting PSF plusgalaxy model profiles (convolved by the PSF) by mini-mizing χ . Close to the center of the source, systematicerrors from the PSF subtraction dominate, so to preventthose errors to dominate the fit, the inner 3 pixel radiuswas downweighted by a factor of 0.5 to ensure that the χ surface was fairly uniform across the fitting aperture. The position of the PSF and the galaxy nucleus was notheld fixed; so in total we fit the flux, angle, axial ratioand position of the host galaxy to the image. We triedto fit both elliptical and exponential profiles to the im-ages, however in all, but one case (F2M1118 − Fig. 5.—
Host galaxies with artificial PSF added. The images show different nucleus to host ratios in increasing order. When the nucleushas the same magnitude as the host galaxy, which is the case for the red quasars in this sample, the interactions are clearly visible, suchas the tidal connection in F2M1118 − with the quasar. This quasar light could be either scat-tered continuum, or extended line emission. The lack ofa clear ionization cone morphology (with the possible ex-ception of F2M0830+3759), however, would suggest thatmost of this blue light is from young stars.We also measured the nucleus to host (N/H) ratiosfrom these magnitudes (Table 6 for the I c -band N/H). Asdiscussed in section 4.1, the N/H for the g ′ -band will belower, because the quasar is more extinguished and mostof the light in that band will come from the host. Overall,the luminosity of the hosts are on average, a little largerthan those of the nuclei, such as the sample of IR-excessquasar sample of Surace, Sanders & Evans (2001). Therelatively low N/H are in sharp contrast to the high N/Hratios of the luminous quasar sample of McLure et al. (1999), which is tied to the Dunlop et al. (2003) sam-ple. However the N/H ratios of our quasars are gener-ally higher than the red quasar sample of Marble et al.(2003).We then calculated the absolute magnitudes ( M B ) ofthe hosts adopting a 1 Gyr post starburst model fromBruzual & Charlot (2003) for the K-correction. At thetypical redshifts of our sample (z ∼ M ∗ B = − . L ∗ luminosity.One measure of the extent of the host galaxy is thePetrosian radius r p (Petrosian 1976). Very irregular orinteracting systems tend to have high Petrosian radii,while compact ellipticals have a low r p . The Petrosianradius is defined when the ratio of he surface brightnessat the Petrosian radius and the average surface bright-ST Observations of Dust-reddened Quasars 13ness at radii below reach a certain value η : η = µ ( r p )¯ µ ( r < r p ) (2)Following SDSS conventions, we set η = 0 .
2. Ta-ble 6 quotes the Petrosian radii of our systems. Withone exception (the very compact F2M0729+3336), mosthave Petrosian radii around 1”-2.5”, which are normalfor galaxies at those redshifts. When comparing thosenumbers to the surface density plots in Figure 3, onecan notice that the nucleus ( r < r p ) of the host galaxyoften has an elliptical profile; only the most irregular sys-tems such as F2M0841+3604 are irregular at low radii.Mostly, the interaction features only appear well beyondthe Petrosian radius and usually have low surface bright-ness.This could be an indication why many authors con-clude that the host galaxies of luminous quasars arefit best mostly by elliptical profiles (e.g. (Floyd et al.2004)), but find so few merger remnants. Perhaps withthe overpowering AGN (high N/H ratios), the low sur-face brightness features or interaction morphology signa-tures are lost. Recently, very deep imaging of some ofthe quasars studied by Dunlop et al. (2003), show eitherlow surface brightness tidal tails or other merger rem-nants in the form of shells (Bennert et al. 2006) furthersupporting this point. The red quasars in this studyare quite extinguished, so the host galaxy features areeasier to discern. We tested how the tidal features candisappear in the presence of a bright quasar by addingpoint sources of different brightness (1,3,10 N/H ratios)to the host galaxy image. When a PSF with a N/H ratioof 10 is added, the quasar is so bright that in all caseswith exception of F2M0841+3604 the red quasars losemost of their interaction features and the quasar dom-inates the image (Figure 5). Even so, after performingPSF subtraction on a N/H=10 quasar+host system, theinteractions should reappear. Also the red quasar hostgalaxies are in an earlier stage of the merger as we willsee next.As we already commented in section 3 the fraction ofour red quasars showing interaction in their host galax-ies is very high. Furthermore, all of the dust reddenedquasars, that is the quasars where the dust reddeningtemplate fit best, show interaction. E(B-V) seems to cor-relate weakly with the amount of interaction; the moreobscured the quasar is, the more disturbed the morphol-ogy of the host galaxy is. We can arrive at these conclu-sions more or less “by eye”, but we support our claimsby parameterizing the host galaxy morphology with theirGini coefficients and their Concentration indices.The Gini coefficient is a non-parametric approach toclassifying a galaxy, and can therefore be easily ap-plied to irregular galaxies or highly merging systems.It is a measure of the cumulative distribution of agalaxy’s pixel values and is a good alternative approachto quantifying the amount of interactions in galaxies.The Gini coefficient is correlated with the concentra-tion index of galaxies for spiral and elliptical galax-ies (Abraham, van den Bergh & Nair 2003) and anti-correlated with the Concentration Index for ULIRGs and other highly irregular systems (Lotz, Primack & Madau2004). A high Gini coefficient therefore indicates eithera high level of interactions or a highly concentrated ellip-tical galaxy, while a low Gini coefficient indicates eitherspiral galaxies or low surface brightness galaxies.We calculated the Gini coefficient of the host galax-ies of the red quasars by following the conventions inAbraham, van den Bergh & Nair (2003). Based on theformalism by Glasser et al. (1962), if we sort the pixels X i ’s flux into increasing order, the Gini coefficient canbe calculated by: G = 1¯ Xn ( n − n X i (2 i − n − X i (3)However, we took notice of the warning byLotz, Primack & Madau (2004) and tried to create seg-mentation maps at the µ ( r p ), the flux threshold abovewhich pixels are assigned to the galaxy. For that wecut out galaxy postage stamps cutouts by eye and in-cluded only pixels within that cutout that had a surfacebrightness higher than µ ( r p ). The Gini coefficient wasthen computed of the distribution of absolute flux valuesof the pixels which corrects for the areas in the centerwhich were PSF-oversubtracted. The Gini coefficients Gcan be found in Table 6.The Concentration Index measures the ratio of the fluxwithing an inner radius to that within and outer circularor elliptical aperture. The main difference in the defi-nitions by different authors is the choice of the radii orsemimajor axes of the two apertures. Conselice et al.(2003) adopts a ratio between the radii containing cer-tain percentages of the total light. Abraham et al. (1994)uses a flux ratio between two normalized radii α and 1,where E(1) is the area encompassing 2 σ flux and α istypically set to 0.3. C = P P i,j ∈ E ( α ) I ij P P i,j ∈ E (1) I ij (4)Since Abraham, van den Bergh & Nair (2003) has al-ready linked the Gini coefficient to the Concentrationindex via a unity slope, we decided to use the Con-centration index defined in equation 4 instead of theConselice et al. (2003) definition. We chose the totalflux as the flux within 1.5 r p and E( α ) as the flux within0.45 r p . From that ratio we are able to obtain C, whichis quoted in Table 6.Figure 6 shows the Gini coefficient plotted againstthe E ( B − V ) reddening derived from the HST quasarmagnitudes on the left side and the Concentration In-dex over the same E ( B − V ) on the right side. Whilethere is a weak correlation between the Gini coefficientand the reddening, we find no correlation between theConcentration Index and E ( B − V ). As mentioned be-fore, high Gini coefficient and low Concentration indicesare an indication for interaction or merger, since thebrightest flux pixels are highly concentrated in a fewpixels (high G), but those pixels are not in the cen-ter (low C). This seems to be the case in the systemswith the highest E ( B − V ) in our sample, while thehosts of quasars with lower E ( B − V ) tend to haveboth Gini coefficients and Concentration indices moreconsistent with those of normal, undisturbed galaxies4 Urrutia et al. TABLE 6Host galaxy properties
Source Host mag I c Host mag g ′ Luminosities Morphologieszero mono total zero mono total N/H a M B L ∗ r p (pix) G CF2M0729+3336 20.71 19.97 19.78 ± ± − ± ± − ± ± − ± ± − ± ± − ± ± − ± ± − ± ± − − ± ± − ± ± − ± ± − ± ± − ± ± − Note . — a Nucleus to host ratios from I c -Band magnitudes.Zero and monoton are model magnitudes, therefore they don’t have errors. Abraham, van den Bergh & Nair (2003). On the lowerpanel of Figure 6 the Concentration Index is plottedagainst the Gini coefficient, with the unity slope shown.Only the two undisturbed galaxies (F2M0834+3506 andF2M1151+5359) are above this slope, with the rest hav-ing high Gini coefficients in relation to their Concentra-tion Indices, again implying high interaction features.While we could already see those results “by eye”,these galaxy parameters are affirmation to the result thatthe higher the reddening in the quasar, the higher thechance of interaction in the quasar host. Furthermore,if the reddening in the spectrum fits a dust reddeningtemplate well, it is also a good indicator for evidence ofinteraction. Nonetheless, these results are based only on13 highly heterogenous systems, so further observationsare needed to improve number statistics and to confirmthis claim. NOTES ON INDIVIDUAL QUASARS
While most of the quasar host galaxies show some de-gree of irregularity and evidence for merger, it is worthit to look at the quasars on an individual basis to furtherinspect some peculiarities in some of the objects.
F2M0729 + This is the only quasar for which SDSS information wasnot available. Even though we detected only 6 photons inthe X-ray observations, all of them are in the hard band,giving this source a hardness ratio of 1.0 and thereforeimplying a high column density.The spectrum has strong Ca H+K absorption lines,but the spectral reddening does fit a dust extinction lawquite well longward of about 5000 ˚A , so the host galaxycontribution to the red optical light in the spectrum isnot going to make a large impact.This particular system has a very compact ellipticalhost, with a very small Petrosian radius and large Ginicoefficient. Beyond that however, large tidal tails extendout to over 20 kpc on both sides of the host, which isevidence for a recent merger event. This makes the to-tal size of the host galaxy over 35 kpc. In the g ′ -band,the quasar (PSF) contribution is almost non-detectable, making this the system with the largest shift in g ′ - I c color of the PSF relative to the total emission from thehost. F2M0825 + This looks like a Type-2 quasar in the optical spec-trum. The color of the quasar (PSF) would suggest thatalso. However, the Spex IR spectrum shows broad quasarlines, however and a much redder spectral slope than im-plied by the optical continuum,which is probably dom-inated by host galaxy emission. The image shows thatwhile the central component is fit well by an ellipticalgalaxy, there is a tidal bridge linking the central galaxyto a companion nearby.The optical spectrum shows double-peaked narrowemission lines, separated by about 600km s − . Thehigher redshift set has relatively strong Balmer lines and[O ii ]3727 emission at z = 0 . z = 0 .
800 is of higher ioniza-tion and probably corresponds to a high ionization out-flow.
F2M0830 + This quasar has the lowest redshift in our sample,therefore features close to the nucleus will be more easilyobservable. The Hubble image in fact shows a lot of ir-regularity near the nucleus with shells of material alongthe major axis, either side of the nucleus. They almostcancel each other out in the radial profile plot, resultingin a deceptively smooth radial profile, while in reality thehost galaxy is very disturbed.F2M0830+3759 was also observed with Chandra andhas a very high X-ray flux (Urrutia et al. 2005). TheX-ray spectrum shows a very broad Iron K α line, hint-ing that we are looking into the broad line region ofthe quasar. It has a moderate absorption, but a higherthan usual spectral slope (Γ ≈ . Fig. 6.—
Upper left graphic: Gini coefficient vs. reddening. There is a weak correlation between the Gini coefficient and the reddening(dashed line). High Gini coefficient indicate either highly irregular systems, such as ULIRGs or very compact ellipticals. Upper rightgraphic: Concentration index vs. reddening. For single systems the Gini coefficient and the Concentration index are correlated. The factthat for high reddening we don’t necessarily have high Concentration indices means that the systems are highly disturbed, usually withmore than one nucleus. Lower graphic: Concentration index vs. Gini coefficient. No correlation is found in our sample in contrast to otherstudies of single galaxies, indicating that a many of the red quasar hosts are in a merger status. The dotted line is the unity correlationfound by Abraham, van den Bergh & Nair (2003). All but two host galaxies (the two undisturbed ones) of the F2M red quasars havehigher Gini coefficients than implied from their Concentration Indices.
F2M0834 + The red quasar system has almost no evidence for in-teraction. The profile could fit an addition of ellipticalprofiles, but the morphology could also represent mergerremnant shells. However, there are no extended featuresor tidal tails.It is unclear whether the blue component just to theNorth of the quasar is associated with it, or whether it isan unrelated object close to the line of sight. There arebroad H α and H β emission lines ∼ − bluewardof the quasar’s redshift, which could be associated withthat component and/or with the quasar. F2M0841 + This object shows the most irregular morphology inthe host galaxy of the Hubble images. While the optical image clearly shows two bright nuclei, of which either orboth could be the quasar, the radio images (from FIRSTand also 6cm VLA A-array observations) show a steepspectrum point source right in between the two brightoptical components. This raises the question of wherethe active nucleus actually is.In addition to the HST and VLA images,F2M0841+3604 has been observed with Chandra(Urrutia et al. 2005). The Chandra photons are veryhard. Unfortunately, the X-ray image had only 7 counts,so the position error on where the X-ray emission comesfrom is quite large. However, the Chandra photons ap-pear to be spread out, so the X-ray emission could comefrom both nuclei. Overall, this is our most spectacularexample of a young evolutionary merger state.
F2M0915 + tail is seen extending to thEast indication a recent interaction. F2M1012 + This red quasar displays two nuclei only 0.15” (1.2 kpc)apart, which would have been missed in ground-basedobservations. The two nuclei are point-like and espe-cially easy to detect in the g ′ -band, but the nucleus ap-pears very “smeared” in the I c -band image. The fittingof quasar and host galaxy is therefore impossible. Thequoted luminosities for the quasar are only for one com-ponent, so the total nuclear luminosity is much higherthan M B = − . F2M1113 + While the position of the quasar is quite certain, be-cause of the clear stellar-like feature, it is noteworthythat the rest of the host galaxy consists of several brightknots, with a low surface brightness tidal tail extendingto 30-35 kpc from the galaxy.
F2M1118 − The image for this red quasar shows it to be a clearmerger with a nucleus and two compact components, onenear the nucleus and the other one at the end of the twotidal tails trailing the system. This is the only systemfor which an exponential profile fit the host galaxy betterthan an elliptical (but only near the nucleus).The optical spectrum for F2M1118 − F2M1151 + F2M1151+5359 is the quasar also fits an elliptical pro-file quite well, but there is still some residual in the hostgalaxy after model subtraction in the I c -band. The blueband fits almost perfectly. So this would be comparableto other “normal”, bright, blue quasars, which have anundisturbed elliptical galaxy as a host. The spectrumshows asymmetric profiles in the emission lines and nar-row MgII absorption lines very close to the quasar red-shift. F2M1507 + This is the highest redshift red quasar imaged withACS. The spectrum fits dust reddening quite well. Alsohere the central component can be well fit by an el-liptical galaxy, but there is a low surface brightnesstidal bridge connecting it to the red companion nearby.F2M1507+3129 is thus a very large (35 kpc) and spec-tacular merger at z ∼ F2M1532 + HST imaging of F2M1532+2415 displays a truly dis-turbed host, with many components and no clear ev-idence of where the quasar is located among the con-stituents of the host galaxy. Radio maps show the syn-chrotron radiation coming from the central red com-ponent. They also show that this quasar is actuallya classical Fanaroff-Riley Type II radio double (FRII,Fanaroff & Riley (1974)), with the outer components be-ing over one arminute away from the central source.The optical spectrum shows a very narrow H β line andno broad MgII line, but in the infrared Pa β is broad. Thequasar host shows extreme interaction. F2M1656 + This optically-faint object shows almost only hostgalaxy emission in the optical part of the spectrum, butin the near-infrared there is a very red continuum and abroad Pa β line. There are three components in the HSTimage, two point-like and one elliptical. The radio fluxpoints to the emission coming from the brightest (SW)optical point source. The morphology of F2M1656+3821could suggest that the system is a gravitational lens withthe lensing galaxy lying between the two point sources.The reddening for this system might stem from the lens-ing galaxy and not the quasar itself, as is implied bythe host galaxy’s very red colors. However, the lack ofany arc-like morphology in the outer optical componentsand the lack of spectral features at lower redshift in thisobject’s spectrum argues against the lensing hypothesis. CONCLUSIONS
We have observed and analyzed a sample of 13highly reddened, luminous quasars from the F2M surveyGlikman et al. (2007) with ACS in both I c - and g ′ -bandto study their host galaxies. The images show interac-tions in 85% of the objects and clear evidence for mergerssuch as tidal tails and multiple nuclei. We fitted modelPSF and galaxy profiles to the images. Within the nu-clear region, most galaxies fit an elliptical profile, in ac-cordance with quasar host studies such as Dunlop et al.(2003), but outside of the nuclear region this fit is nolonger valid as merger features become clear.After performing PSF-fitting, the quasar displays evenredder colors and reddenings than implied from SDSSimaging and ESI-spectra, showing that the optical mag-nitudes of the quasars are significantly contaminatedby the host galaxies in low resolution data. Five ofour galaxies show very dramatic, likely multiple mergermorphologies, reminiscent of the simulations of high- z quasar hosts by Li et al. (2006). There is a widerange in merger phase, with some of our quasar hostsin an apparently early part of the merger sequencewhere the merging galaxies still retain their identity(F2M0841+3604), through to later stages with multi-ple nuclei (F2M1113+1244; F2M1012+2825), and ob-jects at the final merger stage with only a single nucleusand a tidal tail remaining (F2M0915+2418). This wouldbe consistent with quasar activity being triggered rela-tively early in an interaction, and continuing through themerger process.We calculated the Gini coefficients and the Concentra-tion indices of the host galaxies and found a correlation ofthe Gini coefficient with reddening. The systems with theST Observations of Dust-reddened Quasars 17highest quasar E ( B − V ) did not have high Concentra-tion indices consistent with their high Gini coefficients,which implies that the host galaxies have bright, compactcomponents outside of the nuclear region. Using this asa measure of interaction, it seems that the amount ofinteraction is weakly correlated with how obscured thequasar is.Floyd et al. (2004) studied a sample of 17 normalquasar hosts with HST which are closest to our sam-ple in terms of luminosity and redshift. Their sam-ple spanned − < M V < −
28 in quasar luminosityand 0 . < z < .
42 in redshift. Consistent with previ-ous studies, only a small fraction of interactions/mergerswere seen (they have only one object out of 17 with def-inite signatures of an interaction). However, sensitiveobservations of the hosts of unreddened quasars of sim-ilar luminosity are difficult, so it is possible that somesigns of interaction such as faint tidal tails and multiplenuclei close to the quasar will have been missed. Ourcorrelation of reddening with morphology argues againstthis, but our observed correlation is only weak. However,Figure 5 also shows obvious signs of merger even at nu-cleus to host ratios of 10. Another possible source of biasis that, while none of the red quasars is radio-loud, theseobjects are all radio-selected, and fall into the “radio in-termediate” category. The fact that all of our quasars aredisturbed in comparison to, e.g., the Marble et al. (2003)sample, could be a selection effect. White et al. (2006)find that quasars with redder colors than the SDSS com-posite tend to have higher radio-fluxes.These caveats aside, our results may explain why onlyabout 30% of quasars with obvious signs of mergersand interactions have been found (Guyon et al. 2006;Marble et al. 2003). Reddening by dust from the hostgalaxy seems to be the most likely explanation for theredness of the F2M quasars. The host morphologies andcolors are consistent with dusty, merging galaxies. Ifthese objects were red because the line of sight just grazesthe torus we would expect less disturbed host morpholo-gies, in line with those of normal quasars. Host galaxyreddening of the quasar, whilst only mild in our cases (and much less than nuclear reddening by an edge-ontorus) can nevertheless be sufficient to remove the objectsfrom optically-selected quasar samples (or at best renderthem a small minority in such a sample). Given thatthese kind of dust-reddened Type-1 quasars like thesemake up about 30% of the total in mid-IR selected sam-ples, this means the fraction of quasar hosts associatedwith mergers has been significantly underestimated inthe past. Our result is thus consistent with theories inwhich quasars start their lives obscured by dust, and onlyappear in optical or soft X-ray surveys after the dustalong the line of sight to the nucleus has been clearedby quasar winds (Sanders et al. 1989; Sanders & Mirabel1996; Springel et al. 2005; Hopkins et al. 2007).We thank Elinor Gates for assistance with writing the fithost
REFERENCESAbraham, R. G., Valdes, F., Yee, H. K. C., van den Bergh, S. 1994,ApJ, 432, 75Abraham, R. G., van den Bergh, S., Nair, P. 2003, ApJ, 588, 218Antonucci, R. 1993, ARA&A, 31, 473Becker, R.H., White, R. L., Helfand, D. J. 1995, ApJ, 450, 559Bennert, N., Canalizo, G., Jungwiert, B., Stockton, A., Schweizer,F., Lacy, M., Peng, C. 2006, AAS, 209, 251.04Ben´ıtez, N., et al. 2004, ApJS, 150, 1Bertin, E., Arnouts, S. 1996, A&AS, 117, 393Brotherton, M. S., Tran, H. D., Becker, R. H., Gregg, M. D.,Laurent-Muehleisen, S. A., White, R. L. 2001, ApJ, 546, 775Brown, M. J. I., Brand, K., Dey, A., Jannuzi, B. T., Cool, R., LeFloc’h, E., Kochanek, C. S., Armus, L., Bian, C., Higdon, J.,Higdon, S., Papovich, C., Rieke, G., Rieke, M., Smith, J. D.,Soifer, B. T.; Weedman, D. 2006, ApJ, 638, 88Bruzual, G., Charlot, S. 2003, MNRAS, 344, 1000Canalizo, G., Stockton, A. 2001, ApJ, 555, 719Cid Fernandes, R., Heckman, T., Schmitt, H., Delgado, R. M. G.,Storchi-Bergmann, T. 2001, ApJ, 558, 81Conselice, C., Bershady, M., Dickinson, M., Papuvich, C. 2003,ApJ, 126, L1183Cutri, R. M., Nelson, B. O., Kirkpatrick, J. D., Huchra, J. D.,Smith, P. J. 2001, in ASP Conf. Ser. 232, The New Era of WideField Astronomy, ed. R. G. Clowes, A. J. Adamson and G. E.Bromage (San Francisco: ASP), 78 Dunlop, J. S., McLure, R. J., Kukula, M. J., Baum, S. A., O’Dea,C. P., Hughes, D. H. 2003, MNRAS, 340, 1095Eracleous, M., Halpern, J. P. 1994, ApJS, 90, 1Fanaroff, B. L, Riley, J. M. 1974, MNRAS, 167, 31PFitzpatrick, E. L. & Massa, D. 1990, ApJS, 72, 163Fitzpatrick, E. L. 1999, PASP, 111, 63Ferrarese, L., Merritt, David 2000, ApJ, 539, 9Floyd, D. J. E., Kukula, M. J., Dunlop, J. S., McLure, R. J., Miller,L., Percival, W. J., Baum, S. A., O’Dea, C. P. 2004, MNRAS,335, 196Gebhardt, K., et al. 2000, ApJ, 539, 13Gilli, R., Salvati, M., Hasinger, G. 2001, A&A, 366, 407Gilli, R., Comastri, A., Hasinger, G. 2007, A&A, 463, 79Glasser, G. J. 1962, J. Am. Stat. Assoc., 57, 648Glikman, E., Gregg, M. D., Lacy, M., Helfand, D. J., Becker, R.H., White, R. L. 2004, ApJ, 607, 60Glikman, E., Helfand, D. J., White, R.L., Becker, R.H., Gregg,M.D., Lacy, M. 2007, ApJ, in press, astro-ph/0706.3222Guyon, O., Sanders, D. B., Stockton, A. N. 2006, ApJS, 166, 89Heckman, T. M., Blitz, L., Wilson, A. S., Armus, L., Miley, G. K.1989, ApJ, 342, 735Hopkins, P. F., Hernquist, L., Martini, P., Cox, T. J., Robertson,B., Di Matteo, T., Springel, V. 2005, ApJ, 625, 71Hopkins, P. F., Richards, G. T., Hernquist, L. 2007, ApJ, 654, 7318 Urrutia et al.