The Hot Interstellar Medium in Normal Elliptical Galaxies. II. Morphological Evidence for AGN Feedback
aa r X i v : . [ a s t r o - ph ] F e b Draft version November 1, 2018
Preprint typeset using L A TEX style emulateapj v. 11/26/04
THE HOT INTERSTELLAR MEDIUM IN NORMAL ELLIPTICAL GALAXIES. II.MORPHOLOGICAL EVIDENCE FOR AGN FEEDBACK
Steven Diehl and Thomas S. Statler Draft version November 1, 2018
ABSTRACTWe report on the discovery of a new quantitative relationship between X-ray gas morphology andradio and X-ray AGN luminosities in normal elliptical galaxies. This is the second paper in a seriesusing data on 54 objects from the
Chandra public archive and builds on the findings of Paper I, whichdemonstrated that hydrostatic equilibrium in elliptical galaxies holds, at best, only approximatelyand that the shape of the X-ray isophotes is unrelated to the shape of the gravitational potential.Instead, the gas is almost always asymmetrically disturbed. In this paper, we quantify the amount ofasymmetry and study its correlation with other galaxy properties. We also determine radio powersand derive X-ray AGN luminosities for our galaxy sample. We find that the amount of asymmetryin the gas is correlated with both measures of AGN activity, in the sense that the hot gas is moredisturbed in galaxies with higher radio and X-ray AGN luminosities. We find no evidence that galaxydensity has significant effects on gas morphology. We do however find evidence for a correlationbetween gas asymmetry and the presence of hot ambient gas, which we interpret as a signature ofhydrodynamic interactions with an external ambient medium. Surprisingly, the AGN–morphologyconnection persists all the way down to the weakest AGN luminosities in rather X-ray faint galaxies.This is strong morphological evidence that supports the general importance of AGN feedback, evenin normal elliptical galaxies.
Subject headings: galaxies: cooling flows—galaxies: elliptical and lenticular, cD—galaxies: ISM—X-rays: galaxies—X-rays: ISM INTRODUCTION
In the first paper of this series (Diehl & Statler 2007,hereafter Paper I) we conducted a morphological analysisof the hot interstellar medium (ISM) of normal ellipticalgalaxies, derived from archived
Chandra observations of54 objects. This paper introduced a new technique toisolate the diffuse hot gas emission from the contami-nating effects of unresolved point sources, and presenteda gallery of adaptively binned gas-only images. By fit-ting elliptical isophotes to these images, we demonstratedthat the apparent flattening of the X-ray gas is com-pletely uncorrelated with, and often significantly largerthan, that of the starlight. The absence of any correla-tion, at radii inside the optical effective radius, impliesthat the hot gas cannot be in perfect hydrostatic equilib-rium, unless the potential is dominated by a dark mat-ter component completely unrelated to the stellar massdistribution. This is unlikely, since within the effectiveradius gravitational potentials are generally expected tobe stellar-mass dominated (e.g. Mamon & Lokas 2005;Humphrey et al. 2006). Consequently, efforts to measurethe shapes of dark matter halos using X-ray gas isophotesmay be fruitless in most cases (e.g. Buote & Canizares1994; Buote et al. 2002), and radial mass profiles derivedfrom X-ray data (Humphrey et al. 2006; Fukazawa et al.2006) could be in error by factors of order unity.This paper takes up the question left unaddressed by Astrophysical Institute, Department of Physics and Astron-omy, 251B Clippinger Research Laboratories, Ohio University,Athens, OH 45701, USA Theoretical Astrophysics Group T-6, Mailstop B227, LosAlamos National Laboratory, P.O. Box 1663, Los Alamos, NM87545, USA (present address)Electronic address: [email protected], [email protected]
Paper I: what is the origin of the gas flattening, if notgravitational? In Paper I, we already excluded rotationalflattening as the dominant cause. A qualitative analysisof gas morphologies showed that they are often asym-metric and disturbed. In this paper we will quantifythis asymmetry, argue that it and the isophotal ellip-ticity reflect the same underlying disturbance, and showstrong evidence that the root causes are the central ac-tive nucleus—even in those systems with very weak ac-tive galactic nuclei (AGN)—and interactions with ambi-ent external gas.Asymmetries are well known in X-ray studies of galax-ies, groups, and clusters. Chandra and
XMM-Newton ob-servations of clusters of galaxies reveal strong deviationsfrom smooth surface brightness distributions. Many ofthese features are believed to be due to cold fronts,caused by infalling substructure into the cluster center(e.g. Chatzikos et al. 2006), or gas sloshing in the grav-itational potential (e.g. Ascasibar & Markevitch 2006).In other cases, depressions in the surface brightness dis-tribution are found to be coincident with extended radioemission (e.g. Nulsen et al. 2005). These depressions, or“bubbles”, are generally believed to be pockets of low-density extremely hot plasma, inflated by jets poweredby the central AGN (see McNamara & Nulsen 2007, for arecent review). However, there are many cases where thedepressions have no obvious radio counterparts, whichhas brought them the nickname “ghost cavities” (McNa-mara et al. 2001). These cavities are now interpreted tobe relic bubbles that have detached from the radio sourceand are buoyantly rising radially outward in the clustergas.Unexpectedly,
Chandra and
XMM observations of nor-mal elliptical galaxies have also revealed a wealth of Diehl & Statlerhighly disturbed gas morphologies with qualitativelysimilar properties to clusters. Many individual obser-vations have been analyzed in detail and a variety ofdifferent explanations have been proposed. Sharp, one-sided drops in surface brightness are usually interpretedas signs of ram-pressure, distorting the gas in ellipti-cal galaxies as they move through the ambient intr-acluster or intragroup medium, as in NGC 4472 (Ir-win & Sarazin 1996; Biller et al. 2004) and NGC 1404(Machacek et al. 2004). To explain the very asymmet-ric emission in NGC 7618, Kraft et al. (2006) even ar-gue for a major group-group merger. Other observationsare not interpreted as results of environmental effects,but rather as signatures of the central AGN. For a fewcases, such as NGC 4374 (Finoguenov & Jones 2001) orNGC 4472 (Biller et al. 2004), the association with theAGN is indicated clearly by the correspondence betweenthe radio source and depressions in the X-ray gas dis-tribution. While these cavities are generally surroundedby cool rims (McNamara & Nulsen 2007, and referencestherein), two low-power radio galaxies (Centaurus A andNGC 3801) have recently been identified with shocks sur-rounding the radio lobes (Kraft et al. 2003, 2007; Crostonet al. 2007). In other objects, one can directly detect theX-ray counterpart of the radio jet (e.g. Sambruna et al.2004; Harris et al. 2002; Hardcastle et al. 2003). However,most cases are less clear-cut. For example, the origin ofthe shock-like features seen in NGC 1553 (Blanton et al.2001) or NGC 4636 (Jones et al. 2002) is still a mystery.The morphologies in other galaxies qualitatively resem-ble ghost cavities very close to the core (e.g. NGC 5044,Buote et al. 2003), without evidence for an active radiosource.These individual cases each come with their own analy-sis techniques and interpretations, making general state-ments about elliptical galaxies problematic. Detailedanalyses are generally limited to the X-ray brightestgalaxies with clearly identifiable features, which are oftendismissed as special cases. However, our morphologicalsurvey of normal elliptical galaxies in Paper I shows thatdisturbances and asymmetries in the gas are actually thenorm rather than the exception, even for relatively X-rayfaint galaxies. Most galaxies in the
Chandra archive lackthe signal, and therefore the contrast, to reliably identifycavity or shock features individually. In this paper, weintroduce a statistical measure of asymmetry (the asym-metry index, η ) that allows us to treat all objects on anequal footing. We will show that the asymmetry index isstrongly correlated with two different measures of nuclearactivity, and with the presence of ambient hot gas.The analysis in this paper begins with the ellipticalisophotal fits of Paper I, and is organized as follows. In § η for eachgalaxy. We show by simulations that η is a monotonicmeasure of the sort of asymmetry expected from multiplecavities, and that it is independent of non-morphologicalparameters. We describe our techniques to extract thecentral AGN X-ray luminosity and to determine radiopower and environment. In § § § DATA ANALYSIS
Summary of Paper I and Preliminary Analysis
Our full sample, as described in Paper I, consists of54 early-type galaxies observed with the the ACIS-Sinstrument on the
Chandra satellite during cycles 1-4. The data have been homogeneously reprocessed toavoid problems due to changes in the standard calibra-tion pipeline over time. We will briefly summarize thegeneration of those data products essential for the un-derstanding of this paper and refer the interested readerback to Paper I for more details.In Paper I, we presented a new method to isolate thehot gas emission in elliptical galaxies from the contami-nation of unresolved point sources. This method is basedon the fact that point sources and hot gas contribute dif-ferently to soft (0 . − . . − .For a subset of 36 galaxies with sufficient signal, wecharacterize the overall shape of the gas by deriving ellip-ticity and position angle profiles. A qualitative compari-son with optical DSS-2 R -band images and a quantitativecomparison with published optical surface photometryreveal little correlation between optical and X-ray prop-erties. In particular, we find no correlations between theoptical and X-ray ellipticities at any radius. Even withinone effective radius, where stars dominate the gravita-tional potential (e.g. Mamon & Lokas 2005; Humphreyet al. 2006), the two are completely uncorrelated. Thesecond column of Table 1 lists mean gas ellipticities inthe annulus between 0 . . J -band effective radii,from Paper I. These findings establish that the gas in el-liptical galaxies is, in general, not in precise hydrostaticequilibrium, at least at a level that precludes using thegas isophotes to constrain the shapes of dark matter halos(Buote & Canizares 1994; Buote et al. 2002). Instead ofbeing hydrostatically calm, the gas morphology is almostalways disturbed, even for rather X-ray faint galaxies.In this paper, we restrict our analysis to the subset of36 galaxies (Table 1) for which we derive ellipticity pro-files in Paper I. Following the technique used in PaperI to derive photon-flux calibrated images, we also pro-duce calibrated energy-flux images, which are necessary ∼ diehl/WVT orphological Evidence for AGN Feedback in Normal Ellipticals 3 Fig. 1.—
Left panel : Gas X-ray surface brightness (in photons s − cm − arcsec − ), ellipticity and position angle profile for NGC 4649.Solid lines indicate the Chebyshev polynomial fits. Right panel : Adaptively binned gas map of NGC 4649 with the fitted elliptical isophotesoverlaid.
Fig. 2.—
Gas surface brightness maps for three elliptical galaxies, showing the range in asymmetry index η ; Left: NGC 6482 ( η = 0 . η = 0 . η = 0 . × − to 10 − photons s − cm − arcsec − . The axes are labelled according to right ascension and declination (2000). to determine the central X-ray AGN luminosity ( § ∼
100 eV) and generate counts images in each indi-vidual PI channel (14 . Quantifying Asymmetry
Definition of the Asymmetry Index η Since many of the
Chandra data sets lack the requiredsignal and resolution to reliably identify individual asym-metric features in the gas surface brightness maps, weare forced to measure asymmetry statistically. We firstconstruct a smooth, symmetric surface brightness modelto subtract from the gas map. We base this model onthe isophotal profiles computed in Paper I. These profilesgive surface brightness ( I X ), ellipticity ( ǫ X ), and majoraxis position angle ( P A X ) as a function of mean radius r . We fit this surface brightness profile in log r − log I X space with a quadratic Chebyshev polynomial. For twogalaxies with more complex surface brightness profiles(NGC 1316 and NGC 4636), we use a third order poly-nomial, and for one other object (NGC 4472), we resortto fourth order. To accommodate moderate changes inellipticity and isophotal twists, we fit the ellipticity andposition angle profiles in log r − ǫ X and log r − PA X spacewith straight lines. From the fits, we compute a smooth,symmetric model for the gas distribution. The left panelof Figure 1 shows all three profile fits for NGC 4649. Diehl & Statler TABLE 1Chandra X-ray gas morphology, AGN luminosity and temperature gradient
Name ǫ Xa η b r minb r maxb L X , AGNc α IC1262 0 . ± .
09 0 . ± .
042 6 . . < . × . ± . . ± .
12 0 . ± .
040 0 . . . ± . × − . ± . . ± .
06 0 . ± .
032 1 . . . ± . × . ± . . ± .
42 0 . ± .
074 3 . . . ± . × · · · NGC0315 0 . ± .
14 0 . ± .
059 1 . . . ± . × · · · NGC0383 0 . ± .
09 0 . ± .
065 1 . . . ± . × . ± . · · · · · · · · · · · · < . × · · · NGC0507 0 . ± .
17 0 . ± .
055 2 . . < . × . ± . . ± .
04 0 . ± .
032 1 . . . ± . × · · · NGC0720 0 . ± .
07 0 . ± .
042 0 . . < . × − . ± . . ± .
05 0 . ± .
038 1 . . < . × − . ± . · · · · · · · · · · · · . ± . × · · · NGC1132 0 . ± .
25 0 . ± .
047 2 . . < . × − . ± . · · · · · · · · · · · · . ± . × · · · NGC1316 0 . ± .
04 0 . ± .
041 0 . . . ± . × . ± . . ± . < .
095 0 . . . ± . × . ± . . ± .
04 0 . ± .
013 0 . . . ± . × . ± . . ± .
03 0 . ± .
023 0 . . . ± . × . ± . . ± . · · · · · · · · · . ± . × · · · NGC1553 0 . ± .
14 0 . ± .
091 1 . . . ± . × − . ± . . ± .
09 0 . ± .
074 1 . . . ± . × · · · NGC1700 0 . ± .
11 0 . ± .
069 1 . . < . × · · · NGC2434 · · · · · · · · · · · · . ± . × · · · NGC2865 · · · · · · · · · · · · < . × · · · NGC3115 · · · · · · · · · · · · < . × · · · NGC3377 · · · · · · · · · · · · < . × · · · NGC3379 · · · · · · · · · · · · < . × · · · NGC3585 · · · · · · · · · · · · . ± . × · · · NGC3923 0 . ± .
10 0 . ± .
033 0 . . < . × − . ± . . ± .
05 0 . ± .
027 0 . . . ± . × − . ± . . ± .
08 0 . ± .
054 0 . . . ± . × · · · NGC4365 · · · · · · · · · · · · < . × · · · NGC4374 0 . ± .
04 0 . ± .
050 0 . . . ± . × . ± . . ± .
09 0 . ± .
021 0 . . . ± . × − . ± . . ± .
03 0 . ± .
012 0 . . . ± . × · · · NGC4494 · · · · · · · · · · · · < . × · · · NGC4526 0 . ± .
16 0 . ± .
177 0 . . . ± . × · · · NGC4552 0 . ± .
02 0 . ± .
028 0 . . < . × . ± . · · · · · · · · · · · · < . × − . ± . · · · · · · · · · · · · < . × · · · NGC4621 · · · · · · · · · · · · . ± . × · · · NGC4636 0 . ± .
02 0 . ± .
009 0 . . . ± . × . ± . . ± .
03 0 . ± .
035 0 . . . ± . × − . ± . . ± . < .
080 0 . . < . × · · · NGC5018 0 . ± . · · · · · · · · · < . × · · · NGC5044 0 . ± .
05 0 . ± .
022 1 . . . ± . × . ± . · · · · · · · · · · · · < . × · · · NGC5171 · · · · · · · · · · · · < . × · · · NGC5532 · · · · · · · · · · · · . ± . × − . ± . · · · · · · · · · · · · . ± . × · · · NGC5846 0 . ± .
02 0 . ± .
018 0 . . . ± . × . ± . . ± . < .
032 1 . . < . × − . ± . . ± .
12 0 . ± .
067 1 . . . ± . × · · · NGC7618 0 . ± .
14 0 . ± .
060 3 . . . ± . × − . ± . a Mean X-ray gas ellipticity ǫ X between 0 . − . R J ( J -band effective radius from 2MASSextended source catalog, Jarrett et al. 2000). b Asymmetry index η , measured between r min and r max (in kpc). c X-ray AGN luminosity (in ergs s − ) between 0 . − β and S´ersic modelplus point source fits to radial surface brightness profiles. d Mean logarithmic temperature gradient d ln T/d ln r evaluated at radii between 2 and 4 R J orphological Evidence for AGN Feedback in Normal Ellipticals 5The adaptively binned gas map in the right panel hasthe elliptical isophotes of the adopted fits overlaid. Fi-nally, we bin the surface brightness model to match thebinning structure of the adaptively binned gas map andsubtract the binned smooth model to reveal small scaleasymmetries in the gas distribution.To quantify the degree of asymmetry in this residualmap, we define the asymmetry index η in the followingway: η = 1 N N X i =1 "(cid:18) G i − M i M i (cid:19) − (cid:18) σ G,i M i (cid:19) . (1)The asymmetry index is the sum over all pixels i of thesquared relative deviations of the binned gas image G i from the smooth model M i , over and above the expectedstatistical deviations due to Poisson noise σ G,i . We findthat η is an unbiased measure of asymmetry, independentof exposure time, radial fitting range, background levelor signal-to-noise ratio. The errors on the η values areestimated from a bootstrap analysis of 20 Monte-Carlosimulations.Figure 2 shows representative examples spanning therange of η values present in our sample. NGC 6482( η < . η = 0 . η value ( η = 0 .
48) in our sample. This sequencecan serve as a guide for the meaning of η . However, oneshould keep in mind that the examples in Figure 2 havebeen chosen to have high signal-to-noise ratios for displaypurposes. Most galaxies have significantly less signal andthe asymmetric structures cannot be picked out by eyeas easily.The third, fourth, and fifth columns of Table 1 list thevalues of η and the range of radii from which they arederived. We have chosen to calculate η from the full ra-dial range over which isophote fits are possible for eachobject, in order to take advantage of as much detectedsignal as possible. This choice unavoidably creates a cer-tain inhomogeneity in the sample since η is not alwaysbeing measured at the same physical scale. However, wehave tested the sensitivity of all of the results of this pa-per to the size of the extraction annulus by separatinginto subsamples with small and large outer radii r max . Inno case do we find any statistically significant differencebetween the subsamples. Simulations and Tests
To test the behavior of the asymmetry index η , wesimulate images that qualitatively reproduce the varietyof gas morphologies seen in the data. We start with a256 ×
256 pixel realization of a S´ersic model with index n = 4 and a half-light radius of 50 pixels. We then adddisturbances as relative surface brightness depressionssurrounded by enhanced rim emission, mimicking the ap-pearance of AGN-induced “bubbles”, often observed in Equation (1) is equivalent to an area-weighted average overbins. We find empirically that an unweighted average gives toomuch emphasis to bright regions and small-scale structure. clusters of galaxies (e.g. McNamara et al. 2000). Fig-ure 3 shows a one-dimensional cut through our adoptedcircular bubble template profile, which is obtained bysubtracting two e − ( r/σ ) functions with slightly different σ values, and adjusting their normalizations such that in-tegrating the 2-d profile yields a total deviation of zero.We define the “depression strength” as the maximumrelative deviation at the bubble center, indicated by the − η is largely independent of the bubble size,but correlates positively with the number of simulatedbubbles and their relative depression strength, as shownin the left and middle panel of Figure 4. We compute theexpected trends by simulating 20 random spatial config-urations for each parameter set without adding Poissonnoise, and computing the average of all η values of these“perfect” data sets. We find that our asymmetry indexvalues are fair representations of the expected values.The asymmetry index is constructed to account for thestatistical scatter in the data and to be insensitive to dif-ferences in data quality. To verify this, we repeat ourtests, keeping the bubble parameters fixed while varyingthe total galaxy luminosity, background level, cutoff ra-dius, and the exposure time of the observation. We findno correlations with any of these parameters. As an ex-ample, in the right panel of Figure 4 we show one test inwhich we vary the exposure time. One can see that η isnot affected by the increasing resolution that is a resultof the rising total number of counts. We conclude that η is an unbiased measure of the asymmetry present inthe gas maps, and unaffected by non-morphological dataproperties. X-ray AGN Luminosities
We identify the location of the central AGN in the X-ray image by overlaying the high-accuracy center fromthe 2MASS extended source catalogue (Jarrett et al.2000). The 2MASS position is derived from the lumi-nosity weighted center of the co-added K , J and H bandimages and has a typical 1 σ -uncertainty of only 0 . ′′ . We Diehl & Statler Fig. 3.—
One-dimensional cut through a simulated “bubble de-pression”, demonstrating the fractional depression strength of thebubble with respect to its radial extent. The bubble has negativesurface brightness deviations within the bubble radius R Bubble , andslightly enhanced rims around it. The profile shape is chosen suchthat the integrated deviations integrate to 0 over area.
Fig. 4.—
Tests of the asymmetry index using models with sim-ulated bubble-like depressions (see Fig. 3). The asymmetry index η is sensitive to the number of simulated bubbles (left), and therelative strength of the depression (middle), but not to the effec-tive exposure time, characterized by the total number of counts(right). The dashed lines mark the expected behavior for perfect(noise-free) data. combine this uncertainty with Chandra ’s pointing accu-racy , and manually assign the AGN to the brightest X-ray point source in the 0 . − . ∼ − http://cxc.harvard.edu/mta/ASPECT/abs point.html Instead, we decide to spatially disentangle the X-rayAGN emission from its complex background using theenergy-flux calibrated images of diffuse emission. Theenergy weighting makes it easier to distinguish the hardAGN component. We bin the energy-flux image into cir-cular annuli centered on the AGN position and adap-tively change the annular bin sizes to enforce a minimumsignal-to-noise requirement of 2. As this would result invery large numbers of bins at larger radii, we also requirethe annular width to be larger than 10% of the mean binradius to ensure proper azimuthal averaging within a bin,generally resulting in a logarithmic binning at large radii.We fit the inner 32 arcsec of the radial surface bright-ness profile with a two-component model, one represent-ing the point source and one the diffuse emission. Thepoint source profile is obtained from a normalized mono-energetic point-spread function (PSF) centered on theposition of the central AGN and computed at its meanphoton energy. We bin the PSF image to the same cir-cular binning, to give a radial point source model thatmatches our binning structure with only the normaliza-tion flux as a free parameter. To represent the diffuseemission we use a S´ersic or a β model, on top of a uni-form background. Figure 5 shows the inner 10 arcsec oftwo galaxies with a statistically significant AGN detec-tion (NGC 4261, left) and a non-detection (NGC 6482,right). The dashed lines indicate the diffuse componentfrom the S´ersic (grey) and β model fit (black), while thesolid lines show the same fits with the additional AGNcomponent. All fits are inspected by eye and obviouslyunconverged fits are repeated over a different radial ex-tent until a satisfactory fit is obtained.Even though both models generally produce statisti-cally equally good fits, the β model fits systematicallyyield higher fluxes for the AGN, as the β -profile is flatterat small radii. We adopt the average of the AGN fluxesyielded by the β and S´ersic model fits as the best value.The difference between the two individual fits and theaverage value is assumed to represent a 3 σ systematicerror and is combined with the statistical errors yieldedby each profile fitting procedure. Finally, we correct theAGN fluxes for Galactic absorption with the same correc-tion factor that is used to correct the total gas flux, com-puted from integrating the best spectral fit for the gasemission with and without the Galactic absorption fac-tor. This correction may be a slight overestimate, sincethe AGN emission is generally harder and less stronglyaffected by absorption; on the other hand the columndensity is also likely to be higher due to intrinsic absorp-tion inside the galaxy. This correction is generally rathersmall, well below our statistical and systematical errors,and does not affect the computed X-ray AGN luminositysignificantly.The final AGN luminosities L X , AGN are reported inthe sixth column of Table 1, together with the combinedstatistical and systematic errors. Flux values that are de-tected below the 1 σ confidence limit are reported as up-per limits. For objects with insufficient signal to producea surface brightness profile, we sum up the energy-fluxcalibrated image within the region containing 99% of thePSF flux and correct for a flat background extracted froman adjacent annulus extending out to 10 arcsec. Since wehave no way to distinguish between a central peak in gasemission, LMXBs, or an actual AGN for these galaxies,orphological Evidence for AGN Feedback in Normal Ellipticals 7we quote the 3 σ upper bound of the summed flux as anupper limit on their AGN luminosity. Radio Luminosities
We use the NRAO VLA Sky Survey (NVSS; Condonet al. 1998) to derive homogeneous 20 cm radio contin-uum luminosities for our sample. As we are interested inthe AGN’s impact on the ISM, we sum up all associatedradio sources within 3 J-band effective radii from the2MASS catalog (Jarrett et al. 2000) radii to derive theNVSS radio luminosity L NVSS , thus including contribu-tions from extended structures. This results in generalin an extraction radius of approximately 1 . ′ for mostof the sample (see Paper I for a complete list of opticalradii). The formal average NVSS source density for thefull NVSS catalog is about 50 sources per square degree(Condon et al. 1998), resulting in a probability to includeunrelated background sources within 1 . ′ of only 9%. For63% of our sample, this probability lies below 10%, for91% it lies below 20%. To further reduce the chance ofincluding background sources, we inspect all NVSS im-ages by eye and compare radio features with X-ray andoptical images to find possible counterparts. We manu-ally remove obviously non-associated background sourcesfrom the source list.The resulting luminosities are listed in the second col-umn of Table 2. For the majority of our sample, thelimited spatial resolution of the NVSS survey precludesreliably distinguishing between extended and point-likeradio sources. Only where unambiguously possible, wedetermine the NVSS position angle by manually measur-ing the orientation of the major axis.Twenty-three galaxies in our sample have been ob-served in the Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) VLA survey (Becker et al. 1995),which has a much higher spatial resolution than NVSS,with a similar detection limit. We extract FIRST ra-dio luminosities from the inner 3 optical effective radiito match the NVSS extraction region. A comparison be-tween the two radio surveys indicates that both are yield-ing consistent values. However, the higher spatial reso-lution of the FIRST survey gives us the opportunity toextract fluxes from the central point source alone, largelyexcluding contributions from the radio jets. Thus, in-stead of using a large extraction radius, we sum the fluxwithin a smaller 30 arcsec radius to determine the FIRSTluminosity L FIRST , listed in the third column of Table 2.These values are systematically smaller than the NVSSfluxes due to the smaller extraction region, but still verywell correlated. This indicates that our larger NVSS sam-ple is not heavily contaminated by background objects,but that the extra flux is most likely associated with ex-tended radio structure intrinsic to the galaxy.
Correlation between Radio and X-ray AGNLuminosities
Figure 6 compares the AGN X-ray luminosity withboth the NVSS (left) and FIRST (right) 20 cm contin-uum radio powers. Both plots show a clear correlationbetween radio and X-ray luminosities, spanning almostsix orders of magnitude. The slope is consistent with apurely linear relation between both luminosities, as indi-cated by the solid lines. We analyze all of the correlations
TABLE 2Radio luminosities and galaxy environment
Name L NVSSa L FIRSTa log ρ IC1262 1 . ± . × . ± . × . ± . . ± . × · · · . ± . . ± . × · · · . ± . . ± . × · · · . ± . . ± . × · · · . ± . . ± . × · · · . ± . . ± . × · · · · · · NGC0507 6 . ± . × · · · . ± . . ± . × · · · . ± . < . × · · · . ± . . ± . × · · · . ± . < . × · · · · · · NGC1132 1 . ± . × . ± . × . ± . . ± . × · · · · · · NGC1316 1 . ± . × · · · . ± . . ± . × · · · . ± . . ± . × · · · . ± . . ± . × · · · . ± . · · · · · · . ± . · · · · · · . ± . . ± . × · · · . ± . < . × · · · . ± . · · · · · · . ± . < . × · · · · · · NGC3115 < . × < . × · · · NGC3377 < . × < . × · · · NGC3379 3 . ± . × < . × . ± . < . × · · · . ± . < . × · · · . ± . . ± . × · · · . ± . . ± . × . ± . × . ± . < . × < . × . ± . . ± . × . ± . × . ± . . ± . × < . × . ± . . ± . × . ± . × . ± . < . × < . × · · · NGC4526 6 . ± . × . ± . × . ± . . ± . × . ± . × . ± . < . × . ± . × . ± . < . × < . × · · · NGC4621 < . × < . × . ± . . ± . × . ± . × . ± . . ± . × . ± . × . ± . < . × < . × . ± . < . × · · · . ± . . ± . × · · · . ± . . ± . × · · · · · · NGC5171 < . × . ± . × · · · NGC5532 5 . ± . × . ± . × . ± . < . × < . × · · · NGC5846 1 . ± . × . ± . × . ± . < . × · · · . ± . . ± . × · · · . ± . . ± . × · · · . ± . a
20 cm continuum radio luminosity (in ergs s − Hz − ) from NVSSwithin 3 R J , and FIRST within 30 ′′ . b Projected local galaxy density (in Mpc − ), derived from the2MASS extended source catalog (Jarrett et al. 2000). Reported er-rors only include statistical errors, not systematic errors. Diehl & Statler
Fig. 5.—
Radial X-ray surface brightness profiles for the inner 6 ′′ of NGC 4261 (left) and NGC 6482 (right). Thick solid lines indicateS´ersic model fits, thin solid lines show β model fits, including a PSF model for the central AGN. Dashed lines show the same model fitswithout the AGN component. NGC 4261’s AGN is detected at a 7 σ level, NGC 6482 profile reveals no AGN signature. For NGC 6482,the AGN flux yielded by the S´ersic model fit is so small that the thick dashed line is invisible, as it is covered by the thick solid line. in this paper using an algorithm (“ bandfit ”; see Ap-pendix A) that models the distribution of data points as alinear band with finite Gaussian intrinsic width. Bandfit can handle data with errors in either or both variables,as well as censored data (upper or lower limits). The bandfit analysis for the L NVSS – L X , AGN relation putsthe chance for the null hypothesis of no correlation toless than 10 − % ( ∼ σ ). This correlation is not sim-ply a consequence of plotting distance vs. distance; it isequally significant if the distance dependencies of bothparameters are removed. The best fit correlation can bedescribed aslog L X , AGN = 0 . ± .
10) log L NVSS + 19 . ± . . (2)Similar correlations have been found in the past forAGN (e.g. Hardcastle & Worrall 1999; Brinkmann et al.2000, and references therein), with slopes ranging from ∼ . & L X , AGN withthe total gas luminosity L X , gas , we find a steeper slope,similar to previous results for radio galaxies.Hardcastle & Worrall (1999) have argued that the ex-istence of an L x – L radio correlation indicates that bothX-ray and radio emission originate in a strongly Dopplerboosted jet. If this were the case, then measured X-rayor radio fluxes might reflect primarily the orientation ofthe jet relative to the line of sight, rather than the intrin-sic power of the AGN. In this picture, the shallower slopeof the correlation for quasars could be explainable by acombination of beamed and isotropic X-ray emission, im-plying that X-rays would, at least in part, measure AGNpower. Whether the shallow slope we find in the L NVSS – L X , AGN relation indicates a similarity with the quasars isunclear. It is possible that our slope could be affected atthe low-luminosity end by occasional misidentification ofindividual LMXBs in the galaxy centers as AGN. How- ever, if L NVSS or L X , AGN were only reflecting the jetorientation, we would not expect either to be correlatedwith gas morphology, as we do see.A survey of radio emission from normal elliptical galax-ies by Wrobel & Heeschen (1991) suggests that the ma-jority of their radio flux is due to AGN activity ratherthan star formation. The fact that nuclear X-ray and ra-dio luminosities show a tight relationship supports thisconclusion for our sample, as well as indicating that wehave properly identified and isolated the central AGN inthe X-ray images. We conclude that both the radio andthe X-ray AGN luminosities are reliable tracers of thecentral AGN activity in our sample, most likely due tojets powered by gas accreting onto a supermassive blackhole.
The Galaxy Environment
To quantify the likelihood of interactions with neigh-boring galaxies, we measure the projected galaxy den-sity in the vicinity of our sample galaxies by countingthe number of neighbors listed in the 2MASS extendedsource catalog (Jarrett et al. 2000). We identify all galax-ies within a projected radius of 100 kpc at the distanceof the target as “neighbors.” We restrict this distanceto a maximum angular size of 15 arcmin, to minimizecontamination by random foreground or background ob-jects. Despite the smaller physical extraction radius fornearby objects, the number of neighbors is still sufficientto derive a galaxy density, as the incompleteness limitdrops significantly, and lower-luminosity neighbors canbe identified. Because the 2MASS catalog is flux-limited,we have to cope with a distance-dependent completenesslimit. We use the K -band Schechter luminosity func-tion from Kochanek et al. (2001) to correct for incom-pleteness. We integrate the luminosity function downto the completeness limit at the object’s distance anddivide by the integral of the luminosity function downto a reference luminosity, which we arbitrarily chose as M K = − .
5. This way we can calculate the fraction ofall galaxies above the reference luminosity that 2MASS isable to detect. We find that, after correcting for incom-pleteness, our galaxy densities ρ are independentorphological Evidence for AGN Feedback in Normal Ellipticals 9 Fig. 6.—
X-ray AGN luminosity vs. 20 cm continuum radio luminosity with data taken from NVSS within 3 optical radii ( left ) and fromFIRST ( right ) within a 30 ′′ radius. Arrows indicate 3 σ upper limits. The slope of the relationship between radio and X-ray luminosity isclose to linear, as indicated by the solid lines. The dashed line in the left panel shows the bandfit best fit relation, which has a slope of0 . ± .
10. Thus, X-ray and radio luminosities are each measures of AGN activity. of distance. We list the values for ρ in Table 2in units of galaxies per Mpc . The errors include onlystatistical errors due to the galaxy counting statistics;deriving systematic errors due to background objects oruncertainties in the galaxy luminosity function are be-yond the scope of this paper.Ideally, one would also like to restrict the list of neigh-bors in velocity space, to remove galaxies that are simplyaligned along the line of sight. Such a three-dimensionalrestriction requires an additional unbiased source of ra-dial velocities. However, available radial velocities areheavily biased toward “more interesting” regions of thesky and better studied objects. Thus, we refrain fromusing these velocities and use the projected galaxy den-sity ρ instead, a measure similar to the Tully den-sity parameter (Tully 1988), but well-defined for our en-tire sample. All conclusions based on ρ are repro-ducible with the Tully parameter instead.The hot ISM morphology may also be affected by hy-drodynamic interactions with a surrounding intra-clustermedium (ICM) or intra-group medium (IGM). One ex-pects a transition to an IGM or ICM at large radii to beaccompanied by an outward rise in gas temperature, re-flecting the higher virial temperature of the group or clus-ter. To assess the presence of a hotter ambient medium,we compute radial temperature profiles for our sample.The X-ray counts image of each galaxy is divided into el-liptical annuli, according to the X-ray ellipticity profilescomputed in Paper I, except for those with insufficientsignal to fit ellipses, for which we revert to circular an-nuli. We then extract a source and background spectrumfor each annulus and fit them with a two-componentmodel in the CIAO analysis package Sherpa. The firstcomponent consists of an APEC plasma model to rep-resent the hot gas emission. A quantitative compari- Astrophysical Plasma Emission Code son with its better known predecessor, the Mekal model,shows nearly identical results. We fix the gas metal-licity at the solar abundance value. Unresolved pointsources are represented by a power-law model with thepower law index fixed at 1.6. This “universal” spectralmodel is an adequate representation of the emission oflow-luminosity low-mass X-ray binaries, as demonstratedin Paper I and determined independently by Irwin et al.(2003). We also add a multiplicative absorption compo-nent, for which we fix the hydrogen column density tothe Galactic value, evaluated at the target position withthe CIAO tool
Colden . We repeat our spectral analy-sis for a few objects with the gas abundance as a freeparameter, and find that our choice to fix them to thesolar value does not affect the fitted temperature. Sincethe metallicity is poorly constrained by the fits in lowsignal-to-noise systems, we fix the metallicity for all ofour galaxies, in order not to introduce systematic dif-ferences in the analysis. We then fit the temperatureprofile by a power law, and derive a mean logarithmictemperature gradient between 2 and 4 J-band effectiveradii ( R J ): α ≡ d ln T /d ln R | − R J . Values of α arelisted in the last column of Table 1.One can consider the galaxy density ρ as an in-dicator of the likelihood of gravitational galaxy interac-tions through close encounters or mergers, similar to theTully galaxy density parameter. The outer temperaturegradient, on the other hand, is a measure of the presenceof hot ambient gas and thus a proxy for the likelihood ofhydrodynamic interactions with this gas. A much morethorough analysis of the temperature profiles present inour sample will be the main focus of Paper III. RESULTS
X-ray Ellipticity and Asymmetry http://cxc.harvard.edu/toolkit/colden.jsp Fig. 7.—
The X-ray asymmetry index η as a function of X-ray ellipticity ǫ X . The observed positive trend suggests a commonunderlying cause for both the large measured X-ray ellipticities andthe asymmetries. Figure 7 shows a comparison between the mean X-ray ellipticities between 0 . − . η . There is a trend for flatter galaxiesto be also more asymmetric, with a bandfit correlationstrength of 99.72%. There is a clear deficit of galax-ies with flattened X-ray emission, but small asymmetry,which suggests a common underlying cause for both. Theellipticity may best be interpreted as simply a mean mea-sure of the m = 2 Fourier amplitude of the departurefrom circular symmetry. This is consistent with our re-sults from Paper I, where we find that the gas elliptici-ties are uncorrelated with the ellipticities of the starlight,and thus are not simply determined by the shape of thesmooth underlying gravitational potential. AGN Influence on Gas Morphology
Figure 8 shows the dependence of the asymmetry index η on the two independent measures of AGN activity: theX-ray luminosity L X , AGN of the central point source (leftpanel) and the NVSS radio power L NVSS , integrated overthree optical radii (right panel). Both AGN propertiesare correlated with η , in the sense that galaxies hostingstronger AGN have stronger morphological asymmetriesin their hot ISM. Analysis with bandfit rejects the nullhypothesis of no correlation between L NVSS and η at the98.7% confidence level (see Table 3). The same analysisfor L X , AGN and η yields a slightly lower confidence valueof 97.1%. If we regard L X , AGN and L NVSS as indepen-dent measures of the same phenomenon ( § generally responsible for disturbing thehot gas morphology even in normal elliptical galaxies andprovides further support for the importance of AGN feed-back. TABLE 3Significances for correlations with η . Second Variable P Null log L NVSS & log L X , AGN (jointly) 0.04% α L NVSS L X , AGN ρ ρ Tully
A comparison with our simulations in § η values tend to have stronger surface brightness depres-sions, possibly caused by the stronger interaction of morepowerful radio sources with the ISM. Environmental Influence on Gas Morphology
The probability of interactions with neighboring galax-ies through mergers, tidal interactions or close encoun-ters naturally increases with the local number density ofgalaxies. Thus, if interactions are important in influenc-ing the gas morphology of elliptical galaxies, we wouldexpect a correlation between morphological parametersand the projected galaxy density ρ . However, Fig-ure 9 shows that neither the gas ellipticity (upper panel),nor the asymmetry index (lower panel) shows any trendwith this measure of environment. A bandfit analysisyields no statistically significant correlation in either case(Table 3). Bootstrapping and a Spearman rank analysisgive consistent results.To ensure that our conclusions are independent of ourdefinition of galaxy density, we repeat the analysis withother environmental measures of galaxy density: theTully density parameter ρ Tully (Tully 1988), the distanceto the 10th closest neighbor in the 2MASS extendedsource catalog (Jarrett et al. 2000), and the number ofgalaxies associated with a galaxy group, as listed in theLyon Group of Galaxies (LGG; Garcia 1993) catalog. Allof these measures of environment yield the same result:the galaxy density is not correlated with the observedasymmetries or large ellipticities of the hot gas. Thus,the X-ray gas morphology is not mainly driven by in-teractions with companions, although they may still beimportant for some individual objects.In the absence of galaxy-galaxy interactions, hydro-dynamic interaction with a hot ambient medium is stillpossible, through ram pressure, shocks, pressure confine-ment or hydrodynamic instabilities at the contact inter-face. Two galaxies in our sample are already known tobe affected by ram-pressure: NGC 4472 (Biller et al.2004) and NGC 1404 (Machacek et al. 2004). However,in both cases, the radius within which we compute theasymmetry index is smaller than the radius at which theinteraction with the ambient medium becomes obvious.orphological Evidence for AGN Feedback in Normal Ellipticals 11
Fig. 8.—
Asymmetry index η as a function of X-ray AGN luminosity L X , AGN (left) and 20 cm radio continuum power L NVSS taken fromNVSS (right). Both plots indicate that the central AGN luminosity is positively correlated with the observed asymmetry in the gas, withthe correlation extending all the way down to the weakest AGN.
Fig. 9.—
X-ray ellipticity ǫ X (top) and asymmetry index η (bottom) as a function of projected galaxy density ρ . Bothplots indicate that environment is not the driving factor causingthe ellipticities and asymmetries. To assess the sample in general, we use the outer tem-perature gradient α as a proxy for a hot ICM, IGM, orcircumgalactic gas. A larger gradient suggests a largerambient gas pressure. Figure 10 shows that asymmetry Fig. 10.—
Asymmetry index η as a function of the outer temper-ature gradient α , an indicator for the presence of hotter ambientgas. The correlation is such that galaxies in a hot gas environmentare more asymmetric. Note that the sample includes both rising( α >
0, hotter ambient gas present) and falling ( α <
0, noambient gas present) temperature profiles. indeed increases as a function of α . A bandfit analy-sis rejects the null hypothesis at the 99.8% level (Table3), with similar results obtained from bootstrapping andSpearman rank correlations. DISCUSSION
Causes of Disturbances in the Gas Morphology
Our comparison of optical and X-ray morphologies inPaper I revealed that even well within one effective ra-dius, where the gravitational potential is dominated bythe stellar component (Mamon & Lokas 2005; Humphreyet al. 2006), there is no correlation between stellar and2 Diehl & Statlergas ellipticities, very much at odds with predictions fromhydrostatic equilibrium. In fact, many X-ray ellipticitieseven exceed those of the starlight, quite the opposite ofwhat is expected from purely gravitational effects. Wealso noted in Paper I the prevalence of asymmetries inthe hot gas, which we have now statistically quantified inthis paper with the asymmetry index η . We find that thelarge observed ellipticities and the amount of asymme-try in the gas are correlated, pointing toward a commoncause that dictates the overall X-ray morphology.In this Paper, we have shown that feedback from thecentral AGN (Figure 8) and hydrodynamic interactionswith the ambient medium (Figure 10) are the dominantcauses of these asymmetries. Our analysis of the AGN-morphology correlation shows that this phenomenon isnot restricted to powerful radio sources, but forms a con-tinuous sequence down to the weakest AGN. The corre-lation with the outer temperature gradient α , on theother hand, suggests that the presence of a hot am-bient medium has a similarly strong effect on the gasmorphology. Figure 11 shows that both correlations to-gether dictate the disturbance in the gas. In Figure 11awe have colored the residuals in the α − η correla-tion according to AGN luminosity, with red indicatinglow ( L NVSS < erg s − Hz − ), and green high, ra-dio power. In Figure 11b we color the residuals in the L NVSS − η correlation according to α , with red now indi-cating a negative and green a positive temperature gradi-ent. (Black points indicate objects for which these valuesare unavailable.) In each case, the residuals are clearlycorrelated with the third quantity. Replacing L NVSS withthe X-ray AGN luminosities L X , AGN yields comparableresults. Note that the scatter about the mean correla-tions is about the same in each case, indicating that AGNand ambient medium have comparable effects on the gasmorphology.
Implications for AGN Feedback in EllipticalGalaxies
The great majority of galaxies in our sample are notconspicuous AGN; most harbor only weak, unresolved ra-dio sources. Nevertheless, we find a correlation betweenhot gas asymmetries and the radio and X-ray AGN lu-minosities, down to the lowest luminosities detectable byNVSS, FIRST, or
Chandra.
These results solidify theAGN as an influential feedback mechanism for normalelliptical galaxies in general, even where they do not ob-viously harbor powerful nuclear sources.Intermittent AGN activity may stir the hot gas, push-ing it far enough out of equilibrium that informationabout the shape of the underlying potential is lost. Thatwe observe morphological disturbances in nearly all ellip-tical galaxies, and that some objects even reveal signa-tures of multiple outbursts (e.g. O’Sullivan et al. 2005),suggest that the duty cycle of the AGN is shorter thanor comparable to the sound crossing time. Thus, thehot gas is continually disturbed, without having time toresettle into equilibrium.Despite the lack of correlation between optical and X-ray ellipticities (Paper I), we do note a weak tendencyfor the hot gas isophotal major axes to be aligned withthose of the starlight. To quantify this trend, we extracta subsample of 15 galaxies that have X-ray ellipticities ǫ X > . ◦ -wide bins, are shown as the grey-shaded histogram in Figure 12. A Kolmogorov-Smirnov(KS) test puts the probability for a chance correlation atonly 0.05%.While this alignment may reflect the orientation of theunderlying gravitational potential, it may also be the re-sult of interactions with the central radio source. Oursample exhibits an anti-correlation between radio and op-tical major axes, which is shown as the solid histogramin Figure 12. This controversial phenomenon has beenobserved early on by Palimaka et al. (1979), but hasnot been confirmed in subsequent studies (Sansom et al.1987). Nevertheless, a KS-test yields only a 1.1% proba-bility for the null hypothesis of uncorrelated orientationsin our small sample of 11 galaxies with both availableradio and optical major axes.In case of a relatively stable radio jet axis, one wouldexpect the jet to preferentially evacuate gas along theaxis, creating a misalignment between radio and X-ray gas emission. We do observe a very slight anti-correlation, shown in the dashed histogram in Figure 12,but only 12 galaxies have both reliable X-ray and radioposition angles. The large error bars on both radio andX-ray position angles make it difficult to ascertain thesignificance of the correlation, which a KS-test puts nom-inally at 90%. Extracting the X-ray position angles at alarger radius increases the significance; but the detectionof this misalignment is marginal and awaits confirmationwith a larger sample and deeper radio data. With ourcurrent small sample, a two-sided KS test yields a 22.3%probability that the radio–X-ray anticorrelation and theX-ray–optical alignment are drawn from the same distri-bution, which rises to 80.6% at a larger extraction radius.Thus, the data are consistent with the alignment beingcaused by the effects of the radio source on the hot ISM.The competing effect of the interaction with the ambi-ent medium may also work against a proper detectionof this misalignment. However, the position angle mayalso represent the only information that is still inferableabout the underlying gravitational potential, namely itsorientation. Implications for Interactions with the IntergalacticMedium
We find that the observed ellipticities and asymmetriesin the hot gas are not mainly a consequence of inter-actions with neighboring galaxies. On the other hand,interactions with the ambient medium seem to be as im-portant as the interplay with the central radio source.We expect the impact of an external ambient mediumto increase with distance from the center. However, ouranalysis is confined to a region within
Chandra ’s limitedfield of view, restricted further by our need for sufficientsignal to fit isophotes. Thus, our asymmetry index is in-sensitive to structure outside a radius which depends onthe average surface brightness of the object and the ex-posure time of the observation. To verify that the α − η correlation is not simply a consequence of a variable ex-traction radius, we have checked for a correlation be-tween this radius and η and we find none.The majority of X-ray halos associated with ellipticalgalaxies extend well beyond the outer radii in Table 1,as was already demonstrated with observations with theorphological Evidence for AGN Feedback in Normal Ellipticals 13 Fig. 11.—
Similar to Figures 10 and 8.
Left : correlation between η and the outer temperature gradient α . Red colors indicatelow-luminosity ( L NVSS < erg s − Hz − ), and green colors higher luminosity ( L NVSS > erg s − Hz − ) AGN. Right : correlationbetween η and radio luminosity L NVSS . Red points indicate a negative, and green a positive, temperature gradient α . In each case theresiduals from the mean relation are correlated with the third variable. On can consider the more fundamental correlation to be a 2-d planein log L NVSS – α –log η space. Fig. 12.—
Histogram of absolute differences between optical(PA opt ), radio (NVSS 20cm continuum, PA
NVSS ), and X-ray gas(PA X ) position angles. The grey shaded histogram with dottedcontours shows a tendency for galaxies to have X-ray and opticalisophotes aligned. The solid line shows an anti-correlation betweenoptical and radio position angles, and the dashed line a slight anti-correlation between radio and X-ray position angles, suggestingthat the radio source is responsible for the X-ray–optical alignment. Einstein and
ROSAT satellites (e.g. Forman et al. 1985;O’Sullivan et al. 2003). Interactions with the ambientmedium may very well have an even stronger influenceon the gas structure at larger radii. At this point weare not yet able to distinguish the effects of AGN fromthose of environment based on morphology alone. Nev-ertheless it is curious to note that we do not observe acorrelation between the X-ray ellipticity and the outertemperature gradient, as one might expect. Perhaps adifferent definition of statistical asymmetry that is sen-sitive to “lopsidedness” may be useful in this regard.Alternatively, the existence of a positive outer temper-ature gradient could also be interpreted as the presenceof an external fuel reservoir for the central AGN. Best(2004) finds a correlation between radio-loud AGN ac-tivity and environment, in the sense that a larger frac- tion of AGN in richer environment are switched “on”compared to isolated galaxies. In this scenario, the cor-relation between the outer temperature gradient and thegas asymmetry would be a result of an increased dutycycle for AGN outbursts, rather than hydrodynamic in-teractions with the ambient medium. In addition, thecircumgalactic gas reservoir also acts as a pressure con-finement, which would prevent radio jets and inflatedcavities from simply leaving the system and provides abackground contrast to actually identify asymmetries atlarger radii. CONCLUSIONS
We have extended our earlier work from Paper I on nor-mal elliptical galaxies, which showed that optical and X-ray gas ellipticities are uncorrelated in the regions wherestellar mass dominates, contrary to what is expectedfrom gas in hydrostatic equilibrium. Instead, the hot gasis generally disturbed. To elucidate the nature of theseasymmetries, we have introduced the asymmetry index η ,which measures the statistical deviation from a symmet-ric surface brightness model. This allows a systematicstudy of asymmetry even in relatively low-luminosity ob-jects, in which asymmetric features are barely resolved.We find a strong correlation between the asymmetryindex and two independent measures of AGN activity:the radio continuum power at 20 cm from NVSS (Con-don et al. 1998) and the X-ray AGN luminosity extractedfrom Chandra data. The observed AGN–asymmetry cor-relation persists all the way down to the weakest AGN,where the NVSS survey reaches its detection limit. Thisis quite surprising, since these objects generally lack ex-tended jet signatures in their radio images and are mostlydetected as weak central point sources, if at all. Wealso find the asymmetry index to be correlated with theouter temperature gradient α , a proxy for the pres-ence of a hot intragroup or intracluster medium. Thestrength of this correlation is comparable to that of theAGN–asymmetry correlation, indicating that hydrody-4 Diehl & Statlernamic interaction with the ambient medium is compa-rably important to AGN in determining the X-ray mor-phology of normal ellipticals. However, we find no suchcorrelation with the density of neighboring galaxies, andno evidence that galaxy-galaxy interactions play a sig-nificant role in shaping the hot ISM in these systems.Alternatively, the presence of an outer temperature gra-dient may also indicate the availability of a fuel reservoirfor the central AGN. In this case, the increased amountof asymmetries in the X-ray gas could indicate a greaterduty cycle for AGN outbursts in these objects.The emerging picture is consistent with the AGN per-sistently stirring up the interstellar medium through in-termittent outbursts, and strengthens the case for theAGN to be at least partly responsible for offsetting cool-ing in normal elliptical galaxies. We will address the im-pact of the central AGN on temperature and entropy pro-files in detail in Paper III of this series (Diehl & Statler 2008).We thank the anonymous referee for an insightful ref-eree report which helped improve the manuscript. Wehave made use of data products from the Two Micron AllSky Survey, which is a joint project of the University ofMassachusetts and the Infrared Processing and AnalysisCenter/California Institute of Technology, funded by theNational Aeronautics and Space Administration and theNational Science Foundation. Support for this work wasprovided by the National Aeronautics and Space Admin-istration (NASA) through Chandra Awards G01-2094Xand AR3-4011X, issued by the Chandra X-Ray Observa-tory Center , which is operated by the Smithsonian Astro-physical Observatory for and on behalf of NASA undercontract NAS8-39073, and by National Science Founda-tion grant AST0407152.
REFERENCESAscasibar, Y., & Markevitch, M. 2006, preprint (astro-ph/0603246)Becker, R. H., White, R. L., & Helfand, D. J. 1995, ApJ, 450, 559+Best, P. N. 2004, MNRAS, 351, 70Biller, B. A., Jones, C., Forman, W. R., Kraft, R., & Ensslin, T.2004, ApJ, 613, 238Blanton, E. L., Sarazin, C. L., & Irwin, J. A. 2001, ApJ, 552, 106Brinkmann, W., Laurent-Muehleisen, S. A., Voges, W., Siebert, J.,Becker, R. H., Brotherton, M. S., White, R. L., & Gregg, M. D.2000, A&A, 356, 445Buote, D. A., & Canizares, C. R. 1994, ApJ, 427, 86Buote, D. A., Jeltema, T. E., Canizares, C. R., & Garmire, G. P.2002, ApJ, 577, 183Buote, D. A., Lewis, A. D., Brighenti, F., & Mathews, W. G. 2003,ApJ, 594, 741Cappellari, M., & Copin, Y. 2003, MNRAS, 342, 345Chatzikos, M., Sarazin, C. L., & Kempner, J. C. 2006, preprint(astro-ph/0602199)Condon, J. J., Cotton, W. D., Greisen, E. W., Yin, Q. F., Perley,R. A., Taylor, G. B., & Broderick, J. J. 1998, AJ, 115, 1693Croston, J. H., Kraft, R. P., & Hardcastle, M. J. 2007, ApJ, 660,191Diehl, S., & Statler, T. S. 2006, MNRAS, 368, 497—. 2007, ApJ, 668, 150—. 2008, ApJ, in preparationFinoguenov, A., & Jones, C. 2001, ApJ, 547, L107Forman, W., Jones, C., & Tucker, W. 1985, ApJ, 293, 102Fukazawa, Y., Botoya-Nonesa, J. G., Pu, J., Ohto, A., & Kawano,N. 2006, ApJ, 636, 698Garcia, A. M. 1993, A&AS, 100, 47Hardcastle, M. J., & Worrall, D. M. 1999, MNRAS, 309, 969Hardcastle, M. J., Worrall, D. M., Kraft, R. P., Forman, W. R.,Jones, C., & Murray, S. S. 2003, ApJ, 593, 169Harris, D. E., Krawczynski, H., & Taylor, G. B. 2002, ApJ, 578, 60Humphrey, P. J., Buote, D. A., Gastaldello, F., Zappacosta, L.,Bullock, J. S., Brighenti, F., & Mathews, W. G. 2006, ApJ, 646,899Irwin, J. A., Athey, A. E., & Bregman, J. N. 2003, ApJ, 587, 356Irwin, J. A., & Sarazin, C. L. 1996, ApJ, 471, 683+Jarrett, T. H., Chester, T., Cutri, R., Schneider, S., Skrutskie, M.,& Huchra, J. P. 2000, AJ, 119, 2498Jones, C., Forman, W., Vikhlinin, A., Markevitch, M., David, L.,Warmflash, A., Murray, S., & Nulsen, P. E. J. 2002, ApJ, 567,L115Khosroshahi, H. G., Jones, L. R., & Ponman, T. J. 2004, MNRAS,349, 1240 Kim, D.-W., & Fabbiano, G. 2003, ApJ, 586, 826Kochanek, C. S., Pahre, M. A., Falco, E. E., Huchra, J. P., Mader,J., Jarrett, T. H., Chester, T., Cutri, R., & Schneider, S. E. 2001,ApJ, 560, 566Kraft, R. P., Jones, C., Nulsen, P. E. J., & Hardcastle, M. J. 2006,ApJ, 640, 762Kraft, R. P., Nulsen, P. E. J., Birkinshaw, M., Worrall, D. M.,Penna, R. F., Forman, W. R., Hardcastle, M. J., Jones, C., &Murray, S. S. 2007, ApJ, 665, 1129Kraft, R. P., V´azquez, S. E., Forman, W. R., Jones, C., Murray,S. S., Hardcastle, M. J., Worrall, D. M., & Churazov, E. 2003,ApJ, 592, 129Machacek, M., Dosaj, A., Forman, W., Jones, C., Markevitch, M.,Vikhlinin, A., Warmflash, A., & Kraft, R. 2004, preprint (astro-ph/0408159)Mamon, G. A., & Lokas, E. L. 2005, MNRAS, 362, 95McNamara, B. R., & Nulsen, P. E. J. 2007, ARA&A, 45, 117McNamara, B. R., Wise, M., Nulsen, P. E. J., David, L. P., Sarazin,C. L., Bautz, M., Markevitch, M., Vikhlinin, A., Forman, W. R.,Jones, C., & Harris, D. E. 2000, ApJ, 534, L135McNamara, B. R., Wise, M. W., Nulsen, P. E. J., David, L. P.,Carilli, C. L., Sarazin, C. L., O’Dea, C. P., Houck, J., Donahue,M., Baum, S., Voit, M., O’Connell, R. W., & Koekemoer, A.2001, ApJ, 562, L149Nulsen, P. E. J., McNamara, B. R., Wise, M. W., & David, L. P.2005, ApJ, 628, 629O’Sullivan, E., Ponman, T. J., & Collins, R. S. 2003, MNRAS, 340,1375O’Sullivan, E., Vrtilek, J. M., & Kempner, J. C. 2005, ApJ, 624,L77Palimaka, J. J., Bridle, A. H., Brandie, G. W., & Fomalont, E. B.1979, ApJ, 231, L7Sambruna, R. M., Gambill, J. K., Maraschi, L., Tavecchio, F.,Cerutti, R., Cheung, C. C., Urry, C. M., & Chartas, G. 2004,ApJ, 608, 698Sansom, A. E., Danziger, I. J., Ekers, R. D., Fosbury, R. A. E.,Goss, W. M., Monk, A. S., Shaver, P. A., Sparks, W. B., & Wall,J. V. 1987, MNRAS, 229, 15Tully, R. B. 1988, Nearby galaxies catalog (Cambridge and NewYork, Cambridge University Press, 1988, 221 p.)Wrobel, J. M., & Heeschen, D. S. 1991, AJ, 101, 148
APPENDIX
CORRELATION ANALYSIS WITH BANDFIT