Galaxy gas ejection in radio galaxies: the case of 3C 35
aa r X i v : . [ a s t r o - ph . C O ] F e b Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 2 September 2018 (MN L A TEX style file v2.2)
Galaxy gas ejection in radio galaxies: the case of 3C 35
E. Mannering ⋆ , D. M. Worrall , M. Birkinshaw H.H. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 1TL, UK
ABSTRACT
We report results from
XMM-Newton and
Chandra observations of the nearby ( z =0 . t synch ≈
140 Myr, t belt ≈
80 Myr). The destruction of 3C 35’s atmospheremay offer clues as to how fossil systems are regulated: radio galaxies need to be ofpower comparable to 3C 35 to displace and regulate fossil-group gas. We discuss theimplications of the gas belt in 3C 35 in terms of AGN fuelling and feedback.
Radio galaxies are important sources of heating in groupsand clusters, and feedback by active galactic nuclei (AGN)is now considered as the most likely mechanism to balanceradiative cooling (e.g., McNamara & Nulsen 2007). The hotintergalactic medium (IGM) may be directly heated byshocks and sound waves driven by the expansion of radiojets (e.g., Kraft et al. 2012; Forman et al. 2007). Cavitiesinflated by the radio jets may also limit the ability of theIGM to cool, by doing work on the gas, lifting it out from thedensest regions of the group or cluster (e.g., Blanton et al.2009; Bˆırzan et al. 2008). To understand the role of fu-elling/feedback of AGN, it is necessary to study radio jetinteractions on all scales, from AGN to group/cluster envi-ronments. Nearby radio galaxies are obviously best suitedto this task.One such source is 3C 35, a low-redshift ( z = 0 . P
178 MHz =10 . W Hz − sr − , spanning a total angular size of roughly12 . ′ . Burns & Owen (1977) classify 3C 35 as lying in anopen cluster, coincident to within 2 cluster radii of a Zwickycluster (Zwicky et al. 1961), however McHardy (1974) findthis to be a weak cluster environment and suggest 3C 35probably resides in a small group. In contrast, Guindon(1979) classified 3C 35 as a non-cluster source based on thedefinition of an Abell or Zwicky cluster.In this paper, we report X-ray measurements of a gasbelt that we argue may be a disrupted fossil-group X-ray Corresponding to a linear size of 950 kpc at z = 0 . halo, as well as diffuse emission in the lobes of 3C 35. Con-straining the spectral properties and energetics of the con-stituent relativistic particles in the lobes (e.g. Worrall 2009,for a review) and the thermodynamical properties of theIGM provide useful clues as to the origin of the gas belt.3C 35 seems to be the first case where we see the X-ray haloof a fossil group being ejected following an interaction withthe radio structure. Fossil groups are a class of system that has an extended, hotgaseous halo encompassing a single giant elliptical galaxy,but with the gravitating mass of a galaxy group (Hess et al.2012; Jones et al. 2003). They are believed to be the mergerremnants of a galaxy group whose L ∗ galaxies have co-alesced, leaving behind an extended gas halo and dwarfgalaxies (Dariush et al. 2007, and references therein). Fossilgroups are identified observationally by two criteria; an ex-tended X-ray source with L X > h − erg s − (Jones et al.2003), and an optically dominant elliptical galaxy, where thecompanion galaxies are fainter by two magnitudes in R . Thelower limit in X-ray luminosity attempts to exclude normalbright elliptical galaxies exhibiting a hot coronal gas com-ponent, and the magnitude gap in R ensures that a singleelliptical dominates the system. N − body simulations suggest that fossil groups formedearly (D’Onghia et al. 2005), assembling half of their totaldark matter mass at z &
1, with later growth by minor merg-ers only, whereas non-fossil groups formed much later. Theextensive, virialised X-ray halo, lack of ∼ L ∗ galaxies andthe large magnitude gap tend to support this interpretation(Jones et al. 2003; Mendes de Oliveira et al. 2006). Fossilgroups are at least as numerous as all poor and rich clusterscombined (Jones et al. 2003), contributing significantly to c (cid:13) E. Mannering et al. the total mass density of the Universe. Dariush et al. (2007)use the Millennium simulation and semi-analytic galaxy cat-alogues to trace the formation and evolution of fossil groups,and find that fossil groups may be a phase of hierarchi-cal evolution, rather than a final stage of mass assembly(La Barbera et al. 2009).Fossil-group halos are hotter and more X-ray lumi-nous than non-fossil groups and clusters of similar gravita-tional mass (Khosroshahi et al. 2007). Croston et al. (2005)showed that radio-loud ( L . > W Hz − ) non-fossilgroups are hotter at a given X-ray luminosity than radio-quiet groups, indicating AGN activity is heating the IGM.Hess et al. (2012) found a weak correlation between 1.4 GHzluminosity of radio-loud AGN and the X-ray luminosity ofthe halo, tentatively suggesting that the AGN contributesto the energy deposition in the IGM of fossil systems too.Two-thirds of their sample have radio-loud AGN, suggestingthat AGN fuelling continues (or is re-ignited) long after thelast major merger. Lobe radio emission is well described by the syn-chrotron process (O10 for 3C 35), while the extended non-thermal X-ray radiation from the lobes of radio galax-ies and quasars detected in many FR II sources is at-tributed to the inverse-Compton (iC) mechanism (e.g.,Brunetti et al. 1997; Hardcastle et al. 2002; Comastri et al.2003; Croston et al. 2005; Hardcastle & Croston 2005;Isobe et al. 2005; Goodger et al. 2008; Isobe et al. 2011).Modelling lobe-related iC X-ray emission enables constraintsto be placed on the electron energy density ( u e ) and mag-netic field strength ( B ) which otherwise cannot be decoupledusing radio data alone. X-rays from the lobes are a sensitiveprobe of u e in low-energy electrons within the lobes and, incombination with radio mapping, sometimes allow investi-gation of positional variations in B and the electron energyspectrum (e.g., Croston et al. 2005; Goodger et al. 2008).Giant radio galaxies enable exploration of the late phasein the evolution of radio jets, as lobe energetics are an im-portant indicator of past activity because they accumulatethe energy deposited by the jets. If the non-thermal particleand magnetic field energy densities can be measured fromthe diffuse (iC) lobe emission, estimates can be made of thetotal energy (which is potentially available to be transferredto the environment) and the distribution of internal energywithin the source, without assuming equipartition. This inturn constrains the past duration of the active phase of theactive nucleus.
3C 35 is an old source: its spectral age is estimated byO10 as 143 ±
20 Myr. It has been previously observed withthe X-ray Imaging Spectrometer (XIS) on board
Suzaku (Mitsuda et al. 2007), revealing faint extended X-ray emis-sion associated with its radio lobes. Isobe et al. (2011, here-after IS11) integrated the
Suzaku spectrum within a largerectangular region containing the whole radio structure of3C 35. They report iC emission associated with the lobes(power-law index Γ = 1 . +0 . . − . − . , where the first errors are due to signal statistics, and the second are from back-ground uncertainties), as well as a soft thermal componentcorresponding to thermal plasma emission from the hostgalaxy ( kT = 1 . +0 . . − . − . keV). In the case of giant ra-dio galaxies such as 3C 35, the cosmic microwave background(CMB) is the dominant source of photons to be iC scattered(by relativistic electrons of γ ∼ ) up to X-ray energies, al-though nuclear photons and starlight are important in somesources where they are scattered by lower-energy electrons(e.g., Brunetti et al. 1997; Stawarz et al. 2003). Due to lowelectron densities in the large lobe volumes, the synchrotronself Compton process makes a negligible contribution to theX-ray flux (e.g., Croston et al. 2005; Hardcastle et al. 2002).We use new XMM-Newton (hereafter
XMM ) observa-tions and archival
Chandra data to investigate an X-ray gasbelt seen orthogonal to the lobes of 3C 35, and to probe theiC properties of the lobes themselves. In Section 2 we brieflydescribe the
XMM and
Chandra observations and data re-duction. The X-ray morphology of 3C 35, and the results ofspectral fitting are discussed in Sections 3 and 4. In Section5 we present a discussion of our findings. Throughout, weuse a flat ΛCDM cosmology with Ω m = 0 . Λ0 = 0 . H = 70 km s − Mpc − . We used archival NRAO Very Large Array (VLA) and
Chan-dra data in combination with new
XMM
X-ray data to exam-ine the emission mechanisms in the lobes of 3C 35 and probethe physical parameters of the gas belt. Our X-ray data haveimproved sensitivity and resolution compared to the
Suzaku data of IS11. The high spatial resolution of
Chandra en-ables us to reduce contamination from X-ray backgroundsources and to isolate individual source components. How-ever, the small
Chandra effective area leads to a low numberof ACIS counts, whereas
XMM count rates are higher, en-abling analysis of the properties of the X-ray emission andelectron population as a function of position.
We observed 3C 35 with
XMM for just under 100 ks duringFebruary 2011. Data are presented from the EPIC MOS1,MOS2 and pn cameras in full-frame mode using the medium filter. We reprocessed the Observation Data Format (ODF)files using the dedicated pipeline chain tasks emchain and epchain for EPIC MOS and pn, respectively. We ex-tracted single and double events ( pattern pattern XMM
Ex-tended Source Analysis Software procedure (XMM-ESAS;Snowden & Kuntz 2011) and the associated current calibra-tion files (CCF) which contain filter wheel closed (FWC),quiescent particle background (QPB) and soft proton (SP)calibration data . Data were downloaded from the
XMM
Science Operations Cen-tre (SOC) and analysed using SAS version 11.0.0 ftp://xmm.esac.esa.int/pub/ccf/constituents/extras/esas caldb/c (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 Figure 1.
Light curves generated from the 3C 35 MOS1, MOS2and pn event lists. The black line shows the raw data in 60 s binsfor the whole field of view. The red lines show the GTI applied‘clean’ data.
Table 1.
XMM exposures. The duration refers the total livetimeof the observation. Net exposure is the good time remaining after1 . σ background screening.Detector Duration (ks) Net exposure (ks)MOS1 95.4 30.7MOS2 95.4 31.3pn 79.4 21.4 The SP component due to solar flares cannot be wellremoved via background subtraction. It is highly variable,causing the spectrum to change rapidly with time. Frameswith high count rates must be excluded, which unfortunatelycan significantly reduce net exposure time.
In order to create a Good Time Intervals (GTI) fileand ascertain the severity of the increased solar activ-ity on the observation of 3C 35, we followed the pre-scription of background screening outlined in the XMM-ESAS guide (Snowden & Kuntz 2011) and ran the dedicatedESAS tasks mos-filter and pn-filter to generate SPcontamination-filtered products for the field-of-view data.It is clear from the light curves (Figure 1) that high flaringdominated much of the observation. A Gaussian was fitted tothe nominal count rate distribution and a GTI was createdfor those time intervals with count rates within a thresh-old value of an acceptable level ( ± . σ , Table 1). CCD4 on the MOS1 detector was operating in an anomalousstate (the background at E < cheese combines MOS andpn data to generate point-source lists (down to a limiting flux of 1 . × − erg cm − s − ), exposure maps and maskimages for use in creating source-excluded spectra. Tasks mos-spectra and pn-spectra were run in order to ex-tract spectra from the cleaned event files for a given selectionregion. Finally, the tasks mos back and pn back were runto create model particle background spectra for the selectedregions. The XMM background was estimated using meth-ods described in §
3C 35 was observed with
Chandra using the Advanced CCDImaging Spectrometer (ACIS) on 2009 March 8. The front-illuminated ACIS-I3 CCD was placed at the focus of theobservation, configured in timed exposure and VFAINTtelemetry mode. The data were reduced using the latest
Chandra software ( ciao
Chandra analysis threads . We reprocessed the level1 event files, applying vfaint filtering. The latest charge-transfer inefficiency (CTI) correction and time-dependentgain adjustment were applied and a background light curvewas produced. The observation did not suffer from back-ground flaring, and we selected events of grade 0, 2, 3, 4 or6 resulting in a net exposure time of 25.63 ks. The ciao func-tion wavdetect was used to identify point sources from a0.3-12.5 keV exposure corrected image, and regions around62 sources over 4 CCDs were excluded from subsequent spec-tral extractions. X-ray spectra were extracted from regionsusing the specextract task, and grouped to a minimum of30 counts per bin. VLA data taken as part of program AL419 were downloadedfrom the archive. We use 1.42 GHz data from an observationmade on 1997 August 27th in C array. The radio map wasmade using standard AIPS tasks, with phase self calibrationand a restoring beam of 14 × ′′ (FWHM). Data were com-bined using roughly 3 hr of on-source integration time to ob-tain a root-mean-square (rms) sensitivity < .
07 mJy/beamacross the field. Contours are shown in Figure 2.
A heavily smoothed 0.5-7.0 keV, background-inclusive
Chandra image with point sources excluded is shown in Fig-ure 2, with radio contours obtained from our VLA map at1.4 GHz overlaid. The X-ray core is weak (we detected 9counts in the
Chandra data) and offset from the radio nu-cleus (O10) by approximately 0 . ′′ , which is within the sys-tematic uncertainty of Chandra
X-ray astrometry. In orderto study the morphology of the extended emission, we ex-cluded all point sources in the field, including the core. dm-filth was used to fill in the gap regions with intensities fromsurrounding areas. Figure 2 shows the result after smoothingusing the CIAO tool aconvolve , with a Gaussian function http://cxc.harvard.edu/ciao/threads/index.htmlc (cid:13) , 000–000 E. Mannering et al.
Figure 2.
Chandra exposure-corrected image of 3C 35 in the en-ergy range 0.5-7.0 keV. VLA C-array radio contours at 1.42 GHzare overlaid. The peak radio flux is 32.0 mJy/beam and contourlevels are spaced by a factor of 2, the lowest being 0.5 mJy/beam(restoring beam size ≈ × ′′ FWHM). The X-ray data arebinned by a factor of 4 into pixels of size roughly 2 ′′ . Point sources(including the X-ray nucleus) were removed and the values inter-polated from the surrounding pixels using the ciao tasks wavde-tect and dmfilth . Data are smoothed using the ciao task acon-volve with a two dimensional Gaussian function of σ = 10 pixels( ≈ ′′ ). of σ ≈ ′′ as a convolution kernel, and dividing by the expo-sure map in order to correct for varying effective area acrossthe field. A small X-ray source is offset from the radio core;this is extended emission, the X-ray nucleus was properlysubtracted within an exclusion region of radius 3.5 ′′ .We created an XMM image in the soft energy band0 . − .
25 keV and the hard band 2 . − . ′′ pixels and adaptivelysmoothed. We used an off-source region to estimate an av-erage surface brightness of the X-ray background (describedin § Chandra and
XMM data there is a cleardetection of faint, diffuse X-ray emission filling the radiolobes, suggesting a non-thermal inverse-Compton origin aspreviously reported by IS11 on the basis of lower-resolution
Suzaku data. There is no excess X-ray emission associatedwith the radio hotspots. The striking filamentary appear-ance of the X-ray emission in Figure 3 (left) may be an arte-fact of the smoothing, but is evocative of a jet-like channelextending to the edge of the northern hot-spot. With the lownumber of counts, however, we cannot pursue this furtherat present.
Figure 3.
Exposure corrected and background subtractedMOS+pn
XMM (left) and 0.4-1.25 keV (right) im-ages. Point sources in the field have been removed, the data arebinned into 5 ′′ pixels and adaptively smoothed by a kernel of 50counts. Overlaid are the lowest radio contour from Figure 2 andthe region used to define the gas belt. Figure 4.
XMM exposure corrected and background subtractedMOS+pn 0.4-1.25 keV image of the gas belt. The data are binnedto 5 ′′ and adaptively smoothed by a kernel of 50 counts, after pointsources have been removed. Gridlines are spaced at ∆ RA, ∆ δ =(1 . ′ , . ′ ) and are marked in degrees. The
XMM
PSF is a strong function of off-axis angle andphoton energy (Lloyd-Davies et al. 2011), becoming asym-metric at very large off-axis angles. It is important to knowthe variation of the PSF across the FOV in order to cor-rectly define extraction areas for point sources in the gasbelt. The belt extends roughly 2 ′ from the centre of the fieldof view. In table 2 of Ness et al. (2011), the on-axis in orbit1.5 keV PSFs are given as having a FWHM . . ′′ for thepn telescope and . . ′′ for the MOS1 and 2 detectors. Infigure 9 of Dahlem (1999) the radius at which 90% of thetotal energy of a point-source is encircled (W90) is plottedagainst off-axis angle for different energies. W90 for 1.5 keVat 2 ′ is approximately 15 ′′ . c (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 Figure 5.
Chandra postage stamps of 4 gas belt exclusion re-gions. Pixels are 0.492 arcsecond square, and data are displayedaccording to their energy (red 0.2-1.5 keV, green 1.5-2.5 keV, blue2.5-8.0 keV). The positions of the three regions are shown in Fig-ure 6.
We modified the
XMM masks to include also all sourcesdetected in the
Chandra data, increasing the minimum ra-dius of a point-source to 15 ′′ in order to take into accountthe PSF increase off-axis. Less than half of the point sourcesdetected in the Chandra data were detected in the
XMM data. Figure 4 shows the gas belt in the
XMM data afterpoint sources have been excised and the data have beenheavily smoothed. There is a clear detection of X-ray countsin the belt region.
In both the
XMM and
Chandra data, particularly at softenergies, we find evidence of an X-ray gas belt constrainingthe central radio emission from the lobes of 3C 35. Gas cavi-ties associated with the radio lobes are not seen, in contrastto many cluster sources (e.g., Rafferty et al. 2006), and noextended thermal emission has been previously separatedspatially in the field of 3C 35. The belt emission is detectedout to about 170 kpc from the core, and it is coincident witha deficit in the radio emission (Figures 2 and 3). We defineda region for spectral extraction by eye, making use of theradio contours as a guide to capture the ‘dog-bone’ shape ofthe belt emission.In addition to the core, four point sources unresolvedin the
Chandra data (Figure 5) and detected with wavde-tect , are embedded within the gas belt. We examined theDigitized Sky Survey (
DSS , Figure 6) and found that one ofthese X-ray sources is associated with extended optical emis-sion (source 3). Sources 1 and 4 are confirmed as foregroundstars, whereas 2 has no optical detection.The
Hubble Space Telescope (HST)
Wide Field Plane-tary Camera 2 (WFPC2) HST archive (proposal ID 6967, P.I Sparks, W).We examine the calibrated (cosmic-ray free) science image,and confirmed the object marked within the red circle inFigure 6 to be a galaxy, probably a companion of the hostgalaxy of 3C 35. This is further discussed in § Figure 6.
DSS red image at 6100˚A(Second Palomar ObservatorySky Survey (POSS-II)) of the gas belt. Overlaid are positionsof point sources detected with the ciao task wavdetect . Theoutline of the gas-belt region as defined by the X-ray data is shownin black and the
Chandra chip gaps lie within the dashed parallelblack lines. X-ray point source 3 in the gas belt is identified as agalaxy by eye. Gridlines are spaced as in Figure 4. An adjacentgalaxy undetected in X-rays is marked by the red circle.
Inset :Zoomed image of regions 1-3; 2. has no optical counterpart.
X-ray spectra were fit with XSPEC 12.7.0 (Arnaud 1996),using χ statistics with Gaussian errors over the energyrange 0.4-7.0 keV. Metal abundance was fixed at 0 . Z ⊙ ,and a Galactic neutral hydrogen column density of 1 . × cm − (obtained with the Chandra
COLDEN tool,Dickey & Lockman 1990) was adopted in all fits. Quotederrors on parameter values correspond to 90 per cent con-fidence for one interesting parameter ( χ + 2 .
7, all otherinteresting parameters are allowed to vary). The spectral in-dex for non-thermal components, α is defined as S ν ∝ ν − α .The photon spectral index is Γ = α + 1, and the numberpower-law spectral index of radiating electrons is p = 2 α + 1if the radiation has a synchrotron or inverse-Compton origin. The radio structure of 3C 35 (about 13 ′ in the north-eastto south-west direction) extends across all four chips in the Chandra
ACIS-I field of view, and the emission from thelobes is faint and diffuse, making the selection of an ap-propriate background region difficult. We examined spectraextracted using three distinct background regions; a smalllocal background region close to the lobes (possibly includ-ing cluster gas), a larger local background, and a blank skybackground event file from the
Chandra
CALDB. The threeanalyses gave statistically consistent results, with the largelocal background producing the best fits and smaller param-eter errors. Consequently, only the spectral results obtainedwith the large local background region (Figure 8) are de-scribed in this paper. http://cxc.harvard.edu/toolkit/colden.jspc (cid:13) , 000–000 E. Mannering et al.
Table 2.
Principal emission lines used in the model,Ion Energy (keV) Ion Energy (keV)C v viii vi ix ≈ . vi x vii xi vii xiv We reduced the flare-free
XMM data using the XMM-ESASsoftware package. We chose to model the background ratherthan use a local background region due to the large extent of3C 35 and the spatial variability of the background; in par-ticular instrumental emission lines are not uniform acrossthe detectors. Blank-sky files were also inappropriate; resid-ual soft proton and Solar Wind Charge eXchange (SWCX)contamination are highly variable (Kuntz & Snowden 2008),and the periods of high flaring during our observation makecalibration to the blank-sky files difficult. We modelled theoff-source background using spectra extracted from a rectan-gular off-source region, west of the lobes of 3C 35 and awayfrom the gas-belt and group-gas extraction regions. Ourbackground model was then included in fitting the sourceemission.We follow the prescription of Snowden & Kuntz (2011)to model the
XMM background component. The XMM-ESAS task mos back generates a model quiescent particlebackground (QPB) spectrum using data from the FWC ob-servations and unexposed corners of the MOS CCDs (sim-ilarly, pn back generates files for the pn CCDs). The qui-escent instrumental background is excised from the data,while additional components must be modelled explicitly.Instrumental Al K α and Si K α lines were fitted with an un-absorbed Gaussian of zero intrinsic width at energies 1.49and 1.75 keV for the MOS detectors, and Al K α for thepn (the energy range for the pn data was restricted to0 . < E < . E < × s compared to thehard band, indicative of a change in state of the SWCX. Wetherefore include the emission lines for the ten lines listed inTable 2 in subsequent modelling.Diagonal response matrices and a power-law, not folded Figure 7.
XMM pn light curve of the full field of view (afterfiltering for periods of soft proton flaring, see § − . Each lightcurve was created for events with data flag FLAG =0, i.e eventsnot from or near a bad pixel, and binned by 720 s. The X-ray countrates do not have the particle background subtracted. The meancount rate per bin was corrected for any reduction in exposure ofthat bin, and the error shown on each data point is the standarderror on the mean ( σ/ √ N ). The hard band shows some minorscatter, indicating a low level of residual contamination by softproton flares. The soft band includes the majority of the [OVII]and [OVIII] line emission, which is associated with SWCX. Atroughly 5 × s, there is a significant drop in the 0.52-0.75 keVband, highlighting the variability of the SWCX emission. through the instrumental effective areas, were used to modelthe residual SP contamination (Figure 10, the solid darkblue lines show the contribution of the fitted residual SPbackground for each detector). The cosmic X-ray back-ground (CXB) was modelled by an unabsorbed thermalcomponent representing emission from the heliosphere ( E ≈ . E ≈ . − . α ≈ .
46 to repre-sent the unresolved background sources. We used the localRASS spectrum to constrain the contribution of the cosmicbackground (Figure 10, lower blue line).The best-fit model to the off-source data has χ = 98 . We extracted the
Chandra spectrum of the weak X-ray corefrom a circle of radius 1 ′′ (1.3 kpc), using the same back-ground regions as defined in § ± The HEASARC X-Ray Background Tool can be found athttp://heasarc.gsfc.nasa.gov/cgi-bin/Tools/xraybg/xraybg.plc (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 we are able to derive a weak constraint on the nuclear lu-minosity. Assuming an unabsorbed power-law representingthe nuclear X-ray continuum, modified by Galactic absorp-tion (Belsole et al. 2006), the best-fit (using C -statistics)photon index and 1 keV flux density are Γ = 1 . ± . S = 0 . +0 . − . nJy. This corresponds to an unabsorbed0.4-7.0 keV luminosity of 8 × erg s − , which is much lessthan the average core luminosity of L −
10 keV > erg s − found by Belsole et al. (2006) for 10 higher redshift FR IIradio galaxies (0 . < z < . . ± .
02) than thosewith intrinsic absorption greater than 5 × cm − . The5 GHz flux density of the nucleus of 3C 35 is roughly 12 mJy,based on our 1.4 GHz map and a spectral index of 0.7. Thisplaces 3C 35 in a typical L X /L R location for the populationof galaxies that have a nuclear component with intrinsic ab-sorption less than 5 × cm − (Evans et al. 2006, fig 2b).We conclude that there is no evidence for a luminous hidden-core X-ray component in 3C 35.Assuming all the core counts are associated with a coolcorona and not emission from the AGN, we also fitted thecore with a thermal model. The data are best fit with an apec temperature of 3 . +64 − . keV. Although the upper er-ror is unconstrained, there is no indication of a cool core.We froze the temperature to that found in the gas belt( kT = 1 .
28 keV, § . × erg s − , although the fit was poor. Thiswould place 3C 35 at the extreme high-luminosity end ofthe 1.4 GHz population of radio galaxies that host an X-raycorona (Sun et al. 2007, fig 6a). The core is unresolved inthe Chandra data and the counts are concentrated within < The
Chandra spectrum of the lobe X-ray emission was inte-grated within a radio contour of 0.5 mJy/beam (delineatedin Figure 8), excluding point sources and chip gaps. Fig-ure 9 shows the spectrum of the lobes, after subtracting abackground spectrum using the background region definedin Figure 8. Data are grouped to a minimum of 30 countsper bin. The net count rate and area of the extraction regionare given in Table 3.The lobes are best fitted with a power-law model, withparameters as given in Table 4. The best-fit photon indexΓ = 1 . +0 . − . is consistent with that of the radio spectrum(0 . < α r < . α X = α r (in common with e.g., IS11),we fix Γ = 1 . xspec apec model to the fitto model any thermal gas in the lobe region. We find aslightly improved fit with the addition of an apec compo-nent ( χ / dof=20.15/19), although the uncertainties are suchthat the normalisation of the apec model is consistent withbeing zero. Figure 8.
Chandra image of 3C 35 in the energy range 0.4-7.0 keV. Data are binned by 4 and smoothed with a Gaussianfunction of σ = 8 pixels ( ≈ ′′ ). A radio contour of 0.5 mJy/beamis used as the source region for spectral extraction. The back-ground (labelled bkg) is defined as the region outside the largedashed rectangle centred on 3C 35, but within the dashed square.Point sources and chip gaps are excluded from both regions. −3 no r m a li ze d c oun t s s − k e V − ∆ S χ Energy (keV)
Figure 9.
Chandra background subtracted, 0.4-7.0 keV spectrumof the diffuse emission fitted with a power-law model plus Galac-tic absorption (Table 4). The lower panel shows the residuals’contribution to χ . We extracted spectra integrated within a 0.5 mJy radio con-tour of the lobes (excluding the contribution from pointsources). The net count rates for each detector are givenin Table 3. We performed a combined fit of the EPICMOS and pn spectra (Figure 10), excluding energies out- c (cid:13) , 000–000 E. Mannering et al.
Table 3.
Properties of the regions used for spectral extraction. Net count rates are measured in 0.4-7.0 keV. For
Chandra , the numberin brackets is the fraction of counts in the extraction region attributed to the source. For the
XMM data, ‘net’ refers to the source,instrumental line, SWCX, SP, and CXB emission, minus the QPB background, i.e. only the QPB is excised from the data before fitting.Area is the net solid angle of the region, after excluding chip gaps, damaged CCDs and background sources.Net count rate (10 − ct s − ) Area (arcmin ) Chandra
MOS1 MOS2 pn
Chandra
MOS1 MOS2 pnLobes 2 . ± .
23 (23.8%) 6 . ± .
21 6 . ± .
20 17 . ± .
38 31.3 24.4 23.7 30.3Gas belt 0 . ± .
11 (17.3%) 1 . ± .
08 1 . ± .
07 3 . ± .
16 3.58 4.45 4.38 4.06Core 0 . ± .
017 (98.5%) - - - 0.001 - - -Group gas 0 . ± .
43 (2.1%) 1 . ± . . ± . . ± .
21 94.7 12.6 12.6 12.6
Table 4.
Summary of the power-law model fitting to the lobesof 3C 35 for
XMM and
Chandra . The Galactic neutral hydrogencolumn density is frozen for all fits.Model parameter
Chandra XMM N H (10 cm − ) 0.123 0.123Γ 1 . +0 . − . . +0 . − . [ +0 . − . ] S (nJy) 41 . +7 . − . . +2 . − . [ +6 . − . ] χ / dof 22.11/20 214.75/213 side 0 . < E < . XMM background ( § . σ (90%) values and the second setinclude 1 . σ errors on the background components also be-ing fit. The source emission is well fitted by a power lawwith a photon index Γ = 1 . ± .
11 and a flux density at1 keV of S = 48 ± Chandra fit. Under the assumption that α X = α r , we fixΓ = 1 . apec model in order to test for ther-mal emission in the lobe region. Although hinted at in the Chandra data, the small component of thermal gas in thelobe region was not constrained. We froze the backgroundcomponents to the values for the power law alone, assumedtheir contribution to the error is zero, and derived a best-fittemperature of 1 . +2 . − . keV and an absorption-corrected 0.4-7.0 keV luminosity of 2 . +4 . − . × erg s − ; this indicatesa thermal contribution to the total luminosity of the loberegion of roughly 7% . This is consistent with the results forthe gas belt in § H -band (1 . µ m) HST data tomodel the host galaxy of 3C 35, and derive a best-fit half-light radius ( r e ) of 4.6 kpc. If we assume emission from thehost is truncated beyond 2 r e , the XMM
PSF (15 ′′ ) is largeenough to include all X-ray emission from the volume ofthe optical galaxy. Since a region the size of the PSF wasexcluded from all spectral fitting, and our lobe fit is overthe entire lobe region (1 Mpc in diameter; large even forcluster gas) we interpret the detected thermal component as emission from the gas belt, which runs orthogonal to thelobe region and is visually the most prominent feature in theX-ray map after the lobes (see Figure 3). We extracted
Chandra spectra from a region we defined asthe gas belt (Figures 3 and 4), orthogonal to the lobes of3C 35, excluding point sources and chip gaps. We have fittedseveral models to the 124 on-source, background-subtractedcounts. Data are grouped into 32 bins of 32 channels width,and we used the C statistic, which performs better than χ in the low-count regime (Nousek & Shue 1989). XMM spec-tra were also extracted from the MOS1, MOS2 and pn detec-tors for the belt region. The net count rate for each detectoris given in Table 3. We modelled the
XMM background asdescribed in § Chandra spectra are shown in Figure 11, andbest-fit parameters for both instruments are given in Table5. The gas-belt
XMM data fit a single apec plasma model(with a fixed metal abundance at 0 . Z ⊙ ) of temperature2.5 keV. We also fitted the data with an absorbed power-law,and find an acceptable fit with a photon index of Γ = 2 . apec + power law (with afixed photon index Γ = 1 .
7) is the best-fit model. This fitgives a lower temperature of the gas, kT ≈ . xspec normalisation as compared to the single apec model, and there is a good agreement between the XMM and
Chandra measurements. The flux density of the power-lawcomponent is roughly 5 . We extracted the
Chandra spectrum from a large region sur-rounding 3C 35 (within a circle of radius 400 ′′ ) but excludingall point sources in the field, the lobes and the gas-belt re-gion; a net extraction region on the sky of approximately 95square arcminutes. We chose the same background regionas for the lobe spectral extraction, excluding the circular c (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 −4 −3 no r m a li ze d c oun t s s − k e V − ∆ S χ Energy (keV)10 −4 −3 no r m a li ze d c oun t s s − k e V − ∆ S χ Energy (keV)
Figure 10.
XMM data and best-fit model of the diffuse emissionfrom the lobes of 3C 35 and X-ray background.
Top panel : MOSon-source data from 0.4-7.0 keV (pn in grey for comparison). Theblack and red lines are the composite MOS1 and MOS2 modelsrespectively, the blue peaks are the Al K α and Si K α instrumentalemission lines, the orange lines are the cosmic x-ray backgroundcomponents. The green dashed lines show the SWCX emissionlines at [O VII] 0.565 keV, [O VIII] 0.65 keV, [Ne IX] 0.89 keV,[Ne X] 1.053 keV and the purple/dark blue solid line is the uncon-volved power-law representing residual SP emission. The lowerblue line and 5 blue crosses show the RASS data and best fitused to constrain the XMM
CXB background. The dashed blackline is a power-law representing the source emission from the re-gion. The lower panel shows the residuals as their contributionto χ . Bottom panel : pn off-source data from 0.4-7.0 keV, modelcomponents as above. on-source region. The data were grouped into 32 bins of 32channels width. We fitted an absorbed apec model using the C -statistic, and found a best-fit temperature of kT ≈ xspec normalisation of 6 . × − cm − (corrected forexcluded volumes). However, the group emission is weak, wecould not constrain either of these values. Instead we assumethese provide upper limits on the 0.4-7.0 keV luminosity ofthe group gas encompassing 3C 35, L group . . × erg s − (corrected for excluded volumes).We also extracted the XMM spectra for a rectangularregion (12 . ′ to the west of the core) ad-jacent to the gas belt. We chose a small region offset from3C 35 rather than the 400 ′′ circle, due to the difficulty inmodelling the XMM background over large areas. We mod-elled the background components in our extracted region as −3 −3 −3 no r m a li ze d c oun t s s − k e V − −3 −5×10 −4 −4 −3 no r m a li ze d c oun t s s − k e V − Energy (keV)010 −3 −3 −3 no r m a li ze d c oun t s s − k e V − −3 −5×10 −4 −4 −3 no r m a li ze d c oun t s s − k e V − Energy (keV)010 −3 −3 −3 no r m a li ze d c oun t s s − k e V − −4 −4 −3 no r m a li ze d c oun t s s − k e V − Energy (keV)
Figure 11.
Chandra background subtracted 0.4-7.0 keV spec-trum of the emission from the gas belt, fitted with top an apec model, best-fit temperature of ≈ . middle a power-law witha free photon index, Γ ≈ . bottom an apec model (red dot-ted) plus power-law (black dashed) with a frozen Γ = 1 .
7, best-fittemperature of gas ≈ . (cid:13) , 000–000 E. Mannering et al.
Table 5.
Best-fit parameters of the emission from the gas belt of3C 35 for
XMM and
Chandra . Models are absorbed apec , power-law and apec plus power-law. N is the apec normalization quotedin units of 10 − (1 + z ) R n e n p dV/ πD L .Model parameter Chandra XMM apec kT (keV) 3 . +13 . − . . +0 . − . [ +0 . − . ] N (10 − cm − ) 4 . +1 . − . . +0 . − . [ +0 . − . ] χ / dof cstat 24.30/29 power-law Γ 2 . +0 . − . . +0 . − . [ +0 . − . ] S (nJy) 9 . +2 . − . . +1 . − . [ +3 . − . ] χ / dof cstat 21.59/29 apec + power-law kT (keV) 1 . +1 . − . . +0 . − . [ +1 . − . ] N apec (10 − cm − ) 1 . +2 . − . . +1 . − . [ +1 . − . ]Γ 1.7 (frozen) 1.7 (frozen) S (nJy) 5 . +3 . − . . +1 . − . [ +2 . − . ] χ / dof cstat 19.97/28 Table 6.
Summary of the apec fitting to the group gas surround-ing 3C 35 for
XMM and
Chandra . The Galactic neutral hydrogencolumn density is frozen for all fits. N is the apec normalizationquoted in units of 10 − (1 + z ) R n e n p dV/ πD L , corrected forexcluded volumes.Model parameter Chandra XMM kT (keV) ≈ . +3 . − . N apec (10 − cm − ) 6 . +7 . − . . +8 . − . χ / dof cstat 44.38/58 in § χ = 0 . XMM data are in good agreementwith the
Chandra data: we measure a group gas temperatureof kT = 0 . +3 . − . keV. We assumed the group gas occupies asphere of radius of 400 ′′ centred on 3C 35 ( V ≈ . × m )and assumed constant density, as the poor count rates pre-vent us extracting a density profile. We scaled the XMM measurement extracted from the smaller region up to thisvolume (a factor of roughly 10), accounting for projectioneffects in measuring the group-gas emission 4 ′ away fromthe centre of the sphere. The 0.4-7.0 keV luminosity of thegroup medium is then L group = 7 . +10 . − . × erg s − . Thisvalue is consistent with Chandra , although the assumptionof constant density used in scaling the
XMM emission is veryrough.
Table 7.
Physical parameters based on the X-ray emission fromthe lobes of 3C 35. The photon index is fixed at the radio syn-chrotron index, errors correspond to 90 per cent confidence for oneinteresting parameter. L X is the absorption-corrected 0 . − . Chandra XMM
Γ 1.7 (frozen) 1.7 (frozen) S (nJy) 39 . ± .
88 42 . ± . +4 . − . ] χ / dof 22.31/21 214.39/216 L X (10 erg s − ) 3 . +0 . − . . ± . ± . B me (10 − T) 1.58 ± .
04 1.58 ± . P me (10 − Pa) 0.72 ± .
04 0.72 ± . B iC (10 − T) 0.90 ± .
12 0.87 ± . u B (10 − J m − ) 3.22 +0 . − . +0 . − . u p (10 − J m − ) 3.05 +1 . − . +1 . − . P (10 − Pa) 1.12 +0 . − . +0 . − . The lobe X-ray spectra are well described by a powerlaw (Table 4); the agreement between X-ray spectral index( α Chandra = 0 . +0 . − . , α XMM = 0 . ± .
11) and the radiospectral index between 73.8 and 327.4 MHz (0 . < α r < . α r = 0 .
7. The number den-sity spectrum of radiating electrons was assumed to be apower law of the form ∝ γ α r +1 , where the Lorentz fac-tor is 10 < γ < . Source volume was calculated as75 × kpc , and the radio flux density was estimatedas S R = 2 . ± . B me = 0 .
16 nT (in agree-ment with O10) assuming a filling factor of unity, no rel-ativistic protons, and no relativistic bulk motions. Table 7shows the minimum-energy parameters .We measured the corresponding inverse-Compton X-ray flux density (assuming CMB seed photons) from thepower-law fit to the lobes with a photon index of 1.7. Weassume all X-ray emission from this region is from the iCmechanism. The 1 keV X-ray flux densities are shown inTable 7; we calculated the magnetic field in the lobes tobe B iC ≈ .
09 nT, which is about 1.8 times lower than At minimum energy, the energy in the magnetic field u B = u p ( α + 1) /
2, and total pressure (Pa) P me = (3+ α ) B α ) µ , where B is in Tesla. c (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 estimated under the minimum-energy condition, consistentwith Croston et al. (2005) who reported electron dominance( B iC ∼ . B me ) in the lobes of 33 other FR II radio galaxiesand quasars. We then find the internal pressure in the lobesto be 1 . × − Pa, slightly higher than predicted underthe minimum-energy assumption.Using equations 2 and 3 ( § § . +7 . − . × − Pa. The lobes havecomparable pressure with the external medium, within theuncertainties, although the assumption of hydrostatic equi-librium does not exclude some additional pressure source inthe lobes. Strong electron or magnetic dominance are dis-favoured because of the constraints from the iC emission. Arelativistic proton component or decreased filling factor canprovide additional pressure, as can extension of the electronspectrum to lower energies. A possible alternative sourceof pressure in the lobes is a population of old electrons nearthe inner regions of 3C 35, steepening the electron spectrum.The power-law component of the gas-belt fit adds favour tothis explanation, since we find it can be well described byincreasing α towards the inner regions of 3C 35. This is ex-amined in more detail in § R , the ratio of ob-served to predicted X-ray flux at equipartition. The 1.4 GHzradio map predicts a minimum-energy magnetic field of B me = 0 .
16 nT and therefore R ≈ .
8, i.e., the lobes areelectron dominated. This falls within the narrow distribu-tion of R -values found by Croston et al. (2005).We measured the temperature of the group mediumto be kT = 0 . +3 . − . keV. The ratio of surface brightness,Σ, of the lobe X-ray emission to the surface brightness ofthe group medium differs by more than an order of magni-tude, Σ lobesX / Σ groupX ≈
30. The group gas surrounding 3C 35has a bolometric luminosity of L bol ≈ erg s − assum-ing a temperature of 0.9 keV. Croston et al. (2008) exam-ined the luminosity-temperature relation for a sample of 9low power FR I radio galaxies in group environments com-pared to radio-quiet groups. They found that groups con-taining extended radio sources have a higher temperaturefor a given X-ray luminosity. Their best fitting relation forthe radio-quiet groups predicts a group gas temperature of ≈ . L X − T X relation ofCroston et al. (2005). The energy required to raise the tem-perature of the group gas within a sphere of radius 400 ′′ encompassing 3C 35 is E req ≈ × J. The total energyavailable from the current episode of radio activity, esti-mated as 4
P V (this includes the energy stored in the lobes,which is yet to affect the environment) using the total lobevolume and the pressure in the lobes derived from the X-ray data (Table 7) is 10 J. Thus the radio source is suffi-ciently powerful to heat the group gas by 0 . Table 8.
A comparison of results from the
XMM and
Chandra data (assuming the best fit to the gas belt; apec + power-law)with those from IS11 in the
Suzaku fitting. IS11 attribute the softthermal component of their fit to thermal plasma emission fromthe host galaxy of 3C 35, which we compare to our measurementsof the gas belt.Parameter
XMM Chandra Suzaku L X (10 erg s − ) 4 . ± . . ± . . +0 . − . Lobes L X (10 erg s − ) 3 . ± . . ± . . +0 . − . Gas belt/host galaxy kT (keV) 1 . ± . . +1 . − . . ± . L X (10 erg s − ) 2 . +1 . − . . +1 . − . . +2 . − . IS11 integrate the 0 . −
10 keV
Suzaku spectrum within alarge rectangular region containing the whole radio structureof 3C 35. They report iC emission associated with the lobes,as well as a soft thermal component attributed to thermalplasma emission from the host galaxy. We have separatedthese two components based on higher resolution
Chandra data and more sensitive
XMM data. Table 8 shows a com-parison of the luminosities measured from each componentacross the three detectors. Both this paper and the IS11analysis require careful subtraction of the X-ray background.Given this difficulty, the total X-ray luminosities measuredfor the lobes and gas belt of 3C 35 are in reasonable agree-ment.However, the 0.5-10.0 keV absorption-corrected lumi-nosity of the lobe iC component measured by IS11 is a fac-tor of about 2 lower than the emission measured from the
XMM and
Chandra data. The IS11 thermal component isjust consistent within errors with the
XMM luminosity ofthe gas belt and we derive a gas-belt temperature similar tothat of the IS11 thermal component.In § L group = 7 . +10 . − . × erg s − at a tem-perature of 0 . Suzaku , IS11over-allocated power to the mekal component of their fitand underestimated the total emission in the lobes, and in-cluded a contribution from group gas to the thermal com-ponent of their fit. c (cid:13) , 000–000 E. Mannering et al.
Table 9.
Temperature, density and pressure of the X-ray emis-sion from the lobes, belt and group medium from the
XMM data.The error in the volume is assumed to be 10% for the lobes andextended group gas, and 50% for the gas belt, to reflect our un-certainty on its shape. L X is the absorption-corrected 0.4-7.0 keVluminosity attributed to the thermal component of the gas-beltfit, the iC component of the lobe fit, and the thermal componentof the group-gas fit, corrected for excluded volumes.Parameter Lobes Gas belt Group gas kT (keV) - 1 . +0 . − . . +3 . − . Volume (10 m ) 2.2 ± . . ± . × − ± . n p (m − ) - 440 ±
190 60 +35 − P (10 − Pa) 0.12 ± .
03 2 . +1 . − . . +0 . − . L X (10 erg s − ) 3 . ± . . +0 . − . . +1 . − . It is clear from the fits in § apec model with Galactic N H describesthe data (reduced χ = 0 .
84) and the derived parameters areconsistent across the two instruments, although the temper-ature is not well constrained. Similarly a power-law model(Γ = 2 .
1) also represents the data (reduced χ = 0 . . apec model (with a spectral index consistent withthe iC emission detected in the lobes) improved the fit.The temperature of a purely thermal model ( kT ≈ . . apec +power-law. Both the Chandra and
XMM results areshown in Table 5, but we use the more precise
XMM resultsin subsequent analysis.Making use of Worrall & Birkinshaw (2006), we calcu-late a density and pressure of the X-ray-emitting gas belt.The emission measure, defined in terms of the normalisationfactor returned by xspec , N , is10 N = (1 + z ) R n e n p dV πD (1)where D L is the luminosity distance (9 . × cm), V isthe volume (cm ) and n p (cm − ) is the proton density. If n e ≈ . n p due to the presence of elements heavier thanhydrogen in the gas, the proton number density for a region of uniform density is given by n p ∼ s N 4 πD (1 + z ) .
18 V (2)if we assume that n p is constant over the volume V. Thepressure, P (Pa) is given by P = 3 . × − ( n p kT ) (3)where n p is in m and kT in keV (Worrall et al. 2012). Weestimated the volume of the gas belt by assuming it is com-prised of a disk of radius r = 100 ′′ and height 50 ′′ . Thephysical parameters for the gas belt are in Table 9. The gas belt appears to be over-pressured compared to thegroup gas by an order of magnitude, although the errors aresuch that the belt and group gas could be in pressure bal-ance. The belt is more clearly over-pressured (by an extrafactor of about 2) if we do not allow for a power-law com-ponent. This suggests that the belt is expanded gas fromthe central regions, rather than the backflow of group gasdisplaced by the lobes. While these scenarios would suggestdifferent temperature structures in the belt, the errors onthe temperatures are not well constrained and so we cannotinvestigate the origin of the belt in this way.A comparison of the mass of gas in the belt with that ex-pected in an elliptical galaxy provides a useful clue to its ori-gin. An elliptical galaxy will typically contain a hot gaseouscorona, ( kT ∼ &
1% by AGN jetswould raise the temperature of the X-ray coronal gas tothat of the surrounding ambient medium, but the jets tun-nel through the cool core with little interaction. Coronae,therefore, have posed problems for feedback models, sincethey can apparently provide enough gas to fuel an AGN forlong periods, without being heated and swept away by theresultant radio jets. We investigate whether the belt may bea disrupted corona, the destruction of which would regulatethe supply of distant gas to the central engine.S07 studied the coronae of 157 early-type galaxies, 16of which were identified with radio-loud AGN ( L . > W Hz − ). Of the 27 coronae that were resolved in thesample, S07 do not state what fraction host radio-loud AGN.For the resolved subset, they found a typical coronal radiusof 1-4 kpc, although several extend out to ∼
10 kpc. Thecoronal X-ray gas masses were generally in the range 10 . − M ⊙ . Following their sample, we use the term ‘galacticcorona’ to refer to any component of gas with temperature0 . . kT . . . M ⊙ and a radius .
10 kpc.We calculate the mass of the gas belt in 3C 35 to be M belt ≈ (3 ± × M ⊙ , two orders of magnitude abovethe coronal masses in the S07 sample. One of the mostmassive resolved coronae in the S07 sample was that ofNGC 7720, with a gas mass of ≈ M ⊙ . NGC 7720 hostsa large FR I wide-angle tail source, about 400 kpc in size, c (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 and has a high 1.4 GHz radio power similar to that of 3C 35.The discrepancy in mass estimates could be attributed tothe larger size and higher X-ray luminosity of 3C 35 com-pared to NGC 7720, as well as extra mass accumulated viasweeping up IGM gas during expansion.Sun (2009, hereafter S09) expanded on the S07 sam-ple of radio AGN hosting galactic coronae, presenting re-sults for 52 galaxies with L . > W Hz − . Figure1 of S09 shows that BCG (brightest cluster galaxies) in kT < L X /L . space separated from systems with largecool cores (LCC). The 0.5-2 keV luminosity of the gas belt is L belt = 1 . +0 . − . × erg s − , and the 1.4 GHz flux of 3C 35, S = 2 . . × W Hz − .If the gas belt originated from a corona, we would expectthat during expansion its density, and therefore its lumi-nosity, would have decreased. At earlier times, we wouldexpect L corona ∝∼ L belt V belt /V corona (the weak dependencyon temperature is ignored). If we adopt the maximum ra-dius of a corona measured in the S09 sample of coronae(25 kpc) , then for 3C 35 L corona = (5 ± × erg s − .S09 define a dividing line between systems with large coolcores (LCC, sources with cooling cores of cluster/group gasand a central isochoric cooling time of less than 2 Gyr) andsystems with galactic coronae, according to their 0.5-2.0 keVluminosity; L . − . ∼ × erg s − . 3C 35 is beyondthe S09 boundary for galaxic coronae. While our calculated L corona places 3C 35 in the region allowed for a LCC, theextended group environment is weak ( § § z = 0 . kT = 1 . +0 . − . keV, a bolometric luminosityof 5 . × erg s − (twice that of the gas belt in 3C 35),as well as a slightly smaller gas mass of (2 − × M ⊙ .H07 suggest the ridge is intrinsic to the system (rather thanthe interaction of the radio galaxy with the IGM), and mayhave even been present before the radio source switched on.It aligns well with the local distribution of galaxies andstarlight (unaffected by the radio galaxy), so that a frac-tion of the gas in the ridge may have been stripped from although those systems with r >
10 kpc would probably bebetter classified as small cool cores the galaxies that merged to form the current host galaxy of3C 285. A case where an active merger aligns the gas contentof several galaxies orthogonal to the radio lobes is 3C 442A(Worrall et al. 2007).To check the possibility that the gas belt in 3C 35 isthe combined ISM from several nearby galaxies we used theNASA Extragalactic Database (NED) to search for galax-ies within 6 ′ (450 kpc) of 3C 35. A single galaxy is detectedin the 2MASX survey , at a separation of 5 ′ , and does notappear to be associated with 3C 35 (although there is no red-shift). By contrast, H07 find 12 objects classified as galaxieswithin a 5 ′ (450 kpc) radius of 3C 285, 6 of which are opti-cally confirmed (via SDSS) to be associated with member-ship of the group.We examined the galaxy previously identified inthe HST /WFPC2 and
DSS data ( § Spitzer (Fazio et al. 2004). We find the new galaxy to be about41 kpc offset from 3C 35 (assuming it lies at a similar red-shift), and at least 1.6 mag fainter. It is therefore likely tobe a companion, but spectroscopic follow-up is needed toconfirm this galaxy’s association with 3C 35. Thus, although3C 35 may be associated with a few fainter galaxies, the largemagnitude difference suggests that it is relatively isolated,certainly compared to 3C 285.The potential companion galaxies of 3C 35 are far toosmall to contribute significant gas or cause major distur-bance in 3C 35’s gas. If the gas belt were a ‘ridge’ in theIGM, we would expect to have detected an X-ray cool corein the host galaxy, due to the prevalence of such cores inradio-loud systems (Sun et al. 2007; Sun 2009). We find noevidence for a cool core in the host of 3C 35 ( § Spitzer
IRAC data, and 3C 35 has an extensiveenvelope which is not well fitted, so that the two galax-ies may have a total brightness difference of about 2 mag.There are no other galaxies in the field of comparable lumi-nosity. The combination of a luminous central galaxy (the2MASS H -band host galaxy magnitude of 3C 35 of − . . × W Hz − RA , δ = 01 h m s . d m s . (cid:13) , 000–000 E. Mannering et al. (only one fossil group in the Hess sample is more radio lu-minous), yet the X-ray luminosity of the belt is low com-pared to the other fossil-group halos (Figure 12, right). Butthe belt volume for 3C 35 is also large (Figure 12, left). If3C 35’s belt arises from expansion of a fossil-group atmo-sphere, we would expect L FG ∝∼ L belt V belt /V FG , and thisline is plotted in the Figure. The increased luminosity of thehalo for plausible expansion factors of about 10 brings thesystem more in line with the Hess radio-loud fossil groups(Figure 12, right).Therefore, we believe the most likely explanation forthe appearance of a relatively bright gas belt in such a poorenvironment is via the driven expansion of fossil group gas.We may be witnessing the extreme end of feedback, wherethe radio lobes have almost destroyed the atmosphere of afossil group that fed the radio AGN.The lobes and gas belt appear to be closely related mor-phologically (Fig. 3), and have likely influenced one anotherduring their formation; we expect comparable ages for bothfeatures. The pressure of the gas belt is higher than that ofthe lobes, thus the belt is capable of having driven the radiolobes out of the central regions of the source, explaining the“pinch” in the contours of the radio structure between thelobes (Fig. 2). The pressure ratio (see Table 9) between thegas belt and the group medium is 11 +27 − . Using the Rankine-Hugoniot conditions for a strong shock in a monatomic gas(See § . +2 . − . permits much slower expansions. Thetemperature and density predicted for the shocked groupgas by expansion at a Mach number of 3, are allowed withinthe large errors .The sound speed in the group gas around 3C 35 is c s = 0 .
54 ( kT / keV) . kpc Myr − (4)or c s = 0 . +1 . − . kpc Myr − . Assuming a Mach number of 3, v adv = 1 . +3 . − . kpc Myr − . The expansion timescale of thebelt is then t belt = 80 +180 − Myr, for a belt radius of 130 kpc( § t synch = 143 ±
20 Myr. The errorshowever, are too large to properly test the idea that the radiolobes and gas belt evolved simultaneously.A fit to the gas belt of purely thermal origin (Table 5)finds a higher value of the temperature of kT ≈ . n p ≈
680 m − and a subsequent pres-sure of P gas ≈ . × − Pa. The belt would then be over-pressured relative to the group medium by a factor of about30. Following § M ≈
Some luminous elliptical galaxies are isolated, or reside inlow-density environments, and may be mis-classified as fos- The predicted temperature ratio is about 3.7; we measure T belt /T group = 1 . +1 . − . ; similarly, ρ /ρ ≈ .
0, and we measure ρ belt /ρ group = 7 . +9 . − . sil systems. Jones et al. (2003) note that there are few fieldellipticals with L X > erg s − that are not in the cen-tres of groups or clusters and therefore have a contribu-tion to their X-ray halo from IGM gas. Mulchaey & Jeltema(2010) study the L X − L K relationship for a sample of fieldearly-type galaxies, and detect an X-ray thermal compo-nent in all field galaxies with K -band luminosities above L X ≈ . × L K , ⊙ (more massive galaxies are able toretain their halos). The 2MASS K -band magnitude of3C 35 is 12.17 (corrected for Galactic extinction), whichcorresponds to L K ≈ . × L K , ⊙ . Using the corre-lation of Mulchaey & Jeltema (2010), a 0.5-2.0 keV lumi-nosity of L haloX ≈ (1 . − . × erg s − is predictedfor 3C 35 from its stellar properties. We measure L beltX =1 . +0 . − . × erg s − . A typical radius of the X-ray gas inelliptical galaxies is 30 kpc (Memola et al. 2009), which cor-responds to a volume 30 times less than that of the gas belt.If 3C 35’s belt arises by expansion of ISM in a field ellipti-cal, then we would expect L haloX = 3 . +2 . − . × erg s − , anorder of magnitude larger than the maximum 0 . − . × erg s − ,Mulchaey & Jeltema 2010). Fossil groups have been shownto have systematically higher L X and L B than field ellipti-cals and non-fossil groups (Memola et al. 2009; Hess et al.2012), thus we believe that the gas belt is likely to be ac-cumulated ISM plus IGM from a merged group of galaxies,rather than a hot gas halo surrounding an isolated field ellip-tical. To firmly distinguish between the two possible evolu-tionary scenarios (fossil group or isolated galaxy), a detailedinvestigation of the dark matter potential around 3C 35 isnecessary. The mass and age of the gas belt, as well as the morpholog-ical structure of the radio source provide a convincing argu-ment that the central AGN activity has disturbed the gasmorphology of the fossil group’s X-ray halo, causing an ejec-tion event and thus affecting the supply of large-scale gas tothe central engine. We now address the physical plausibilityof this scenario, given the power of the AGN outburst.If we assume the 3 × M ⊙ of X-ray gas in the beltis being ejected from the galaxy, then it has been given akinetic energy U belt ≈ J. The total energy availablefrom the current episode of radio activity was estimated in § U lobes ≈ J. Thus the lobes can easily supplyenough energy to lift the belt gas.Following Buttiglione et al. (2010), we derived the blackhole mass from the host luminosity using the correlation ofMarconi & Hunt (2003)log( M BH ) = − . − . M H (5)where the optical host magnitude of 3C 35 is − .
17 (2MASS H -band, Skrutskie et al. 2006), giving a black hole mass of M BH = 10 . M ⊙ . How much of the ejected gas must remainin order to fuel the AGN? To produce energies of 10 Jin the radio structure, a mass of at least 10 . M ⊙ musthave been accreted onto the central black hole. If the BH over a 23 . ′′ × . ′′ integration areac (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 Figure 12.
Left:
Halo 0.5-2.0 keV luminosity against halo volume for fossil groups from Hess et al. (2012); Santos et al. (2007) with thegas belt of 3C 35 marked with a diamond symbol. Those systems with a radio-loud ( > W Hz − ) AGN are shown in red. The lineshows the luminosity and volume of the gas belt at earlier times, assuming L FG ∝ L belt V belt /V FG . The errors on the volume estimatedfor the gas belt reflect our uncertainty on its shape. Right: is accreting from a thin disk at 1% efficiency , 10 . M ⊙ isrequired after the ejection of the X-ray halo, i.e., less than1% of the gas mass of the belt, if left behind in the core of3C 35, would be sufficient to fuel the radio structure.It is conceivable, that the dynamic gas in the belt orig-inated as an X-ray halo of a fossil group, accumulated as aresult of the merger of L ∗ galaxies at z ≈
1, and was drivenoutwards by the expanding radio jets before continuing toexpand outwards at the sound speed. Its mass is comparableto that of the X-ray emitting IGM in a fossil group system,and the luminosities of the gas belt and radio galaxy are inline with brightest fossil group radio galaxies. The appar-ent influence of the radio structure on the belt implies thetwo have likely affected one another over the past roughly150 Myr. At earlier times the lobes would have been over-pressured relative to the belt, and probably shaped its for-mation. The similar evolution timescales of the radio sourceand gas belt support the idea that their evolution was cou-pled.
The non-thermal component required in fitting the gas beltimplies a detection of inverse-Compton emission. The ratioof the surface brightness of the lobe X-ray emission (Table 4)to the X-ray surface brightness attributed to the power-lawcomponent in the belt is Σ lobesX / Σ belt X ≈ .
0. This is inconsis-tent with the radio data; from the VLA 1.4 GHz radio map,the flux density in the gas-belt region is 30 ± lobes R / Σ belt R ≈ .
8. We would expect the X-rayand radio surface brightness ratios to be comparable, if themagnetic field strength and number densities of both the iCand synchrotron emitting electron populations are constant An efficiency of up to about 0.4 is possible for > M ⊙ BH(Davis & Laor 2011). across the lobe and gas belt. In other words, in compari-son with the lobes as a whole, the belt region appears tobe more successful in producing power-law X-ray emissionthan would be expected on the basis of its radio flux.There are three possible scenarios to explain the vari-ation of the X-ray/radio emission across the lobes and gasbelt of 3C 35. (i) another process is boosting X-ray emis-sion from the central regions. (ii) a population of low-energyelectrons in the central regions is contributing to the iCmechanism, or (iii) the magnetic field strength is varyingas a function of position, such that the inner lobes are moreelectron-dominated.Nuclear photons ( γ ∼ ) and starlight are importantas extra sources of seed photons for iC scattering in somesources (e.g., Brunetti et al. 1997; Stawarz et al. 2003), andcould provide a boost in the non-thermal X-ray emissionseen from the belt of 3C 35. Inverse-Compton scattering ofnuclear photons should dominate at (projected) distances R kpc from the core; R kpc < L (1 − µ ) (1 + z ) − (6)where L is the isotropic nuclear luminosity (10 erg s − )in the FIR to optical band and µ is the cosine of the anglebetween the radio axis and the line of sight (Comastri et al.2003). We assume µ = 90 deg for 3C 35. Chiaberge et al.(2000) find an upper limit on the optical luminosity of3C 35 to be L ˚ A < . × erg s − Hz − using Hub-ble Space Telescope (HST) data from the Wide Field andPlanetary Camera 2 (WFPC2). Using infrared UKIRT ob-servations of 3C 35 (Lilly et al. 1985), in four bands (JHKL)with fixed aperture size of 7.5 ′′ , we deduce the integratedIR to optical luminosity of the core region to be L IR − opt < . × erg s − .Since the gas belt extends over 100 kpc in radius, but R kpc . .
3, we conclude that iC scattering of nuclear pho-tons does not contribute significantly to any power-law X-ray emission from the belt. It is also unlikely that a boost in c (cid:13) , 000–000 E. Mannering et al. gas-belt iC emission is due to the upscattering of starlightfrom faint galaxies in close proximity to 3C 35, although wecannot rule out some iC emission originating from this pro-cess.Scenario (ii), in which differences in the electron spec-trum are responsible for the variation in X-ray/radio ra-tio, seems favourable. A population of lower energy elec-trons may exist in the gas-belt region. These would raisethe X-ray iC emission, but not significantly contribute tothe synchrotron mechanism at 1.4 GHz. A steeper electronspectrum near the core (more older electrons) may add anadditional component of pressure to the gas belt. A typicalLorentz factor for a population of electrons emitting syn-chrotron radiation at 1.4 GHz in a fixed magnetic field of0 .
09 nT is γ ∼ . . In the case of CMB seed photons, theiC emission requires electrons with γ ∼ , the less ener-getic end of the synchrotron emitting population.Radio images of O10 (figure 1) support this explanation.The emission at low frequency shows a steeper radio spectralindex near the core of 3C 35: α r changes from 0.6 to 1.7between 327 MHz and 1.4 GHz in the inner region of thelobes. A steeper spectrum implies more low-energy electronsnear the centre of 3C 35 than we estimated in § α = 0 . α = 1 . B iC = 0 .
09 nT, we would expect to measure an increasedflux at 1 keV of 9.9 nJy. This would give a surface brightnessratio of Σ lobesX / Σ belt X ≈ . B iCbelt = 0 .
02 nT (for α = 0 . α = 1 . In this paper, we have suggested that 3C 35 is destroyingthe surrounding IGM. We have found that the gas belt hasan age comparable to that of the radio source and thatthe morphologies of the radio lobes and gas-belt point toco-evolution. The gas belt is unlikely to be asymmetrically aligned group gas accumulated from galaxies in the plane ofthe belt, as 3C 35 resides in a poor environment, neither isit consistent with disrupted galactic coronal gas. We favourthe argument that the gas belt is expanded fossil group gas,driven outwards by the radio lobes. We have demonstratedthat the radio source is powerful enough to unbind the groupgas from the central galaxy potential well, and that only 1%of the mass of the belt is required to have fuelled the radiostructure.We find the pressure of the external group medium tobe up to two orders of magnitude lower than the gas belt,implying the belt is an expansion of gas from the centralregions, rather than a backflow of heated IGM displacedby the radio lobes. However, without robust temperaturemeasurements across the gas belt, we cannot completely ruleout the latter scenario. A backflow could provide fuel to theAGN, continuing its outburst. This would be the type offeedback loop on scales of hundreds of kpc necessary forradio sources to respond to their large scale environments(a weakness of feedback models has been the absence ofa responsive mechanism over large distances, Dubois et al.2010).The detection of radio-loud AGN in fossil groups(Hess et al. 2012, and this work) implies these systems arenot as quietly and passively evolving as previously thought.The short lifetime of radio synchrotron emission (less than afew hundred Myr) after the source has turned off (Sun et al.2004, and references therein) compared to the fossil groupformation timescale inferred from simulations (several Gyr),as well as the long cooling time for the X-ray emittinggas ( t cool of the gas belt is estimated as 20 −
50 Gyr;Worrall & Birkinshaw 2006, figure 5) suggests that a rec-curing process drives AGN activity.The mechanisms governing the fuelling of AGN, andthe processes involved in turning nuclear activity on and offin these dynamically old systems, remain unclear. In 3C 35it appears not to be merger-driven. We see no stucture inthe
HST and
Spitzer images that might support a majormerger in this system for more than about 1 Gyr. The op-tical nuclear spectrum of 3C 35 (Buttiglione et al. 2009) isabsorption-line dominated, with no sign of optical AGN ac-tivity or a recent strong burst of star formation (in the last ≈
20 Myr).A relationship between AGN and their host galax-ies has been well documented (BH-bulge mass correlation,e.g. Ferrarese & Merritt 2000) and discussed in terms ofAGN-induced feedback through several proposed mecha-nisms; AGN-driven winds, starburst-driven superwinds, andjet-induced outflows (e.g., Holt et al. 2008; Hambrick et al.2011, and references therein). AGN feedback remains themost likely mechanism to balance radiative cooling of theIGM in systems with large cool cores (Sun 2012, and refer-ences therein). There has been no clear example of feedbackbetween radio sources hosted in fossil systems and the largescale gas in the literature. 3C 35 could be the first case wherewe are seeing the X-ray halo of a fossil group during the endphase of disruption. c (cid:13) , 000–000 alaxy gas ejection in radio galaxies: the case of 3C 35 We have discussed the extended X-ray emission seen in new
XMM and archival
Chandra observations of the FR II giantradio galaxy 3C 35. The properties of the lobes are consistentwith earlier work (IS11, O10), and show a clear detectionof inverse-Compton emission. The implied departure fromequipartition is comparable to the range observed in othersources. We report detections of an extended group-scaleenvironment, and place weak constraints on the non-thermalemission from the nucleus.More importantly, we report the detection of a gas belt,orthogonal to the radio lobes and lying between the radiolobes seen in our 1.4 GHz radio map. We conclude that theX-ray emission from the belt is most likely a combination ofthermal and power-law emission. The thermal componentis likely to be from hot gas originating in ∼ L ∗ galaxiesthat merged over a Gyr ago and was driven outwards bythe expanding radio structure before continuing to expandat its own sound speed. The age of the radio structure of3C 35 is consistent with the lifetime of the expanded beltfeature. The higher than predicted (from 1.4 GHz data) X-ray flux of its non-thermal component is attributed to apopulation of low energy electrons boosting iC emission incombination with positional variation of the magnetic field.3C 35’s power is sufficient to place it close to the peak of thedistribution of radio power in the local universe, and thusin the critical range where radio-mode feedback should beseen (Worrall 2009), if it is indeed an important mechanismin the required regulation of structure evolution.In order to support the mechanism being common andto determine a possible age/power/environmental depen-dency on the ejection of the X-ray halos of fossil groupsby radio-loud AGN requires a sample of potentially simi-lar cases. We are currently investigating a sample of ninesuch sources (Mannering et al. in prep). Simulations wouldprovide beneficial insight into this phenomenon. This paper is based on observations obtained with
XMM-Newton and
Chandra . XMM-Newton is an ESA science mis-sion with instruments and contributions directly funded byESA Member States and NASA. We thank the CXC forits support of Chandra observations and data analysis, andthe SAO R&D group for DS9 and FUNTOOLS. This workhas also used data from the VLA. NRAO is a facility ofthe National Science Foundation operated under coopera-tive agreement by Associated Universities, Inc. EM thanksthe STFC for support.
REFERENCES
Arnaud K. A., 1996, in G. H. Jacoby and J. Barnes ed.,Astronomical Data Analysis Software and Systems VVol. 101 of Astronomical Society of the Pacific Confer-ence Series, XSPEC: The First Ten Years. p. 17Bell, E. F., McIntosh, D. H., Katz, N., & Weinberg, M. D.2003, APJS, 149, 289Belsole E., Worrall D. M., Hardcastle M. J., 2006, MNRAS,366, 339 Bˆırzan L., McNamara B. R., Nulsen P. E. J., Carilli C. L.,Wise M. W., 2008, ApJ, 686, 859Blanton E. L., Randall S. W., Douglass E. M., SarazinC. L., Clarke T. E., McNamara B. R., 2009, ApJL, 697,L95Brunetti G., Setti G., Comastri A., 1997, AAP, 325, 898Burns J. O., Owen F. N., 1977, ApJ, 217, 34Buttiglione, S., Capetti, A., Celotti, A., et al. 2010, AAP,509, A6Buttiglione, S., Capetti, A., Celotti, A., et al. 2009, AAP,495, 1033Carter J. A., Sembay S., Read A. M., 2011, AAP, 527, A115Chiaberge M., Capetti A., Celotti A., 2000, AAP, 355, 873Comastri A., Brunetti G., Dallacasa D., Bondi M., PedaniM., Setti G., 2003, MNRAS, 340, L52Croston J. H., Hardcastle M. J., Birkinshaw M., WorrallD. M., Laing R. A., 2008, MNRAS, 386, 1709Croston J. H., Hardcastle M. J., Harris D. E., Belsole E.,Birkinshaw M., Worrall D. M., 2005, ApJ, 626, 733Dahlem M., 1999, ESA: XMM-Newton SOC “XMM-Newton Users Handbook, 1999, Issue 1.1”Dariush, A., Khosroshahi, H. G., Ponman, T. J., et al. 2007,MNRAS, 382, 433Davis S. W., Laor A., 2011, ApJ, 728, 98Dickey J. M., Lockman F. J., 1990, ARAA, 28, 215D’Onghia, E., Sommer-Larsen, J., Romeo, A. D., et al.2005, APJL, 630, L109Dubois, Y., Devriendt, J., Slyz, A., & Teyssier, R. 2010,MNRAS, 409, 985Evans D. A., Worrall D. M., Hardcastle M. J., Kraft R. P.,Birkinshaw M., 2006, ApJ, 642, 96Fanaroff B. L., Riley J. M., 1974, MNRAS, 167, 31PFazio, G. G., Hora, J. L., Allen, L. E., et al. 2004, ApJs,154, 10Ferrarese L., Merritt D., 2000, ApJl, 539, L9Floyd D. J. E., Axon D., Baum S., Capetti A., ChiabergeM., Macchetto D., Madrid J., Miley G., O’Dea C. P., Perl-man E., Quillen A., Sparks W., Tremblay G., 2008, ApJS,177, 148Forman W., Jones C., Churazov E., Markevitch M., NulsenP., Vikhlinin A., Begelman M., B¨ohringer H., Eilek J.,Heinz S., Kraft R., Owen F., Pahre M., 2007, ApJ, 665,1057Forman W., Jones C., Tucker W., 1985, ApJ, 293, 102Goodger, J. L., Hardcastle, M. J., Croston, J. H., Kassim,N. E., & Perley, R. A. 2008, MNRAS, 386, 337Guindon B., 1979, MNRAS, 186, 117Hambrick D. C., Ostriker J. P., Naab T., Johansson P. H.,2011, ApJ, 738, 16Hardcastle M. J., Birkinshaw M., Cameron R. A., HarrisD. E., Looney L. W., Worrall D. M., 2002, ApJ, 581, 948Hardcastle M. J., Croston J. H., 2005, MNRAS, 363, 649Hardcastle M. J., Kraft R. P., Worrall D. M., Croston J. H.,Evans D. A., Birkinshaw M., Murray S. S., 2007, ApJ, 662,166Hess, K. M., Wilcots, E. M., & Hartwick, V. L. 2012, AJ,144, 48Holt J., Tadhunter C. N., Morganti R., 2008, MNRAS, 387,639Isobe N., Makishima K., Tashiro M., Hong S., 2005, ApJ,632, 781Isobe N., Seta H., Gandhi P., Tashiro M. S., 2011, ApJ, c (cid:13) , 000–000 E. Mannering et al. c (cid:13)000