© 1998 RAS

Mon. Not. R. Astron. Soc. 298, 966–976 (1998)

A radio-jet–galaxy interaction in 3C 441

Mark Lacy, Steve Rawlings, Katherine M. Blundell and Susan E. Ridgway

Astrophysics, Department of Physics, Keble Road, Oxford OX1 3RH

Accepted 1998 February 27. Received 1998 February 9; in original form 1997 August 27

A B STR ACTMultiwavelength imaging and spectroscopy of the z\0.708 radio galaxy 3C 441 anda red aligned optical/infrared component are used to show that the most strikingaspect of the radio–optical ‘alignment effect’ in this object is due to the interactionof the radio jet with a companion galaxy in the same group or cluster. The stellarpopulation of the red aligned continuum component is predominantly old, but witha small post-starburst population superposed, and it is surrounded by a low-surface-brightness halo, possibly a face-on spiral disc. The [O III] 500.7/[O II] 372.7 emission-line ratio changes dramatically from one side of the component to the other, withthe low-ionization material apparently having passed through the bow shock of theradio source and been compressed. A simple model for the interaction is used toexplain the velocity shifts in the emission-line gas, and to predict that the ISM of theinteracting galaxy is likely to escape once the radio source bow shock has passedthrough. We also discuss another, much fainter, aligned component, and thesubarcsecond-scale alignment of the radio source host galaxy. Finally, we commenton the implications of our explanation of 3C 441 on theories for the alignmenteffect.

Key words: galaxies: active – galaxies: individual: 3C 441 – galaxies: interactions –intergalactic medium – galaxies: jets.

1 INTRODUCTION

The consequences of radio jets impacting on density inho-mogeneities have been invoked to explain many propertiesof high-redshift radio sources, such as bending and asymme-tries in arm length. Evidence for this is seen in the form ofcorrelations between emission-line gas and the side of theradio galaxy with the shorter arm-length, or higher depolar-ization (Liu & Pooley 1991; McCarthy, van Breugel &Kapahi 1991). The extent to which this is related to the so-called ‘alignment effect’, whereby the axis of extended opti-cal continuum and line emission is found to be co-alignedwith the radio axis, is unclear, although many attempts toexplain radio–optical alignments involve a dense externalmedium. Such a medium is needed, for example, to act asthe source of a scattering surface for quasar light from theAGN (Bremer, Fabian & Crawford 1997), or as a mediumfor jet-induced star formation (e.g. Best, Longair & Rott-gering 1996, and references therein).

3C 441 is a z\0.708 radio galaxy with an asymmetricradio structure, which appears to be a rare example of a

radio source with a red aligned component outside the radiolobes. Such a component is clearly hard to obtain in eitherjet-induced star formation or scattered quasar models. Thisred component is seen just beyond the end of the shorter,north-west radio lobe (component ‘c’ of Eisenhardt &Chokshi 1990; see Fig. 1). This lobe appears to possess aradio jet which bends round to the west at the tip of thelobe. Just to the south of the red component, mostly withinthe radio lobe, is an arc-shaped clump of emission-line gasseen in the [O II] 372.7 emission-line image of McCarthy,Spinrad & van Breugel (1995). Spectroscopy of the [O II]emission line by McCarthy, Baum & Spinrad (1996) showsthat this emission-line material is redshifted by 2800 kmsÐ1 with respect to the radio galaxy, and weak emissionextends to just beyond the red component.

In this paper we use our own spectroscopy, infrared andradio imaging, and archive data from the Hubble SpaceTelescope (HST) (also presented in Best, Longair & Rott-gering 1997b) to attempt to explain the properties of 3C 441in terms of models for the interaction of radio jets with theirenvironments. In particular, we address the problems of the

origin of the red continuum light from the aligned compo-nent and the properties of the extended emission-lineregion.

We assume an Einstein–de Sitter cosmology with aHubble constant H0\50 km sÐ1 MpcÐ1.

2 OB SERVATIONS

We obtained an optical spectrum with ISIS on the 4.2-mWilliam Herschel Telescope (WHT) on 1993 August 20with the aim of establishing the nature of component ‘c’.

Both the red and blue arms of ISIS were employed, thered arm detector being the EEV5 CCD and the bluearm detector the TEK1. Observations were taken using a3-arcsec-wide slit at position angle PA\45°, aligned so as topass through both the host galaxy (‘a’ in Fig. 1) and the knot‘c’. The airmass was between 1.02 and 1.06 and the seeing0.9 arcsec during the observations, which consisted of a totalof 3600 s exposure in the blue arm and 3500 s in the redarm.

Observations in J- and K-bands were made with IRCAM3on the United Kingdom Infrared Telescope (UKIRT) on

A radio-jet–galaxy interaction in 3C 441 967

© 1998 RAS, MNRAS 298, 966–976

Figure 1. Overlay of the WFPC2 images of 3C 441 on our VLA 8-GHz B-array radio map. (a) The F555W image smoothed with a s\0.1arcsec Gaussian, with radio contours and magnetic field polarization vectors superposed. The optical components referred to in the text arelabelled. Contours are spaced at intervals separated by factors of 4 from 0.4 mJy beamÐ1, and the polarization vectors are scaled so that alength of 1 arcsec represents 25 per cent polarization. (b) The F785LP image smoothed with a Gaussian with s\0.2 arcsec, with radiocontours again superposed. Contours are spaced at intervals separated by factors of 21/2, starting at 0.3 mJy beamÐ1. The radio beam FWHMis indicated in the bottom-right corner of both plots.

1996 November 6 for 20 and 40 min respectively. Conditionswere non-photometric. The seeing was 0.8 arcsec with thefast guider employed. The K-band image showed patternnoise in the form of stripes across the image, these wereremoved with the IRAF task fit1d.

A J-band spectrum was obtained with CGS4 on UKIRTon 1994 June 17, covering the region of the Ha line. Condi-tions were non-photometric, with some light cirrus present.3C 441 was observed for 3200 s with the long camera and the75 line mmÐ1 grating in second order. The 2-pixel (3-arcsec)wide slit was aligned at the same PA as for the opticalobservations. The combined optical and infrared spectraare shown in Fig. 2.

Radio observations were made in B-array with the VeryLarge Array radio telescope at 8.3 GHz on 1992 January 11.

The exposure time was 1950 s and the bandwidth 50 MHz.The data were calibrated for phase and amplitude in thestandard manner and analysed in AIPS. One of the two IFswas removed due to interference problems. A correction forRicean bias was made to the polarization map using the AIPS

task POLCO.HST imaging data were obtained from the archive,

and consisted of a total exposure of 1700 s in the 555W and785LP filters with the WFPC2. The radio galaxy was centredon the WF3 CCD, on which the scale is 0.1 arcsec pixelÐ1.The filters both contain strong redshift emission lines: theF555W filter contains the [O II] 372.7 emission and theF785LP filter the [O III] 495.9 and 500.7 lines. Fig. 1 showsan overlay of the 8-GHz radio map contours on grey-scalesof the two HST images.

968 M. Lacy et al.

© 1998 RAS, MNRAS 298, 966–976

Figure 1 – continued

3 DISCUS SION

3.1 The continuum knot

Component ‘c’ in Fig. 1(a) and in Eisenhardt & Chokshi(1990), the red aligned knot beyond the north-west radiohotspot, is slightly resolved on the HST images. Its spectralenergy distribution (SED) seems to be consistent with thatof an old stellar population (Fig. 3), although the amplitudeof the 4000-Å break seems lower than that of the model6-Gyr-old stellar population (corresponding approximatelyto the age of the Universe in our assumed cosmology) whichis otherwise a good fit to the SED.

A stellar origin for the light is indeed confirmed by a closeinspection of the spectrum of ‘c’ (Fig. 4) in which stellarabsorption features, in particular the ‘G-band’ from CHabsorption in cool stellar envelopes, can be seen at z\0.714,redshifted by 1000 km sÐ1 with respect to the radio galaxy(z\0.708). Also visible are Hd and the calcium H and Klines, the latter possibly blended with He. The weakness ofthe 4000-Å break, despite the apparent dominance of theSED by an old stellar population, may then be explained bythe existence of a small post-population of A-type stars, andindeed the detection of Hd absorption in our spectrum isconsistent with this. Such populations have been detected inother AGN host galaxies and companions (Tadhunter,Dickson & Shaw 1996; Canalizo & Stockton 1997), and maytrace mergers or interactions in small clusters or groups(Zubludoff et al. 1996).

Whether there is a link between a starburst 1108 yr agoin a close companion or the host itself and the triggering ofthe AGN is unclear – if so, there must certainly be a delaybetween the starburst and the development of a radiosource if typical radio source lifetimes are 1107 yr (Alex-ander & Leahy 1987). In the case of 3C 441, component ‘c’is 2125 kpc from the host galaxy, so it is conceivable that ifits relative velocity with respect to the host in the plane ofthe sky is of the same order as its relative velocity along theline of sight, it could have been close enough to interact 108

yr ago. There is, however, no sign of a corresponding star-burst in the host galaxy, and E+A galaxies are relativelycommon in high-z clusters (Dressler & Gunn 1990; Caldwell& Rose 1997). Furthermore, the high-velocity encounterimplied by this may not have produced sufficient disruptionto initiate the starburst.

3.2 The emission line gas in ‘c’

Contour plots of the 2D spectra around the [O II] 372.7 and[O III] 500.7 emission lines are shown in Fig. 5. There is astriking change in the ionization and luminosity across theposition of the continuum knot: the side nearest to the radiogalaxy has a high luminosity, and a low ionization as mea-sured by the ratio of [O III] 500.7/[O II] 372.7 emission lines(0.35¹0.1). In contrast, beyond the knot the [O III] 500.7/[O II] 372.7 ratio is much higher (a10¹3), and the totalluminosity lower by a factor of 26. There is a velocity

A radio-jet–galaxy interaction in 3C 441 969

© 1998 RAS, MNRAS 298, 966–976

Figure 2. The spectrum of the radio galaxy ‘a’. The ISIS red-arm optical spectrum has been smoothed with a five-pixel (1.4-nm) box-car filterand the CGS4 near-infrared spectrum by a nine-pixel (7.2-nm) one.

970 M. Lacy et al.

© 1998 RAS, MNRAS 298, 966–976

gradient of 2300 km sÐ1 in the line emission across thecontinuum knot, in the sense that the side nearest the radiogalaxy is blueshifted, and that on the far side is, within errorsof 2100 km sÐ1, at rest with respect to the starlight in ‘c’.There is also a faint tail of yet more highly blueshiftedemission (up to 2600 km sÐ1) on the side nearest the radiogalaxy, pointing towards the radio galaxy.

Note that there is no sign of emission lines in Fig. 4, whichused a narrow extraction about the continuum peak. In

Figure 3. The SED of the compact red knot ‘c’. The continuousline is from our own spectroscopy, the near-infrared points fromEisenhardt & Chokshi (1990) (open squares), and our own near-infrared and optical photometry (solid triangles; note that the J andK points have no associated error bars, as the data were non-photometric). The dotted line is the model spectral energy distri-bution of a 6-Gyr-old galaxy in which star formation proceeded ata constant rate in a 1-Gyr burst before ceasing (Bruzual & Charlot1993).

Figure 4. The red-arm spectrum of knot ‘c’ of 3C 441, showing the stellar absorption features mentioned in the text. The raw spectrum hasbeen smoothed by a nine-pixel (2.4-nm) box-car filter and corrected for atmospheric absorption.

Figure 5. 2D spectrum of 3C 441 with the slit aligned along the axisjoining the radio galaxy and ‘c’. Left: the [O II] emission line withwavelength on the horizontal axis and distance along the slit verti-cally. The radio galaxy is to the top, and component ‘c’ below it.Right: the [O III] 5007 emission line. The figures are 11.2 nmÅ30arcsec in size (11.2 nm25280 km sÐ1 close to [O II] and 23930 kmsÐ1 close to [O III]). The contour levels are spaced at intervals of2Å10Ð21 W mÐ2 nmÐ1 per pixel; starting from 2Å10Ð21 W mÐ2

nmÐ1, each pixel was 0.28 nmÅ0.33 arcsec in size.

contrast, in Fig. 5 the extended emission-line flux from ‘c’ isclearly visible and is quite strong when integrated over theentire emission region (Table 1).

3.3 The radio structure

The asymmetric radio structure of 3C 441 is interesting inthe context of the models to explain the aligned emission.The asymmetry in jet brightness either side of the nucleus isvery pronounced, and can be interpreted either as an asym-metry produced by Doppler boosting, or in terms of 3C 441having a radio structure transitional between FRI and FRII,with one side FRII-like and the other more FRI-like. Thelack of radio central component, commonly seen in radiogalaxies with Doppler-boosted one-sided jets (e.g., 3C 22;Rawlings et al. 1995), argues strongly in favour of the latterexplanation, and indeed the flaring of the jet just to thenorth-west of the host galaxy is reminiscent of the ‘Machdisc’ structure seen in the M87 jet (Owen, Hardee & Corn-well 1989).

3.4 The radio galaxy

The spectrum of ‘a’, the host galaxy of the radio source isshown in Fig. 2. It is a typical moderately high-ionizationnarrow-line radio galaxy spectrum, with strong emissionlines superposed on a stellar spectrum dominated by an oldstellar population with a strong 4000-Å break.

Close inspection of the images shows, however, that evenin this case, where the integrated light is dominated by oldstellar populations there is evidence for morphologicalpeculiarity. In particular, there is a distinct blue alignedcomponent to the south-east of the host galaxy peak (Fig.6). Whether this represents a spiral arm, a merger remnantor some form of radio source-induced aligned component isunclear. As its separation from the radio galaxy is onlyabout 0.5 arcsec, it is hard to tell from the spectra whetherit is line- or continuum-dominated, but the more diffuse

material extending 22 arcsec to the south of the radiogalaxy is definitely continuum-dominated. Though muchweaker relative to the smooth underlying host galaxy in theF785LP image, the aligned component is neverthelessvisible, along with some more diffuse aligned emission onthe other side of the peak, to the north-west.

In Fig. 7 the position angle of the host galaxy (derivedfrom the second moments of the flux distribution) is plottedfor various isophotes and apertures. This shows twointeresting aspects of the alignment. First, the alignmentpersists into the K band, suggesting that the aligned light isfairly red. Second, in the HST images there is evidence forisophotal twisting as the aperture size is increased to includethe inner aligned components highlighted in Fig. 6(PA1150°), and later the low-surface-brightness emissionround the host at PA10°, seen best in Fig. 1. The low-

A radio-jet–galaxy interaction in 3C 441 971

© 1998 RAS, MNRAS 298, 966–976

Table 1. Emission-line fluxes from 3C 441.

Notes: za and zc are the mean redshift of the emission line incomponents ‘a’ and ‘c’ respectively. No aperture correctionshave been applied. The optical continuum flux matched thebroad-band flux measurement in a 3-arcsec-diameter apertureto 20 per cent, but the infrared continuum flux was low by 0.5mag, suggesting that an aperture correction of 21.6 should beapplied to Ha+[N II] flux. No attempt was made to deblend theHa+[N II] complex due to the low signal-to-noise ratio of thedetection.

Figure 6. Close-up of the host galaxy (‘a’), grey-scaled so as toshow the aligned components near the nucleus: (a) F555W image;(b) F785LP image. Both images are 10 arcsec2 and have beensmoothed with a s\0.05 arcsec Gaussian.

surface-brightness material to the north-west may be mostlyline emission, although the emission-line image of McCar-thy et al. (1995) shows that the [O II] emission is also roughlyaligned along PA\0°.

3.5 The relationship of the radio and opticalcomponents

The relative astrometry was performed using the results ofthe APM scans of the Palomar Sky Survey plates. The posi-tions of four stars were used to align the corners of the radiomap with the optical images. The uncertainty in the overlayfrom the scatter in the fit was 20.3 arcsec. This astrometryplaces the host galaxy (‘a’ in Fig. 1a) just to the south-east ofthe first appearance of the northern radio jet. Component‘d’ is apparently aligned along the jet direction, though posi-

tioned to the side of it. The point of deflection of the jet isapproximately coincident with the peak of the [O II] 372.7emission, about 3 arcsec south-east of the peak of the con-tinuum knot ‘c’. In the F555W image (Fig. 1a), there is anarc of emission just to the north of the north-west radiohotspot and apparently centred on ‘c’. Fig. 1(b) shows theF785LP image; here there is a halo of diffuse emissionaround ‘c’.

By subtracting a spectrum centred on the continuum knotin an aperture 1.7-arcsec wide from the total spectrum ofthe aligned component ‘c’ (in a 6.8-arcsec-wide aperture),we have been able to estimate the line contribution to thecontinuum magnitudes measured for the nebular regionsurrounding ‘c’. In both filters this is 210 per cent overall,but in the regions of brightest line emission in the inter-action region to the south of ‘c’ our spectrum suggests that

972 M. Lacy et al.

© 1998 RAS, MNRAS 298, 966–976

Figure 7. Position angle as a function of aperture size and iso-photal cut-off for the host galaxy of 3C 441. The isophotal cut-offlevel is in units of the sky noise, and the aperture radii (indicated bycircles of increasing size) are 0.6, 0.9, 1.1, 1.7 and 2.4 arcsec in (a),and 0.3, 0.4, 0.6, 0.8, 1.2, 1.7 and 2.4 arcsec in (b) and (c). The radiosource PA (defined as the PA of the line joining the hotspots) isindicated by a dotted line.

the line contribution of [O II] to the F555W flux rises todominate the overall flux in the filter, consistent with the arcof emission seen in the F555W region just above the radiohotspot consisting entirely of line emission. The continuumemission present in this extraction is bluer than the emissionfrom the knot.

The linear object ‘d’ (Fig. 8) is reminiscent of the alignedcomponent in 3C 34 (Best, Longair & Rottgering 1997a).Like the object in 3C 34, it has no emission lines visibleeither in our spectra or in the narrow-band image of McCar-thy et al. (1995), but is well aligned with the radio structureand lies within the radio lobe. Its optical–near infraredcolours are bluer than ‘c’ (K\20.8; Ja22.7; m785\23.3;m555\25.1 in 3-arcsec-diameter apertures, where m785 andm555 are AB magnitudes in the two HST images), but showa break between the 785LP and 555W filters, consistent witha 4000-Å break at the redshift of the radio source.

3.6 Other companion objects

There are several objects around the radio galaxy whichappear to be members of a group or cluster around it.3C 441 is in a sample of radio galaxies whose clusteringproperties we are currently evaluating, and a formal esti-mate of the clustering amplitude, Bgq, for this radio sourcewill be presented in Wold et al. (in preparation). For thepurposes of this paper, we simply compared the counts of

galaxies with magnitudes m1 to m1+3 (where m1 is themagnitude of the radio source host galaxy) in the objectframe of the F785LP image (the WF3 CCD) with the aver-age of those in the two side frames (WF2 and WF4). Thisrevealed an excess of 11.5¹7.1 galaxies, indicating the pos-sible presence of a cluster, but not at a high confidence level.Clearly, however, this is likely to be an underestimate of thecluster richness, as many cluster members may be outsidethe restricted field of the CCD, and present on the otherframes, increasing the estimate of the background count.

4 INTERPR ETATION OF THE A LIGNED STRUCTUR ES

4.1 Component ‘c’

We argue here that the evidence presented above allows usto deduce that 3C 441 includes an example of a radio-jet–galaxy interaction. (Best et al. 1997b have also proposed ajet–galaxy interaction at ‘c’, but were unable to confirm itusing the HST imaging data alone.)

First, the evidence that the northern radio jet is beingdeflected by an increase in density just south of knot ‘c’ isgiven by the fact that the radio jet appears to bend round tothe west 23 arcsec south-east of ‘c’, consistent with simula-tions of radio jets interacting with gas clouds (Higgins,O’Brien & Dunlop 1996). The polarization B-field vectors(Fig. 1a) are also consistent with an interaction scenario,tending to be perpendicular to the emission-line arc. Therelatively high polarization of the interaction region (up to30 per cent) suggests that we are seeing the interactionthrough the radio lobe (contrast PKS 2250Ð41; Clark et al.1997, in which the polarization of the interacting lobe at 8GHz is only 2.6 per cent, so in that case the emission-linegas is almost certainly covering the radio lobe). Providedthat the rotation measure is R200 rad mÐ2 in the rest frameof the source, the angle of the polarization will be rotated byR5°.

Second, the evidence that the radio jet is shocking theemission-line gas to the south of knot ‘c’ is strong. The factthat emission-line gas is present beyond the interactionregion shows that photoionization is probably responsiblefor most, if not all, of the line emission. The main photo-ionization mechanism could either be photoionization bythe nucleus, or photoionization by hot, post-shocked gas.Hot stars in H II regions of the galaxy could also contributeto the photoionizating flux. Irrespective of the ionizationmechanism, compression of the shocked gas would lead to adecrease in the ionization parameter (Clark et al. 1997), andhence the difference in ionization between the two sides ofknot ‘c’. The velocity structure of the emission lines is alsoconsistent with shocking, with the emission being blue-shifted on the radio galaxy side of the continuum knot,suggesting that the radio jet is approaching the observer.The mean velocity of the knot, a redshift of +1000 km sÐ1

is assumed to come from the random motion of the galaxy ina cluster around 3C 441.

Third, the halo-like appearance of the diffuse emissionaround a central continuum knot and the distribution of[O III] 500.7 emission in the spectrum is hard to explainunless the stars and gas are associated with a galaxy. Thisgalaxy has an red old or E+A-type central concentration of

A radio-jet–galaxy interaction in 3C 441 973

© 1998 RAS, MNRAS 298, 966–976

Figure 8. Close-up of component ‘d’: (a) F555W image; (b)P785LP image. Both images are 5 arcsec2 and have been smoothedwith a s\0.1 arcsec Gaussian.

stars with a surrounding lower surface-brightness disc orhalo, in which there is a significant quantity of gas. Thespectrum of the diffuse emission shows that the SED of theextended continuum is consistent with that of a late-typespiral, with rest frame UÐB2Ð0.2 (Coleman, Wu & Wed-man 1980). The most straightforward interpretation of thisis therefore that we are seeing the interaction of the radiojet with a face-on spiral galaxy in the same cluster, with thebulge component providing the central high-surface-bright-ness knot of red light (with little gas to provide line emissionfrom this region) and the gas and stars in the surroundingdisc providing the diffuse halo of emission. The [O II] emis-sion, however, appears to trace the bow-shock, presumablydue to compression behind the shock increasing the densityand lowering the ionization parameter. The size of the disc(44 kpc in our assumed cosmology) is typical of the size ofspiral discs seen in H I (e.g. Broeils & van Woerden 1994).The problem of survival of the gas halo or disc in the clusterenvironment that may surround 3C 441 suggests that thegalaxy may only recently have fallen into the cluster.

4.2 A simple model for the interaction

We consider a simple model in which the radio jet interactswith a two-phase interstellar medium of the spiral galaxy.Similar models have been developed to investigate theinteractions of supernova shocks with the interstellarmedium, and the case of the interaction of radio jets with atwo-phase medium has been discussed by Rees (1989).These models show that the shock wave propagates thoughthe hot phase, initially by-passing the cold clouds (in whichthe shock speed is much lower). The velocity given to thecold clouds can be estimated by considering the effect of thepressure differential as the shock passes. For a cloud ofradius r and a jump in pressure DP across a shock of speedvsh, the velocity given to the cloud is

vcl1r 2DP 2 r

vsh3 1

rcl r31

DP

rcl Mch

,

where rcl is the density of the cold clouds, M the Machnumber of the shock (in the pre-shock medium), and ch thespeed of sound in the hot-phase gas – i.e., independent ofthe size of the cloud. From the limiting forms of theRankine–Hugoniot relations in the case of a strong shock,

DP1P1M 2,

where P1 is the pre-shock gas pressure. Hence

vcl1P1M

rclch

1c 2

cl

c 2h

vsh

1100 Tcl /104 K

Th/106 K(vsh/104 km sÐ1) km sÐ1,

where Tcl and Th are the temperatures of the cold and hotphases respectively. If we take vsh\vrs10.03c, typical for aradio source advance speed (Alexander & Leahy 1987), thisprocess can easily account for the modest velocity shiftsseen in the gas to the south of the knot relative to theunshocked gas in the north.

974 M. Lacy et al.

© 1998 RAS, MNRAS 298, 966–976

The ionization parameter (here defined as the dimen-sionless ratio of the number density of photoionizingphotons to the total number density of hydrogen atoms,U\nph/nH) will change across the shock as nH is increased(nph is likely to be similar either side of the knot, whether theionization is due to shocks or to the nucleus). Typically[O III] 500.7/[O II] 372.7 is expected to be a linear function ofU over the range we are considering (e.g. Penston et al.1990). To obtain the observed change in the [O III] 500.7/[O II] 372.7 ratio therefore requires an enhancement of nH

in the post-shock region by a factor of 130. This demon-strates the importance of the two-phase assumption: if onlyone phase were present, the Rankine–Hugoniot relationspredict a density enhancement of at most a factor 24. Thecloud gas should also be heated by the compression follow-ing the shock, but deeper spectroscopy is required tomeasure the [O III] 436.3 temperature diagnostic line eitherside of the shock.

Numerical simulations (e.g. Stone & Norman 1992, whoexamine the physically similar situation of supernovashocks) show that, in fact, the hot phase, which is moving ata high speed with respect to the cold clouds, interpenetratesthe cold medium and eventually (within a few sound-cross-ing times of the cloud) causes turbulent mixing of the hotand cold phases. This is probably required to produce thehigh-velocity tail of blueshifted emission-line gas. Gas withhigh velocity shifts and widths (z1000 km sÐ1) is seen in anumber of high-z radio galaxies (e.g., 3C 368), and probablyarises from this mechanism (Stockton, Ridgway & Kellogg1996; Tadhunter 1997).

4.3 Component ‘d’

The apparent close correlation of the position angle of ‘d’with the direction of the radio structure may, of course, becoincidental, and ‘d’ may well be a companion, foregroundor background galaxy which just happens to align with theradio axis. In Section 3.5 we showed that the colours of ‘d’were at least consistent with it being in the same group orcluster as the radio galaxy and ‘c’, but there is no otherevidence to associate ‘d’ with these two galaxies. Neverthe-less, we can speculate on what ‘d’ might be if its alignment isnot a coincidence.

One possible explanation is that ‘d’ may be a tidal rem-nant of an encounter between the radio galaxy and ‘c’ (seeSection 3.1). If, however, as previously argued, any inter-action of the radio galaxy and ‘c’ is unlikely to have beenvery disruptive, this leaves the possibility that the alignmentof ‘d’ is caused by a process involving the radio source, mostplausibly either starlight produced by jet-induced star for-mation, or optical synchrotron emission.

If ‘d’ is synchrotron emission, it is not emission from thejet centre, which clearly passes to one side. Emission from aboundary layer of the jet cannot be ruled out, however, andindeed an interaction of the boundary layer of the jet with agalaxy or gas clump might be expected to be a region of highmagnetic field and current particle acceleration, in amanner analogous to that seen in supernova remnants (e.g.Anderson et al. 1994). We nevertheless cannot rule out ajet-induced star formation origin for this component, on thelines of that proposed by Best et al. (1997a) for the similar

component in 3C 34. Probably the best argument againstthis (apart from the lack of emission lines) is that the objectis parallel to the current jet direction, not the line joining itto the host galaxy as one might naively expect if the galaxywas formed as a result of an interaction with the head of theexpanding radio source.

4.4 The host galaxy

Though relatively subtle, as Fig. 7 demonstrates, the lightfrom the host of 3C 441 is aligned along the direction of theradio jet in both HST passbands (which straddle the 4000-Åbreak) and in the near-infrared. The fraction of aligned light(defined here as the fractional excess of emission within¹22 °. 5 of the radio axis) is largest in the F555W image, atabout 20 per cent, falling to 10 per cent in the F785LP imageand only 4 per cent in the K band. Thus, although 3C 441 isone of the least ‘active’ (by the definition of Best, Longair &Rottgering 1997b) galaxies in the 3C catalogue, and has nodetectable polarization in the infrared to a limit of 25 percent (Leyshon & Eales 1998), we cannot eliminate thepossibility that the aligned light is scattered. Nebular con-tinuum is unlikely to contribute, however, because theobserved limit on the Ha emission suggests that nebularcontinuum accounts for s1 per cent of the total K-bandflux, and, although in the optical emission lines may contri-bute to the aligned components, there is no strong emissionline in K band.

The aligned structures themselves are not closely associ-ated with visible radio emission, rendering both synchrotronand inverse-Compton mechanisms (Daly 1992) unlikely.Jet-induced star formation, however, cannot be ruled outhere, although the colour difference between the alignedemission north and south of the galaxy argues somewhatagainst this, as in the jet-induced star formation model stel-lar populations at equal distances from the nucleus shouldhave similar colours. However, differential dust extinctioneither side of the nucleus could be responsible for the col-our difference.

The aligned structure seen particularly in the F555Wimage (Fig. 6a) is a particularly intriguing feature of thisobject, reminiscent of the more spectacular features in3C 280 (Ridgway & Stockton 1997). As it seems fairly blue,it may represent an AGN-produced component, or pre-existing material brightened by the interaction with theradio jet, for example, nebular continuum emission fromshocked gas, and it may therefore contribute little to thenear-infrared aligned light.

The presence of a good near-infrared alignment mayimply a link between the position angle of the underlyingold stellar population and the radio jet axis, perhaps pro-duced by a selection effect (Eales 1992). Evidence in favourof this is the surprisingly good correlation often seen in 3Cradio galaxies between the infrared PA and the radio axis.Dunlop & Peacock (1993) even claim that it is better thanthat seen in the optical. Although subsequent HST studieshave shown some of Dunlop & Peacock’s infrared align-ments to be fortuitous (e.g., 3C 175.1; Ridgway & Stockton1997), it is also clear that in some cases (e.g., 3C 280; Ridg-way & Stockton 1997) the infrared emission may be wellaligned even after subtraction of some extended alignedcomponents seen in the optical.

Alignments between cD galaxies and the projected distri-bution of galaxies within clusters have been noted by Rhee,van Haarlem & Katgert (1992). Given the common occur-rence of clusters around radio sources at z10.5, it seems notunlikely that the central galaxy is aligned with the clusterpotential, and the selection effect pointed out by Eales(1992) will then tend to favour aligned radio sources (butsee West 1994) for an alternative explanation of the align-ment of the radio source with the cluster/cD axis).

5 IMPLICATIONS OF R ADIO -SOURCE–GA L AXY INTER ACTIONS

3C 441 is a striking example of the ‘alignment effect’,whereby the radio and optical axes of za0.7 powerful radiogalaxies are frequently found to be co-aligned. In this case,if we are correct, the most spectacular aspect of the align-ment, that of knot ‘c’ with the radio axis, has a relativelysimple explanation – namely the interaction of the radio jetwith another galaxy in the same group or cluster. A jet–galaxy interaction has also been invoked to explain theproperties of the aligned emission in the z\0.3 radio galaxyPKS 2250Ð47 (Clark et al. 1997). How relevant these are tothe more problematic cases seen at higher redshifts (andusually higher radio powers too) is unclear. For example,the host galaxy and knot ‘c’ of 3C 441 are s5 per centpolarized (Tadhunter et al. 1992; Leyshon & Eales 1998),but in many examples of high-z radio galaxies the alignedlight has a polarization of 210 per cent. Also, there are veryfew examples where the aligned component is outside theradio lobes.

Jet-induced star formation, the process whereby radiojets impacting on gas clouds promote star formation,appears not to be happening in the aligned component ‘c’ of3C 441. Although the appropriate conditions appear to bein place (namely a powerful radio jet impacting on denseclouds of gas), there is apparently no significant extra bluecontinuum associated with the interaction region that mightcome from a population of very young stars (the youngeststellar population detectable in ‘c’ is the spiral disc popula-tion, which is equally bright on both sides of the centralknot). However, it may simply be that the interaction hasoccurred too recently for the stars to begin to form.

The effects of the interaction on galaxy ‘c’ may bedramatic. The post-shock temperature of the hot (106 K)component of the ISM of ‘c’ will be raised as it passesthrough the shock by a factor 1M 2/4. Assuming vsh1104 kmsÐ1, this is a factor of 1103. Mixing with the cold mediumand radiative losses will cool it somewhat, but it is likely thatthe resulting gas will still have thermal velocities exceedingthe galaxy escape velocity (corresponding to T1107 K), andwill ultimately evaporate into the intracluster medium(ICM). Thus radio source interactions may provide amethod of stripping gas from cluster galaxies, and therebyadding metal-enriched gas to the ICM.

The importance of jet–companion galaxy interactions inproducing radio–optical alignments is still to be fully quan-tified. It is likely that interactions of this sort produce largeamounts of line emission and some nebular continuum, and,if they occur close to the nucleus, the continuum emissonwill be further brightened by scattering of the hidden quasarlight. Any old stellar population in the companion galaxy

A radio-jet–galaxy interaction in 3C 441 975

© 1998 RAS, MNRAS 298, 966–976

contributes light in the near-infrared, and, if the selectioneffect pointed out by Eales (1992) is effective, we wouldexpect to see more companion galaxies along the radio axesthan elsewhere. Furthermore, if radio galaxies form by theaccumulation of smaller subunits as in the ‘bottom-up’structure formation hierarchy, close companions will be acommon feature of high-z radio galaxy hosts.

ACKNOWLEDGMENTS

We thank Chris Simpson for assistance with the UKIRTobservations, and we are grateful to the support staff at theWHT and UKIRT for their help with the other observa-tions. We also thank the referee for a number of usefulcomments. This paper was partly based on observationsmade with the NASA/ESA Hubble Space Telescope,obtained from the data archive at the Space TelescopeScience Institute. STScI is operated by the Association ofUniversities for Research in Astronomy, Inc. under theNASA contract NAS 5-26555. The National Radio Astro-nomy Observatory is a facility of the National Science Foun-dation operated under cooperative agreement byAssociated Universities Inc. The WHT is operated on theisland of La Palma by the Royal Greenwich Observatory inthe Spanish Observatorio del Roque de los Muchachos ofthe Instituto de Astrofisica de Canarias, and the UKIRT bythe Royal Observatory Edinburgh on behalf of the UKPPARC. We are grateful to Mike Irwin and Richard McMa-hon for making the APM scans of the Palomar Sky Surveygenerally available.

976 M. Lacy et al.

© 1998 RAS, MNRAS 298, 966–976

R EFER ENCES

Alexander P., Leahy J. P., 1987, MNRAS, 225, 27Anderson M. C., Jones T. W., Rudnick L., Tregillis I. L., Kang H.,

1994, ApJ, 421, L31Best P. N., Longair M. S., Rottgering H. J. A., 1996, MNRAS, 280,

L9Best P. N., Longair M. S., Rottgering H. J. A., 1997a, MNRAS, 286,

785Best P. N., Longair M. S., Rottgering H. J. A., 1997b, MNRAS,

292, 785Best P. N., Longair M. S., Rottgering H. J. A., 1998, MNRAS, 295,

549

Bremer M. N., Fabian A. C., Crawford C. S., 1997, MNRAS, 284,213

Broeils A. H., van Woerden H., 1994, A&AS, 107, 129Bruzual A. G., Charlot S., 1993, ApJ, 405, 538Caldwell J. A., Rose N., 1997, AJ, 113, 492Canalizo G., Stockton A. N., 1997, ApJ, 480, L5Clark N. E., Tadhunter C. N., Morganti R., Killeen N. E. B.,

Fosbury R. A. E., Hook R. N., Siebert J., Shaw M. A., 1997,MNRAS, 286, 558

Coleman G. D., Wu C.-C., Weedman D. W., 1980, ApJS, 43,393

Daly R. A., 1992, ApJ, 386, L9Dressler A., Gunn J. E., 1990, in Kron R. G., ed., ASP Conf. Ser.

Vol. 10, Evolution of the Universe of Galaxies. Astron. Soc.Pac., San Francisco, p. 200

Dunlop J. S., Peacock J. A., 1993, MNRAS, 263, 936Eales S. A., 1992, ApJ, 397, 49Eisenhardt P., Chokski A., 1990, ApJ, 351, L9Higgins S., O’Brien T., Dunlop J., 1996, in Ekers R., Fanti C.,

Padrielli L., eds, Proc. IAU Symp. 175, Extragalactic RadioSources. Kluwer, Dordrecht, p. 467

Leyshon G., Eales S., 1998, MNRAS, 295, 10Liu R., Pooley G., 1991, MNRAS, 253, 669McCarthy P. J., van Breugel W., Kapahi V. K., 1991, ApJ, 372,

478McCarthy P. J., Spinrad H., van Breugel W., 1995, ApJS, 99, 27McCarthy P. J., Baum S. A., Spinrad H., 1996, ApJS, 106, 281Owen F. N., Hardee P. E., Cornwell T. J., 1989, ApJ, 340, 698Penston M. V. et al., 1990, A&A, 236, 53Rawlings S., Lacy M., Sivia D. S., Eales S., 1995, MNRAS, 274,

976Rees M. J., 1989, MNRAS, 239, 1P

Rhee G., van Haarlem M., Katgert P., 1992, AJ, 103, 1721Ridgway S. E., Stockton A. N., 1997, AJ, 114, 511Stockton A. N., Ridgway S. E., Kellogg M., 1996, AJ, 112, 902Stone J. M., Norman M. L., 1992, ApJ, 390, L17Tadhunter C. N., 1997, in Clark N., ed., Jet–cloud Interactions in

Active Galaxies. http://www.shef.ac.uk/1phys/research/astro/conf/tadhunter/tadhunter.html

Tadhunter C. N., Scarrott S. M., Draper P., Rolph C., 1992,MNRAS, 256, 53

Tadhunter C. N., Dickson R. C., Shaw M. A., 1996, MNRAS, 256,53

West M. J., 1994, MNRAS, 268, 79Zabludoff A. I., Zaritsky C., Lin H., Rucker D., Hashimoto Y.,

Shectman S. A., Oemler A., Kirshner R. P., 1996, ApJ, 466,104