Mon.Not.R.Astron.Soc.000,1–30(0000) Printed1February2008 (MNLaTEXstylefilev1.4) The X-ray Evolution of Merging Galaxies A. M. Read1,2 & T. J. Ponman2 1 Max-Planck-Institut fu¨r extraterrestrische Physik, Postfach 1603, D-85740, Garching, Germany 2 School of Physics and Astronomy, Universityof Birmingham, Edgbaston, BIRMINGHAM, B15 2TT, U.K. 1February2008 8 9 ABSTRACT 9 1 We present here the first study of the X-ray properties of an evolutionary sample n of merging galaxies.Both ROSAT PSPC and HRI data are presented for a sample of a eightinteractinggalaxysystems,eachbelievedtoinvolvea similarencounterbetween J two spiral discs of approximately equal size. The mergers span a large range in age, 0 from completely detached to fully merged systems. 3 AgreatdealofinterestingX-raystructureisseen,andtheX-raypropertiesofeach individual system are discussed in detail. Along the merging sequence, several trends 1 are evident: in the case of several of the infrared bright systems, the diffuse emission v is very extended, and appears to arise from material ejected from the galaxies. The 7 onset of this process seems to occur very soon after the galaxies first encounter one 1 3 another,andtheseejections soonevolveintodistortedflows.Moremassiveextensions 1 (perhaps involving up to 1010M⊙ of hot gas) are seen at the ‘ultraluminous’ peak of 0 the interaction, as the galactic nuclei coalesce. 8 TheamplitudeoftheevolutionoftheX-rayemissionthroughamergerismarkedly 9 different from that of the infrared and radio emission however. Although the X-ray h/ luminosityrisesandfallsalongthesequence,thefactorbywhichtheX-rayluminosity p increases,relative to the optical, appears to be only about a tenth of that seen in the - far-infrared. This we believe, may well be linked with the large extensions of hot gas o observed. r t The late, relaxedremnants,appear relatively devoidof gas,and possess an X-ray s halo very different from that of typical ellipticals, a problem for the ‘merger hypothe- a : sis’, whereby the merger of two disc galaxies results in an elliptical galaxy. However, v thesesystemsarestillrelativelyyoungintermsoftotalmergerlifetime,andtheymay i X still have a few Gyr of evolution to go through, before they resemble typical elliptical galaxies. r a Keywords: galaxies:individual:(Arp270,Arp242,NGC4038/9,NGC520,Arp220, NGC 2623, NGC 7252, AM 1146-270) - galaxies: interactions - galaxies: evolution - galaxies: ISM - galaxies: peculiar - X-rays: galaxies 1 INTRODUCTION fact depend mainly on the number and severity of merger events in its past history. Pure disc systems, formed from Galaxieswereoncethoughtofas‘islanduniverses’,evolving relatively isolated protogalactic gas clouds, appear at one slowly in complete isolation. This is now known not to be end of the Hubble sequence, the giant ellipticals, possibly the case. Galaxies interact in a variety of ways with their produced through mergers of similar spirals, appear at the environment, both with satellite and neighbouring galax- other,andinbetween,mergersbetweengalaxiesofdiffering ies, and, in the case of galaxies within groups and clusters, mass produce galaxies with a wide range of bulge to disc with large masses of tenuous hot gas. Collisions and merg- ratios. ers of galaxies are now thought to be one of themost dom- This merger hypothesis discussed above, the idea that inant evolutionary mechanisms (Schweizer 1989). Indeed, if elliptical galaxies might be formed from the merger of two eventsduringgalaxyformationarecounted,thereareprob- disc galaxies, was first suggested by Toomre (1977). The ably veryfew galaxies thatwerenot shaped byinteractions conversion of orbital to internal energy during a close tidal or even outright mergers (Toomre 1977). The position of a encountercausesthetwoprogenitorsystemstosinktogether galaxy in Hubble’s (1936) morphological sequence may in andcoalesceviolentlyintoacentrallycondensedsystem,dis- c 0000RAS (cid:13) 2 A. M. Read & T. J. Ponman rupting any pre-existing discs, and largely randomizing the lowasubstantialimprovementinourknowledgeofthecom- stellar motions, as shown in many N-bodysimulations (e.g. plex X-ray properties of interacting galaxies over what was Barnes1988).Theproblemthatellipticalshavecentralmass possible with Einstein (see Fabbiano 1989 and references densities some 102 103 times higher than normal spirals therein).ThevariationinX-rayemission withmergerstage − hasbeenovercomebythediscoverythatyoungmergerrem- and position can be compared with corresponding multi- nants contain huge central concentrations of molecular gas wavelength results to reveal new clues as to how the tidal (Sanders et al. 1986; Sanders et al. 1988a; Scoville & Soifer forces trigger star formation, and to constrain the nature 1991). Also, elliptical-like luminosity profileshavebeen dis- and duration of theactivity which ensues. covered in the infra-red underlying the disturbed optical Studies of the nearest, and therefore best resolved, in- morphology ofmerging galaxies (Wrightetal. 1990). There teractingsystemsareespeciallyimportant,andmanyofthe have been many attempts to disprove the merger hypoth- most attractive targets havebeen observed by ROSAT. We esis. Counter-evidence cited includes the fractional abun- have undertaken a programme of ROSAT PSPC and HRI dances of ellipticals in clusters, the colour-luminosity and observations of a carefully chosen chronological sequence of metallicity-luminosity relationships, radial gradients within merging galaxies in ordertostudytheevolution of theirX- ellipticals andwiththeincidenceofglobularclusterswithin ray properties through the merging process. We have com- ellipticals(seeHernquist1993).Thesecriticismscanmostly bined our own ROSAT data with other archival ROSAT be overcome by appealing to a variety of ideas, such as the data and with individual studies published recently. Where growth of substructure in hierarchical cosmogonies, the in- possible, we have attempted to analyse these systems in a complete nature of violent relaxation, and merger induced uniform manner. In particular we have carefully separated star-formation. point sources from what appears to be diffuse X-ray emis- Studies of the evolution of merging galaxies have been sion, and havedetermined thespectral properties of each. performedinmanyregionsoftheelectromagnetic spectrum In the present paper we describe the analysis we have (Joseph&Wright1985;Telesco,Wolstencroft&Done1988; performed,andpresentthebasicresultsforeachofoursam- Casoli etal. 1991; Hibbard& vanGorkom 1996), andthese ple systems in turn. In some cases, we are able to compare arediscussedindetailinbothSection2andSection5.One with previous authors’ results from the same data, but in region of the electromagnetic spectrum that has been ne- many instances our analysis is the first reported ROSAT glected so far though, has been theX-ray. analysis, and in some cases it represents the first picture of The ROSAT X-ray telescope (XRT),with the Position theX-ray emission from an imaging telescope. Sensitive Proportional Counter (PSPC) (Pfeffermann et al. The plan of this paper is as follows. After some com- 1986) at the focal plane, offers three important improve- ments (Section 2) on the selection of the sample, Section 3 mentsoverpreviousX-rayimaginginstruments,suchasthe describestheobservationsandthedatareductionmethods. EinsteinIPC.Firstofall,itsspatialresolutionisverymuch The results for each galaxy are presented in Section 4, to- improved, the 90% enclosed energy radius at 1keV being getherwithnotesontheindividualsystems.InSection5,we 27′′ (Hasingeretal.1992). Secondly,thespectralresolution comment on the range of galaxy properties which emerges, ofthePSPCissignificantlybetter(∆E/E 0.4FWHMat and discuss the X-ray evolution of the sample as a whole. ∼ 1keV) than earlier X-ray imaging instruments and this al- Finally, in Section 6 we present ourconclusions. lowsthederivationofcharacteristicsourceanddiffuseemis- siontemperatures.Finally,thePSPCinternalbackgroundis verylow( 3 10−5cts−1 arcmin−2;Snowdenetal.1994), ∼ × 2 THE MERGING GALAXY SAMPLE thus allowing the mapping of low surface brightness emis- sion.TheHighResolutionImager(HRI)ontheotherhand, Thebasicproblemwithachronologicalstudyistoestablish becauseofitsexcellentspatialresolution( 5′′)andrelative the ordering of the sample. The most famous chronological ≈ insensitivity to diffuse emission, is an ideal instrument for sequenceofmerginggalaxiesisundoubtedlythe‘Toomrese- furtherinvestigation into thepoint source populations. quence’ (Toomre 1977), containing the 11 best examples of Furthermore,asROSAT’senergybandisrelativelysoft ongoing mergers from the New General Catalogue of Neb- compared to that of the Einstein IPC (and to other previ- ulae and Clusters of Stars (NGC). These systems appear, ous instruments), the hot halos now known to exist around whenordered chronologically, torangefrom wellseparated, many nearby normal galaxies (e.g. Bregman & Pildis 1994; strongly interacting systems, apparently destined to even- Wang et al. 1995; Read, Ponman & Strickland 1997), may tually merge, through systems in the throes of merging, to be observable within these more distant, though more ac- well merged systems. This sequence has formed the basis tivesystems.Indeedasshallbediscussedlater,ahothalois for other evolutionary studies made at other wavelengths, known to exist around the merging galaxy pair, the Anten- as discussed below, and indeed, forms the basis for the se- nae (NGC 4038/9) (Read,Ponman & Wolstencroft 1995). lection of thepresent sample. In the ROSAT band, it appears that the X-ray emis- Relevant studies at other wavelengths were, until very sion from normal spiral galaxies is made up of a complex recently,rare.Joseph andWright(1985) studiedhighlydis- mix of sources. Both stellar sources such as low- and high- turbedsystemsinwhichthetwodiscgalaxieshadlosttheir massX-raybinaries,cataclysmicvariablesandnormalmain individualidentitiesandappearasasinglecoalescedobject, sequence stars, and interstellar sources such as supernova the primary morphological indication of which is the pres- remnants(SNRs)andthehotphasesoftheISM,contribute ence of the two tidal tails. Dynamical simulations (Toomre tothetotalX-rayemission.Themergeroftwosimilarspiral and Toomre 1972) suggested that the tails will persist for galaxies may lead to additional sources of X-ray emission. 109 years, and the faintness of these tails, together with ∼ Theadvancesininstrumentationdiscussedaboveshouldal- the degree of coalescence, were used as indicators of rel- c 0000RAS,MNRAS000,1–30 (cid:13) The X-ray Evolution of Merging Galaxies 3 ative merger age. Comparing the infrared luminosities of Toomre & Toomre (1972) have shown that proper tail thesegalaxiestheyfoundthatthetwoyoungestandtwoold- construction requires that the perturbing mass be at least est systems were less luminous than the four ‘middle-aged’ comparabletotheperturbed,andthattherefore,majortidal systems, suggesting that merging galaxies undergo a phase tails should protrude from both of two interacting discs of of starburst activity of characteristic luminosity 1012L⊙. roughly equal dimensions as in NGC 4676 (the Mice) and ∼ Another useful indicator of the progress of a starburst was NGC4038/9(theAntennae).Forsystemswherethecentral found by Telesco, Wolstencroft & Done (1988); for compa- parts of the galaxy have not yet merged, tail length seems rablysizedinteractinggalaxies,thedustcolourtemperature the best indicator of age, increasing with time (Toomre & andtheinteractionstrengthisgreatestforthepairswiththe Toomre 1972), although tail length also increases with the smallestseparation.Furthermore,Casolietal.(1991),while closeness of the encounter. Once the nuclei of the galaxies studyingthe molecular gas content within a series of merg- have merged, the tails begin to diffuse away, and so using ingsystems,discoveredthatthefar-infrared colourtemper- themorphologyofthesystemstoorderthemchronologically ature, the LFIR/LB ratio and the LFIR/M(H2) ratio all comes down to a combination of tail length, tail faintness evolve along the sequence, increasing in the early stages of and degree of coalescence of thenuclei. thecollision, reachinga climax, and then decreasing to val- Listed below, in chronological order, are the eight sys- ues typicalof ellipticals. tems within the sample, selected on the basis of the above Much more recently, Hibbard & van Gorkom (1996) considerations. Optical pictures of the eight systems are havepresentedHi,Hii,andR-bandobservationsofanevo- shown in figure 1. All are taken from the digitised sky sur- lutionarysequenceofmergingsystems,again takenprimar- vey,thesouthernsystemsfromtheU.K.Schmidtplatesand ily from the Toomre sequence. Their five systems, four of thenorthern systems from thePalomar Schmidt plates. which appear in the Toomre sequence, and three of which Arp 270 (NGC 3395/6) consists of two low IR-flux appear in this present paper, span a large range in merger galaxies, obviously involved in some sort of interaction, stage, unlikethework of Joseph & Wright(1985) discussed though the lack of tidal tails is suggestive of them not hav- above, where only middle stage mergers were considered. ingproperlyencounteredoneanotheryet.Thissystemisin Much of Hibbard & van Gorkom’s (1996) work is very rel- fact a pre-Toomre sequence system. evant to the present paper, and many of their results are Arp 242, the Mice (NGC 4676), occurs second in the used in our discussion (Section 5). They find that, moving Toomre sequence and, though it appears rather similar to from early to late-stage mergers, a larger fraction of the Arp 270, the presence of tidal tails and an obvious bridge neutral hydrogen is found outside of the optical confines of between the two galaxies, together with an increase in IR the system, and in the last stage, all of the Hi is found activity, indicates that the galaxies have begun to interact, in the tails, with none in the central regions. They suggest and havepassed each other. that the interaction-induced relaxation causes most of the NGC 4038/9, the Antennae (Arp 244), is a classic atomic gas within the original discs to compress, condense, exampleofaninteractingsystemwithtwoequallylongtails form stars and/or shock-heat to X-ray temperatures, any and distinct central masses. Although it occurs first in the residual atomic gas remaining in the remnant being heated Toomresequence,CO(Stanfordetal.1990),radio(Hummel via theresultant starburst. & van der Hulst 1986) and X-ray observations (Read et al. Inordertomaketheestablishmentofareliablechrono- 1995) indicate that the discs havebegun to merge, and the logical ordering as easy as possible, the systems were care- Antennae therefore should really be placed sometime after fully chosen on the basis of both close attention to the theMice. work of the previous authors’ described above, especially Toomre’s (1977) sequence, and thefollowing important cri- NGC 520(Arp157),seventhintheToomresequence, teria: and classified as an intermediate-stage merger by Hibbard &vanGorkom(1996),isasradioandinfraredbrightasthe Allthesystemscontain,orappeartohaveevolvedfrom, Antennae, and has two smaller tails as well as two nuclei • two spiral galaxies of fairly equal mass. andtwovelocitysystemsinitsspectra,indicativeofayoung A wide range of merger stages, wider in fact than the merger. • Toomresequence,fromcompletelydetachedtofullymerged Arp220, one of the superluminous IRAS galaxies, is a systems, is covered. prototypicalmerger.It isthebrightest ofthesample across Allthesystemsareinfraredandradiobright,indicating the whole of the electromagnetic spectrum (except for the • thepresence of unusualstar-formation activity. optical). Huge Hα filaments exist (Heckman, Armus & Mi- Lowabsorbingcolumns(1.2 5.2 1020cm−2)existin ley1990),suggestiveofvigorousstarburstactivity.Twodis- • − × thedirections ofall thesesystems,maximizing thesensitiv- tinct (though very close) nuclei are visible in the infrared ity to soft X-rayemission. (Majewski et al. 1993). Allthesystemsarelargeenoughforworthwhilespatial • NGC 2623(Arp243),appearingeighthintheToomre resolution to be achieved with the PSPC, and for detailed sequence,is,likeArp220,asuperluminousIRASgalaxy,and mapping of point sources with theHRI. is also very bright in the radio. Very long tails are visible Much additional multiwavelength information is avail- • butthecentralmasseshavebecomeindistinguishable.Only able for all these systems. one truenucleusis thought to exist. Optical morphology (especially theappearance of tidal NGC 7252 (Arp 226), the atoms for peace galaxy, tails, and the proximity of the two nuclei) has been used is the prototypical merger remnant, and is the last sys- as the primary measure of merger stage, but this has been tem within the Toomre sequence. It has two equally long, supplemented with thefollowing additional indicators. very faint tails emanating from a disturbed, double motion c 0000RAS,MNRAS000,1–30 (cid:13) 4 A. M. Read & T. J. Ponman spheroidal body whose mean light distribution closely fol- own pointed observations. The remaining PSPC datasets lows thede Vaucouleursr1/4 law, typical of ellipticals. wereobtainedfromtheUKROSATDataArchiveCentreat AM 1146-270, beyond the end of the Toomre se- the Department of Physics and Astronomy, Leicester Uni- quence, is a spheroidal galaxy with a small tail-like struc- versity U.K. ture, appears to be the site of a great deal of recent star formation, hence the many blue knots surrounding its nu- 3.1 PSPC Observations and Data Reduction cleus. Table 1 lists the sample systems together with the fol- All of the PSPC datasets have been analysed in essentially lowingbasicproperties.Distancesare,wherepossible,taken thesame way using theSTARLINKASTERIXX-rayanal- from Tully (1988). These are based on H0 = 75km s−1 ysissystem.Thismethodisdescribedbelow,andingreater Mpc−1, and assume that the Galaxy is retarded by 300km detail in Read et al. (1997). Departures from this standard s−1 from universalexpansionbythemassoftheVirgoclus- procedure are described in detail in the notes on the indi- ter. For the more distant galaxies, distance values are ei- vidual systems (sections 4.1 4.8). In every case, the target ther taken from Condon et al. (1990), where the same H0 systeminquestion layatthe−centreofthePSPCfield,thus and Virgocentric flow correction is used, or are calculated minimizing the effects of vignetting and blurring. from the group-flow-corrected radial velocities given in the Once the data were ‘cleaned’ of high background pe- Second Reference Catalog of Bright Galaxies (RCBG) (de riods ( 2 3% of the data), they were binned into a Vaucouleurs, de Vaucouleurs & Corwin 1976), using a H0 (0.1 2.≈3keV−) spectral image (or ‘data cube’) of size ap- of75kms−1 Mpc−1.AM1146-270’s radialvelocityistaken prox−imatelytwicethatofeach system.Anannulussituated from Sekiguchi & Wolstencroft (1993). outsidethePSPCcentralsupportring,andwith thebright Optical(B)luminosities forallthesystemsinthesam- sources within it removed, was formed from the data, and ple are calculated as in Tully (1988), i.e. as a re-expression thisbackgroundregion was,togetherwithknowledgeofthe oftheabsolutemagnitude,itselffollowingfromtheblueap- PSPC vignetting function, used to construct a background parent magnitude BT, and thedistance D: model. When this was subtracted from the data cube, the resultantbackground-subtracteddatacubecouldbeusedto logLB(L⊙)=12.192 0.4BT +2logD. − form images in different spectral bands. Blue apparent magnitudes are, again where possible, PSS(Allan,Ponman&Jeffries,inpreparation),apoint taken from Tully (1988). Values for the more distant sys- source search program which uses a likelihood techniqueto temsaretakenfromRCBG(deVaucouleurs,deVaucouleurs search for enhancements above the background, was used & Corwin 1976) and from NGC 2000.0 (Dreyer 1988) (the to search for point-like emission within these images. The value for AM 1146-270 is taken from Smith & Hintzen basic method of PSS involves a comparison of the input 1991)). dataset (in this case, the image) with a model, comprising FIR luminosities are calculated from IRAS 60 and of a background and a scaled PSF. The positions of the 100µm fluxes (taken from the IRAS Point Source Cata- detected point sources, i.e. those with a significance 4σ ≥ logue) using the expression (exceptwherestatedinthetext),werecross-correlatedwith avarietyof stellar and non-stellarcatalogues, includingthe LFIR =3.65×105[2.58S60µm+S100µm]D2L⊙, SIMBADcatalogueandtheEinsteinandIRASpointsource (e.g. Devereux & Eales 1989). Here D is the distance catalogues. inMpcandS60µm andS100µm aretheIRAS60and100µm Asin Read et al. (1997), we attempted toseparate the fluxes(inJanskys).Alsogivenintable1aretheinfraredto diffusefrom thesourceemission byremovingdatafrom the blue luminosity ratio, L /L , and the dust colour tem- background-subtracted data cube at the position of each FIR B perature, S60/S100. These two infrared indicators are the PSS source over a circle enclosing 93% of the energy of a two suggested by previous studies (Joseph & Wright 1985; 0.5keV point source (as all the sample systems subtended Telescoetal.1988) tobeworthwhileindicatorsofstarburst small angles, this extraction radius was never greater than activity,and itcan beseen thattheyriseand fall along the abouthalfanarcminute).Theremainingdatawerethencol- chronological sequence in just the way one might expect if lapsed into a spectrum and corrected for vignetting effects a burst of star formation were triggered during a merger. and exposure time. To account for the diffuse flux lost in Finally, included also in Table 12 are the radio luminosi- thesourceremovalprocess, thediffusespectrum wasrenor- tiesfrom Condonetal.’s(1990) 1.49GHzatlasoftheIRAS malised using a‘patched’image, where the‘holes’ left after brightgalaxysample,supplementedinthecasesofArp242 source removal were filled by bilinear interpolation. and NGC 7252 (upper limit), with the luminosities (con- This method of separating the point source emission verted to H0 = 75km s−1 Mpc−1) from Heckman’s (1983) fromthediffuseemissiondidnot,ingeneral,workaswellas study. inReadetal.(1997).Thisisnotsurprisingasthesystemsin the present study are at much greater distances than those studied in Read et al. (1997), and both limitations on sen- sitivity and on resolution havehad a more serious impact. 3 OBSERVATIONS AND DATA REDUCTION As in Read et al. (1997), source spectra were binned Both PSPC and HRI data have been analysed within this directly from the raw data from circles of the same size as study,andthemethodsofdataextractionandreductionfor those used in the source removal procedure above. To ob- bothtypesofdataaredescribedbelow.AlloftheHRIdata tain true source spectra however, both the local true back- and the PSPC data for three of the systems (NGC 4038/9, ground flux, obtained by scaling the background emission NGC 2623 and AM 1146-270) were obtained through our from source-free regions of the field using knowledge of the c 0000RAS,MNRAS000,1–30 (cid:13) The X-ray Evolution of Merging Galaxies 5 Figure1.Themerginggalaxysample:A)Arp270,B)Arp242,C)NGC4038/9,D)NGC520,E)Arp220,F)NGC2623,G)NGC7252, H)AM1146-270 System Othernames Distance logLB logLFIR LFIR/LB S60/S100 logLrad (Mpc) (ergs−1) (ergs−1) Whz−1 Arp270 NGC3395/6 28 44.08 43.61 0.335 0.486 21.93 Arp242 NGC4676 88 44.06 44.09 1.069 0.534 22.18 NGC4038/9 Arp244 25 44.19 44.20 1.026 0.516 22.64 NGC520 Arp157 28 43.87 44.15 1.900 0.651 22.24 Arp220 UGC9913 76 43.96 45.52 36.04 0.884 23.39 NGC2623 Arp243 78 43.78 44.87 12.28 0.862 22.85 NGC7252 Arp226 63 44.28 44.00 0.526 0.565 <22.78 AM1146-270 25 42.93 42.55 0.416 0.590 - Table 1.Themerginggalaxysample,listedinorderofevolutionaryage.Sourcesforthetabulatedparameters aregiveninthetext. PSPC vignetting function, and the local diffuse flux, cal- An integrated Spectrum was extracted from a cir- • culated by scaling the integrated spectrum of the diffuse cular area of diameter slightly greater than the optical di- emission with an estimate of the fraction of the diffuse flux ameter of each system. Although particle and cosmic X- underlyingeachsource,hadtoberemovedfromeachsource ray backgrounds were removed, no sources were excluded. spectrum. Standard spectral models (power law, bremsstrahlung and So, in summary, three types of spectrum were formed Raymond and Smith (1977) hot plasma) were used in the from each dataset: fitting. Even though these spectra contain emission from c 0000RAS,MNRAS000,1–30 (cid:13) 6 A. M. Read & T. J. Ponman everything around each system’s position, including galac- tic sources, galactic diffuse emission, and foreground and System BR BR backgroundsources(bothGalacticandextragalactic),some 0.3keV 3.0keV of the resultant spectral fits are very acceptable. The net NGC520 6.47 10−4 3.50 10−4 number of counts, the χ2 values, number of degrees of NGC2623 4.54×10−4 2.31×10−4 freedom (n.d.o.f.), and the integrated X-ray luminosities NGC7252 7.11×10−4 4.99×10−4 (0.1 2.0keV) inferred from the spectral fitting, are shown × × − in each of Tables3 to10. Table 2.Counts-to-fluxConversionfactorsfortheROSATHRI ADiffuse spectrumwasextracted,usingthemethod detector for the three individual HRI datasets and for two sep- • described above, from an area the same size as that arate thermal bremsstrahlung models. The values are corrected used for the integrated spectrum. The correctly normal- for Galactic absorption and are in units of 10−11erg cm−2 s−1 ized (0.1 2.0keV) X-ray luminosities of each system’s dif- cts−1. − fuseemissionaretabulatedinTables3to10,alongwiththe net diffuse counts and the results of the spectral fits (us- ingagain thesamemodelsasthoseusedabove).Becauseof rate( 1.1 10−6cts−1arcsec−2;Davidetal.1993).Firstly, low count rates, χ2 fittingcould only beused in thecase of a pol≈ar pr×ofile of the data was formed about the centre theAntennae. Inall othercases, amaximumlikelihood cri- of the system, with radial bins of width 0.001◦. This was terion, which allows for the Poissonian nature of the data, seen to level out at large radius to a constant background had to be used. Unfortunately, when using likelihood fit- level.Thesecondmethodinvolvedanevaluationofthemean ting, the absolute value of the statistic is not informative value,onceallthebrightfeatureshadbeenremovedfromthe (and hence we havenot given it). The Cash statistic (Cash data.Thebackgroundvaluesobtainedfromthetwomethods 1979) can be used however, to compare the relative quality agreed well in all threecases. These values were subtracted ofdifferentfits,andtoderiveconfidenceintervalsformodel from therawdata(oversuchasmall area, thevignettingis parameters,sincedifferencesinthevalueoftheCashstatis- essentially flat,and can beignored). tic from one model to another, are distributed in the same The resultant image was then exposure corrected, and wayasχ2.Furthermore,onlythefreezingofcertainparam- searched for sources, again using PSS. Again, a detection eters at sensible values led to acceptable fits in a few cases threshold of 4σ was assumed. Images, smoothed on many (these are indicated by a bracketted ‘F’). All these results different spatial scales (from 3′′to 18′′), were formed to aredescribedindetailinSections4.1to4.8. Becauseofthe search for extended low-surface brightness features. largedistancesinvolved,manyofthesesystems’diffuseemis- Count rates of the detected point sources within the sion spectra will be complicated, involving some genuinely three HRI datasets were converted into fluxes (and then diffuse gaseous emission, and some contribution from both into luminosities), assuming 0.3keV and 3.0keV thermal unresolved point sources and from ‘stumps’ left in the sub- bremsstrahlung models. HRI conversion factors for both traction of the bright sources. Hence, we attempted, where models from number of counts into fluxes (taking into ac- possible,tofita2-componentmodeltothediffusespectrum, count the different exposure times and Galactic foreground comprisingofaRaymondandSmithhotplasma(represent- N values) are given in Table 2. H ingthetrulydiffusegas)andahighlyabsorbedhot(10keV) Although the spectral response of the HRI is rather component (representingunresolved sources). These results poor, it is thought (Fraser 1992) that the ratio of thenum- are also described in more detail later in the notes on the ber of counts in channels 1 5 to the number of counts in − individualsystems. channels6 11, used by Wilson et al. (1992), does seem the − Individual source spectra were extracted as de- mostsensitiveenergyindicatorthatcanbeconstructed.The • scribed above. Sourcesknown notto beconnectedwith the above analysis was also repeated for data within these two galaxywereexcludedfromtheanalysis,asweresourcesthat energy bands. lay outside both the optical emission and any diffuse X- ray emission. The results of the spectral analysis of all the sources within each system are tabulated below in each of 4 RESULTS theindividualgalaxysections.Thepointsourcespectragen- erallycontainedlownumbersofcounts,andGaussianstatis- The following eight subsections describe the results of the ticscouldnotbeassumed.Amaximumlikelihoodcriterion, analyses ofROSAT PSPC and,whereapplicable, HRIdata as described above, was used. ofthesample systems.Wheretheanalysis techniquediffers from that described in Section 3 this is noted in thetext. WithineachofSections4.1to4.8,atableisgiven,show- 3.2 HRI Observations and Data Reduction ing the main results of the analysis of each set of ROSAT The three HRI datasets (NGC 520, NGC 2623 and PSPC data. The results of the spectral analyses of each of NGC 7252) were analysed in exactly the same way, again the three types of spectra formed as described in Section 3 using the STARLINKASTERIX X-ray analysis system. In (an integrated spectrum, a diffuse spectrum and individual each case the data were first binned into a 0.4◦ 0.4◦ im- source spectra), are given. The number of counts, plus the age with a3′′ resolution, thusexploitingtheHRI×’ssuperior statistical uncertainty, contained within each spectrum are spatial resolution. Background subtraction of HRI data, al- given, together with the results of the best spectral fit to though significantly easier than of PSPC data, is still non- each spectrum. A quoted photon index indicates that the trivial,andtwomethodswereutilised,whichwerecompared best fit is a power law model, whereas if instead, a temper- both with each other and with the mean HRI background ature is given, the best fit is either a Raymond & Smith c 0000RAS,MNRAS000,1–30 (cid:13) The X-ray Evolution of Merging Galaxies 7 hot plasma model (if the metallicity is quoted), or it is a bremsstrahlung model (if no metallicity is quoted). Where χ2-fitting has been used (in fitting most of the integrated spectraplusasmallnumberofotherhigh-countratecases), the values of the χ2 statistic are given together with the number of degrees of freedom. The (0.1 2.0keV) X-ray lu- − minosities tabulated are those escaping from the system in question,i.e.beforeabsorptioninourowngalaxy.Alsogiven in each case is the value of L (diffuse+sources), the sum X of the diffuse X-ray luminosity and each of the individual source luminosities. Unlessotherwisestated,wealsopresentforeachsource a PSPC image showing contoursof background-subtracted, exposure-correctedX-rayemissioninthe0.1 2.3keVband, − superimposed on an optical image of the system. These X- ray images have a resolution of 5′′ and have been lightly smoothedwithaGaussianofstandarddeviation10′′tosup- press noise. The contour levels increase by factors of two from 7.2 10−4ct s−1 arcmin−2. × 4.1 Arp 270 Figure 2. Contours of (0.1 2.3keV) X-ray emission shown su- − Arp270(alsoVV246;Vorontsov-Velyaminov,1959)iscom- perimposed on an optical image of Arp 270. Contour levels in- creasebyfactorsoftwofrom7.2 10−4cts−1 arcmin−2.) posed of two galaxies, thought to be of comparable masses × (Davis & Seaquist 1983), NGC 3395, to the west, and the somewhat smaller,moreirregulargalaxy,NGC3396, tothe east. The angular separation of their nuclei ( 90′′) corre- each galaxy, and be compressed into star-formation. This ≈ sponds to a separation of about 12kpc at the distance of process however, as Huang et al. note, will take some time, 28Mpcassumed within thispaper.Inthefar-infrared, they andthestar-formationmaywellbeginbeforethegasreaches appear as a strong, though unresolved source (Soifer et al. the nuclear regions. Non-nuclear bursts of star-formation 1987), and in the UV, they are noted in the KISO cata- canoccurin thecollision betweentwogas-rich spiral galax- logue(Takase&Miyauchi-Isobe1987)asaninterestingpair ies,whenaburstofmassivestar-formation isfirsttriggered of UV-excess galaxies. Many radio continuum studies have in the overlapping/interpenetrating regions of the galaxies been made, the most noteworthy being that of Huang et (Jog & Solomon 1992). This overlapping region will at first al. (1994), who find that the radio emission in each galaxy be the very outer discs of the two galaxies. In the case of comespredominantlyfrom thatpartofeachgalaxynearits Arp270 though,it appearsthat thisveryinitial phasemay companiongalaxy.WhereasNGC3396showstwostrongra- be nearly over, as gas is now seen close to the nuclear re- dio peaks and one weak one, its partner has a much more gions,i.e.theforcingofgasthroughtheinteractiontowards complex structure, with six or more peaks in the north- the galactic nuclei has begun. The fact that, in the case of eastern part of the galaxy alone. A clear bridge between NGC 3396, this hot gas is seen only close to the nucleus, thetwogalaxiesisalsoseenintheradio,roughlycoincident not yet within it, may indicate that this second phase, the with theoptical. tidal forcing of gas towards the two nuclei, may have only Arp 270 has been observed in X-rays previously to just begun. Furthermore, there is quite a good correlation ROSAT, with the Einstein IPC. The Einstein image (Fab- between the X-ray contours in Fig. 2 and the radio data of biano et al. 1992) shows very little structure, what there is Huangetal.(1994)–boththetentativeX-ray‘bridge’,seen appearing to be centred on NGC 3395. Suggestions of ex- apparentlyconnectingthetwogalaxies,andtheradiobridge tensionsarealsovisibletothenorth-westandsouth-westof appear toconnect thetwosystems along thesouthern edge thesystem. of theoptical bridge. The ROSAT image (Fig. 2) shows far more structure Both source spectra are best fit with low-temperature, thanseenwithEinstein.Emissionisseentocomefromboth absorbed plasma models, suggestive of the emission from galaxies,andthisemissionappearstobeenshroudedinmore both regions being due to hot gas, rather than to evolved diffuseemission.TheX-rayfeatureassociatedwiththewest- stellar components. The fact that no evidence was found ern galaxy (source A; see Table 3) appears to be centrally in the radio for an active galactic nucleus in either galaxy, positioned and lies less than 2′′ from the position given in and that theradio spectra of thegalaxies are relatively flat the Third Reference Catalog of Bright Galaxies (RC3) (de (atlowfrequencies)(Huangetal.1994),indicativeofagood Vaucouleurs et al. 1991). Source B however, does appear, dealofsupernovaactivity,addscredencetothehotgasidea. like the radio emission discussed above, to emanate from a The two other detected sources in the field, the strong positioninNGC3395nearertoitscompaniongalaxy(X-ray point-likesourcetothenorthofNGC3396,andtheweaker sourceBliessome27′′ fromtheRC3positionofNGC3396. sourcewithintheX-ray‘tail’tothesouth-westofNGC3396, AsHuang et al. (1994) point out, this is not expected from appear to explain the suggested extensions tentatively seen dynamical simulations. Instead, the gas should be forced in the Einstein image. The northern source is unlikely to through thetidalinteraction towards thenuclearregions of haveanythingto dowith the system, and is most probably c 0000RAS,MNRAS000,1–30 (cid:13) 8 A. M. Read & T. J. Ponman abackgroundquasar,itsbestspectralfitbeingapower-law talR-bandluminosityofthesystem,andhaveahighatomic fit with a photon index, α = 2.3 0.5, consistent with the gas content (Hibbard & van Gorkom 1996). mean spectrum of quasars in the±ROSAT band (2.2 0.2; The northern galaxy appears to exhibit a 6.6h−1kpc ± Branduardi-Raymont et al. 1994; Roche et al. 1995). The plumeof Hα along its minor axis, and the southern galaxy X-rayfeaturetothesouth-west ofNGC3395 however,may possessesanionizedgasbar,asproducedinBarnes&Hern- wellbeassociatedwiththesystem.Theradioobservationsof quist’s (1991, 1996) merger simulations, offset with respect Huangetal.(1994)doshowthatsomeemissionhasdiffused to the stellar bar (Hibbard & van Gorkom 1996). Angular beyondtheopticaledgeofthegalaxywithinthisregion,per- momentum transfer between the two bars is able to force hapsindicativeofprevioussupernovaoutbursts.Thiswould large amountsof gas towards thegalactic centre. helptoexplaintheexistenceofhotX-rayemittinggasalso. Arp 242 has never been observed in X-rays previously. Thenumberofcountswithinthisregionmakeitdifficultto The X-ray image (Fig 3) shows rather an amorphous X- constrain its spectral properties. ray structure. The one source detected (labelled ‘A’) lies The ‘diffuse’ spectrum of Arp 270 as a whole appears further than the ROSAT rms attitude uncertainty ( 6′′; ≈ severely contaminated, both by unresolved point sources, Hasinger et al. 1992) from the nuclei of either galaxy, lying and by ‘stumps’ of source emission left in the source- approximately 40′′ south of NGC 4676a and 20′′ west of subtraction procedure (see Section 3.1). This is reflected in NGC4676b.Itmaybe,thatthesourceisactuallyassociated theratherwideconfidenceerrorsquotedonthebestfitval- with the contact region between the two galaxies, and is uesin Table 3(remember, nogoodness of fit can bequoted either collisionally heated gas or possibly due to enhanced as a maximum likelihood techniquehad tobe used because starformationtakingplaceinthedensemolecularcloudsin ofthesmall numberofcounts).Asdiscussedin Section3.1, thecontactregion.Unfortunately,duetolackofcountsand we attempted to fit a two-component model to the diffuse limitations on resolution, very little can be said about the spectrum,comprisingaRaymond&Smithhotplasma(rep- spectral properties of thesource, or its trueposition. resenting the hot gas), and a highly absorbed hot (10keV) Leavingasidethissourceforthemoment,theremainder component (representing the source contamination compo- of the emission does seem, in general, to follow the optical nent).Thistwo-componentfitsuggests,asshowninTable3, ‘heads’ of the galaxies, running from the south-east to the that approximately two-thirds of the‘diffuse’ emission may north-west. The little ‘spur’ to the south-west, though in- be truly diffuse gas at a very low (<0.2keV) temperature, teresting,isprobablyspurious,orifnot,unlikelytobeasso- while theremaining third of the emission can be accounted ciated with thesystem. The feature to thenorth-west how- forintermsofhighlyabsorbed,hard(10keV)sources.AnF- ever,maywellbeassociated withtheNGC4676aHαplume testshowsthattheimprovementinfitqualitytothe‘diffuse’ (Hibbard&vanGorkom1996)extendinginthesamedirec- spectrumfromincludingthesecondcomponentissignificant tion. The correlation of Hα and X-ray features extending at over99% confidence. alongtheminoraxisofgalaxies,isoftenagoodindicatorof a starburst-driven galactic wind taking place, both the Hα andtheX-rayemissionbeingduetoclouds(ofdifferentden- sities)beingshockedbythehot,fastwindfromthegalactic 4.2 Arp 242 nucleus(Heckman,Lehnert&Armus1993).Whilethefitted OneoftheoriginalsystemspresentedbyToomre&Toomre temperaturetothediffuseemissionspectrumisinverygood (1972) as a classic example of a pair of galaxies undergo- agreement withfittedtemperaturesofmorenearby,known, ing tidal interaction is Arp 242 (the Mice). Also known as starburstwinds(Heckman1993;Readetal.1997),thesmall NGC 4676, this system lies second in the proposed evolu- numberofcounts,makeitimpossibletobeconfidentofthis tionary sequences of both Toomre (1977) and Hibbard & result. van Gorkom (1996), as well as in the present work. Recent kinematicalwork(e.g.Mihos,Bothun&Richstone1993)has 4.3 NGC 4038/9 strengthened further the tidal interpretation of the Mice’s structure, and it is now generally agreed that both galax- The Antennae, NGC 4038/9 (also Arp 244) is perhaps the ies, inthethroesof aprogradeencounter,havetheirnorth- classic example of a pair of galaxies in gravitational inter- ern edges moving away from us, such that NGC 4676a’s action, with spectacular tails spanning nearly 150kpc. It tail is on the very furthest side, swinging away from us, has been the study of many dynamical models (Toomre & and NGC 4676b is rotating clockwise, its northeastern por- Toomre 1972; Barnes 1988), and the basic validity of these tion, the closest to us (Hibbard & van Gorkom 1996). Hib- models has been confirmed by studies in neutral (van der bard&vanGorkom (1996) findthesekinematicsingeneral Hulst1979) andionised (Amram et al. 1992) hydrogen gas. agreement with their Hi velocity data. What discrepancies Hα emission knots (Rubin,Ford & Dodorico 1971), coinci- they do see allow them to derive a mass ratio of 2:1 for dentwithpowerfulradiofeatures(Hummel&VanderHulst NGC 4676b:NGC 4676a. 1986; Wolstencroft 1988) are found throughout the central Both thenortherngalaxy (NGC4676a) and thesouth- parts of both galaxies, and Van der Hulst (1979) showed erngalaxy(NGC4676b)appeartohaveshapesandcolours that about 70% of the Hi in the system lies in the optical consistent with those of early-type spirals, though the disc tails. Both infrared (Bushhouse & Werner1990; Vigroux et regionsarestronglydistortedorabsent.Bothtails,although al. 1996) and CO (Stanford et al. 1990) observations indi- bluer than the galaxies’ central colours, are in agreement cate that both the nuclei, and the contact region between withthecoloursofouter-discregions(Schombert,Wallin & the two galaxies, are likely to be sites of very active star Struck-Marcell1990).Theyaccountfor16%ofthetotalHα formation. emission,arequiteluminous,containingone-thirdoftheto- The results of the Einstein observation of the Anten- c 0000RAS,MNRAS000,1–30 (cid:13) The X-ray Evolution of Merging Galaxies 9 SourceSpectrumAnalysed Net logLX BestSpectral FitResults R.A. Dec Counts ergs−1 column photon temp. Z χ2 (n.d.o.f) (2000.0) (2000.0) (0.1 2.0keV) 1020cm−2 index keV (solar) − Integrated 396 37.4 40.48 7.76+2.94 3.11+0.44 20.16(19) ± −2.03 −0.43 ‘Diffuse’1-component fit 40.04 5.10+5.46 2.56+9.51 0.00+0.38 −3.18 −0.30 −0.00 ‘Diffuse’2-component fit 168 34.2 (RS)40.01 2.01+72.1 0.17+0.11 1.30+5.42 ± (Br)39.75 150+−903.4.48 10−(F0.)11 −0.44 −71.7 SourceA 104949.58 +325851.6 112 13 39.92 6.49+3.02 0.57+0.21 SourceB 104954.53 +325916.9 95.3±12 39.81 4.91−+21..3708 0.63−+00..3103 ± −1.60 −0.17 Best‘Diffuse’+sources 40.49 Table 3. The results of the PSPC spectral analysis of Arp 270. Tabulated luminosities are those escaping from the system. Only the bestmodelfitisshown.Errorsonthespectralfitparametersare1σ foroneinterestingparameter. SourceSpectrumAnalysed Net logLX BestSpectralFitResults R.A. Dec Counts ergs−1 column photon temp. Z χ2 (n.d.o.f) (2000.0) (2000.0) (0.1 2.0keV) 1020cm−2 index keV (solar) − Integrated 53.8 19.0 40.76 4.37+5.53 0.10+0.13 0.05+118 41.78(18) ± −4.37 −0.10 −0.05 Diffuse 31.2 18.4 40.66 0.00+2.63 0.35+0.39 0.04+0.45 ± −0.00 −0.19 −0.04 SourceA 124609.96 +304320.8 17.5 5.5 40.22 66.2+121 0.24+0.58 ± −63.3 −0.15 Diffuse+sources 40.80 Table 4. The results of the PSPC spectral analysis of Arp 242. Tabulated luminosities are those escaping from the system. Only the bestmodelfitisshown.Errorsonthespectralfitparametersare1σ foroneinterestingparameter. nae (Fabbiano & Trinchieri 1983) were rather inconclusive, Nearly half of the Antennae’s X-ray flux appears dif- thoughtheemissionwasseentobeextended,andcontained fuse, and a two-component spectral fit (see Table 5), indi- a soft X-ray contribution. Harder emission and a possible cates that almost all of it is genuine, hot (4 106K), low- × hard point source at thecontact region were also seen. The metallicity gas (the rest being due to unresolved sources), ROSAT PSPC observation sheds far more light on to the with a total mass perhaps exceeding 109M⊙. The bulk of matter, and full details of theobservation, theanalysis and this emission appears to envelope the entire optical sys- the results, are given in Read et al. (1995), though a sum- tem, except for the tidal tails. An enhancement (accom- mary is presented below. panied by a rise in temperature to over 107K) is seen in The ROSAT PSPC data were processed initially in a this diffuse emission at the contact point between the two manner very similar to the method presented here, though galaxies, where it is believed the two discs are colliding. further,moresophisticated techniqueswere usedin thelat- This may be due to gas heated by the collision, or to trig- terstagesoftheanalysis(seeReadetal.1995).Usingamax- gered star-formation in the vicinity. Of great interest are imum entropyreconstruction technique(Gull1989; Skilling thetwoelongatedstructuresextendingtothenorthwestand 1989), a number of discrete components to the X-ray emis- to the southwest, extending from the galactic discs to radii sionwereidentified(A G;seeTable5andFig.4).Manyof of 30kpc, and culminating in symmetrically-positioned, these have counterpart−s at other wavelengths; B and D are app∼arently point-like, sources (P and Q). These streamers associated with the northern (NGC 4038) and the south- most likely indicate the existence of galactic winds, their ern (NGC 4039) nuclei respectively. Their properties, the normalbipolarstructurepossibly havingbeendisturbedby factsthattheyarebestfittedwithlow-temperatureandrel- therapiddynamicalevolutionofthesystem(seeSection5.3 ativelylow-columnplasmamodels,thatthereareoffsetsbe- forfurtherdiscussionofthis).Containing 10%ofthetotal ≈ tween theX-raypositions andtheradio positions, and that hotgaspresent,thesestreamerscouldrepresentasignificant it is known that there is a great deal of supernova activity loss in mass and energy from thesystem. taking place at these sites, suggest that they are bubbles The point sources at the ends of the two arms are a of hot extranuclear gas, rather than the central starbursts realpuzzle.Thesourcescouldberelatedtothearmsifthey themselves.FeaturesA,C,andpossiblyEhavemassiveHα were massive ( 108M⊙) objects, ejected from the galactic ∼ knots and radio features as counterparts, and are likely to nuclei (see Saslaw, Valtonen & Aarseth 1974), the ejection begiant Hiiregions. processgivingrisetothejuxtaposed‘wakes’ofhotgas.How- c 0000RAS,MNRAS000,1–30 (cid:13) 10 A. M. Read & T. J. Ponman Figure 3. Contours of (0.1 2.3keV) X-ray emission shown su- Figure 4. Contours of (0.1 2.3keV) X-ray emission shown su- − − perimposed on an optical image of Arp 242. Contour levels in- perimposed on an optical image of Arp 4038. Contour levels in- creasebyfactorsoftwofrom7.2 10−4cts−1 arcmin−2.) creasebyfactorsoftwofrom7.2 10−4cts−1 arcmin−2.) × × evertherearevariousdifficulties withthisidea(Read etal. containstwointeractingdiscswhichcollided 3 108 years 1995).Alternativelytheycouldbeforegroundorbackground ∼ × agoandUGC957wasonlyinvolvedintheproducingofthe objects, for example quasars or stars, completely unrelated northern half of a tidal tail. tothediffusearms.Thislatterpossibilityissupportedbyre- TovmasyanandSramek(1976)foundthatthecompact centopticalobservationsofsourceP,theopticalfeatureco- radiosourceinNGC520issituatedinthedarklanebetween incidentwiththesourceattheendofthenorthernstreamer, the two visible parts of the system. Condon et al. (1982) which indicate that it is a distant (z = 0.155) galaxy with later resolved this into a 6′′ east-west extension, consistent an activenucleus (L.R.Jones, privatecommunication). withanedge-ondisc,lying3◦ offeast-west.Millimeter-wave TheAntennaehasalsobeenobservedwiththeROSAT interferometer maps of the 2.6µm CO emission (Sanders et HRI (Fabbiano, Schweizer & Mackie 1996), and the results al. 1988a) show a strong peak at the position of this ra- are generally consistent with the PSPC results. A complex dio source; approximately 1.9 109M⊙ of molecular gas is X-raystructureisseeninvolvingfilamentaryregions,closely × concentrated in a region approximately 0.8kpcin size. following the Hα distribution, emission peaks coincident Muchoftheextranuclearregionsofbothgalaxieswithin with Hii regions, super-Eddington sources, and prominent NGC520experiencedaperiodofenhancedstarformation nuclearsources. Thereis also some evidenceof nuclearout- 3 108 yearsago.Themainsequenceremnantsofthisburs∼t flows and superbubbles. × are the A stars whose features are evident in the optical spectra (Stanford 1991). The burst within the less massive (the north-western) nucleus occurred slightly later than in 4.4 NGC 520 the extranuclear regions, but the star formation rate has ThenatureofthepeculiarsystemNGC520wasonceapuz- returned to a nominal level. The more massive, optically zle; is it one disturbed galaxy or two interacting galaxies? hidden nucleus (to the south-east) produces stars at a rate Recent studies have clarified the interacting galaxies inter- of 0.7M⊙yr−1 andisthecurrentdominantsourceofstar ∼ pretation. Stanford and Balcells (1990) have detected two formation in this system. The star formation rate within galacticnuclei,justvisibleintheopticalbutmoreclearlyin this region is 35 times higher than for an isolated disc ∼ their K-band image. The less massive component (by per- galaxy. This region dominates the mid-infrared flux of the hapsmorethanan orderofmagnitude) isthenorthwestern system,andprobablyproducesmostofthefar-infraredflux knot,which is optically brighter than themain component. seen in NGC 520. It is still unclear, though, whether the The main component is optically weak because it is seen massivestarcomponentissufficienttopowertheactivityat edge-on,andthelightfromitscentralregionisabsorbedby theverycentre,eventhoughmulti-wavelengthobservations interstellardustinitsdisc(visibleasadarklaneinFigure1 argue against the presence of a nonstellar compact power ofBernl¨ohr1993).Furthermore,Twohypotheses,eitherthat source within thesouth-eastern nucleus(Stanford 1991). the nearby dwarf galaxy UGC 957 might be primarily re- The NGC 520 PSPC data were processed exactly as sponsible for the disturbed morphology of a single galaxy describedinsection3.Figure5showscontoursof(0.1 2.3) − in NGC 520, or that two interacting disc systems formed X-ray emission, superimposed on an optical image. Table 6 NGC520,weretestedwithnumericalsimulations(Stanford showstheresultsofthespectralfittingtothePSPCdataas and Balcells 1991). The simulations indicate that NGC 520 describedinSection3.Theonesourcedetectedinthevicin- c 0000RAS,MNRAS000,1–30 (cid:13)