Galaxy Evolution in Hickson Compact Groups: The Role of Ram Pressure Stripping and Strangulation
Jesper Rasmussen, Trevor J. Ponman, Lourdes Verdes-Montenegro, Min S. Yun, Sanchayeeta Borthakur
aa r X i v : . [ a s t r o - ph ] M a y Mon. Not. R. Astron. Soc. , 1–21 (2008) Printed 31 May 2018 (MN L A TEX style file v2.2)
Galaxy evolution in Hickson compact groups: The role of rampressure stripping and strangulation
Jesper Rasmussen, ⋆ † Trevor J. Ponman, Lourdes Verdes-Montenegro, Min S. Yun and Sanchayeeta Borthakur Observatories of the Carnegie Institution, 813 Santa Barbara Street, Pasadena, CA 91101, USA School of Physics and Astronomy, University of Birmingham, Edgbaston, Birmingham B15 2TT Instituto de Astrof´ısica de Andaluc´ıa, CSIC, Apdo. Correos 3004, E-18080 Granada, Spain Astronomy Department, University of Massachusetts, Amherst, MA 01003, USA
ABSTRACT
Galaxies in compact groups tend to be deficient in neutral hydrogen compared to isolatedgalaxies of similar optical properties. In order to investigate the role played by a hot intragroupmedium (IGM) for the removal and destruction of H I in these systems, we have performed a Chandra and
XMM-Newton study of eight of the most H I deficient Hickson compact groups.Diffuse X-ray emission associated with an IGM is detected in four of the groups, suggestingthat galaxy–IGM interactions are not the dominant mechanism driving cold gas out of thegroup members. No clear evidence is seen for any of the members being currently stripped ofany hot gas, nor for galaxies to show enhanced nuclear X-ray activity in the X-ray bright ormost H I deficient groups. Combining the inferred IGM distributions with analytical modelsof representative disc galaxies orbiting within each group, we estimate the H I mass loss dueto ram pressure and viscous stripping. While these processes are generally insufficient toexplain observed H I deficiencies, they could still be important for H I removal in the X-raybright groups, potentially removing more than half of the ISM in the X-ray bright HCG 97.Ram pressure may also have facilitated strangulation through the removal of galactic coronalgas. In X-ray undetected groups, tidal interactions could be playing a prominent role, but itremains an open question whether they can fully account for the observed H I deficiencies. Key words: galaxies: evolution — galaxies: interactions — galaxies: ISM — X-rays: galaxies— X-rays: galaxies: clusters
The origin of the galaxy morphology-density relation is still oneof the most important unsolved problems in astrophysics. Not onlyare spiral galaxies less common within dense cluster environments,but those which are present tend to be deficient in H I , and this de-ficiency itself correlates with projected local galaxy density (e.g.,Giovanelli & Haynes 1985). The mechanisms responsible for thechanges in the morphology and gas content of galaxies are unclear,with gas stripping, tidal shocks, and galaxy interactions and merg-ers all contenders.Traditionally, clusters of galaxies have represented the envi-ronment of choice for attempts to unravel the nature of the rele-vant processes. Recent results, however, strongly suggest that theorigin of the environmental modification of galaxies which under-pin the morphology-density relation lies, not in the cores of galaxyclusters, but in smaller groups and cluster outskirts. For exam- ⋆ E-mail: [email protected] † Chandra Fellow ple, spectroscopic studies of the effects of the cluster environmenton galaxies (e.g., Lewis et al. 2002) show that the suppression ofstar formation takes place in cluster outskirts rather than in thecore, and is modulated by local galaxy density. Moreover, X-raybright groups show a morphology-density relation stronger thanthat of clusters (Helsdon & Ponman 2003), a result adding to theaccumulating evidence that cluster galaxies have often been ‘pre-processed’ in groups prior to their assembly into larger systems(see, e.g., Cortese et al. 2006). In addition, it is becoming clear thatprocesses once thought to be exclusive to the cluster environment,such as ram pressure stripping (Rasmussen, Ponman & Mulchaey2006) and strangulation (Kawata & Mulchaey 2008), may play arole also in much smaller systems. In order to elucidate the ori-gin of the morphology-density relation, it is therefore necessary tostudy the processes acting on galaxies within groups.The compact groups in the catalogue of Hickson (1982) of-fer particularly interesting opportunities in this regard. Many ofthese groups are spiral-rich, but their galaxy population is, on av-erage, deficient in H I by a factor of ∼ compared to loose groups(Williams & Rood 1987). Furthermore, some of these groups have c (cid:13) J. Rasmussen et al. exceptionally compact galaxy configurations, and so may repre-sent pre-virialisation systems close to maximum collapse, in whichgalaxies are suffering strong environmental modification but haveyet to be converted into early-types. While recent work has iden-tified tidal interactions and mergers as playing an important rolein the morphological transformation of spirals in some compactgroups (Coziol & Plauchu-Frayn 2007), it is still unclear what iscausing the observed H I deficiencies and to what extent this is re-lated to the processes modifying the stellar component of the groupmembers.Verdes-Montenegro et al. (2001) presented a detailed study ofthe H I content of a sample of 72 Hickson compact groups (HCGs),with integrated H I masses from single dish measurements, and de-tailed Very Large Array (VLA) mapping of a subsample. DefiningH I deficiency ∆ HI as ∆ HI ≡ log M HI , pred − log M HI , obs , (1)where M HI , obs is the observed H I mass of the group galaxies and M HI , pred is that predicted for isolated galaxies of similar morphol-ogy and optical luminosity (Haynes & Giovanelli 1984), this workconfirmed the deficiency in HCGs, and allowed a search for corre-lations between deficiency and other group properties. One of thestrongest relationships found was that with detectable intergalacticX-ray emission; almost half of the groups with significant H I de-ficiency showed diffuse intragroup X-ray emission in the ROSAT survey of Ponman et al. (1996, hereafter P96). More recently, sim-ilar results have been reported also for groups outside Hickson’scatalogue (Sengupta & Balasubramanyam 2006).The increased prevalence of X-ray detected systems amongH I deficient (compact) groups may suggest a picture wherebyH I is stripped from spiral galaxies within virialising groups, andthen destroyed due to heating by a surrounding hot intragroupmedium (IGM). However, for most of the compact groups in theVerdes-Montenegro et al. (2001) sample, ROSAT data were eithernot available or of insufficient quality to permit any detailed studyof the relationship between the hot and cold gas. With only veryshallow
ROSAT
All-Sky Survey data at their disposal for a numberof these systems, Verdes-Montenegro et al. (2001) could not estab-lish the amount and properties of any hot IGM in these groups.In order to explore the processes destroying H I in these sys-tems, we have therefore embarked on a programme to obtain highquality X-ray and radio data for the most H I deficient compactgroups, adding in existing archival X-ray data wherever relevant.Since ram pressure stripping can only operate where a significantIGM is present, while tidal stripping of gas requires only that galax-ies be interacting, a key discriminator for the mechanism of H I re-moval from galaxies is whether or not there is a correlation betweenH I deficiency and the properties of a hot IGM, and in particularwhether a hot IGM is present in the highly H I deficient systems.Clarifying these issues represents the goal of the present study, inan attempt to shed light on the role played by galaxy–IGM inter-actions in destroying H I and establishing the morphology-densityrelation in these compact groups. The focus of this paper thusrests mainly on the X-ray properties of any hot intergalactic gaswithin the groups, while a forthcoming companion paper (Verdes-Montenegro et al., in prep.) will discuss in more detail the H I andradio continuum emission in the groups, the detailed relationshipbetween X-ray and H I morphology, and the multi-wavelength prop-erties of the individual group galaxies.In Sections 2 and 3 we outline the sample selection and X-ray data analysis, respectively. Section 4 presents the results forthe X-ray properties of the hot gas and galaxies within each group, and compares the derived IGM properties to the observed H I defi-ciencies. In Section 5 we construct an analytical model of a repre-sentative late-type galaxy orbiting within the derived gravitationalpotential of each X-ray bright group, allowing us to evaluate theimportance of ram pressure and viscous stripping for H I removal.The results are discussed in Section 6 and summarized along withthe main conclusions in Section 7.A Hubble constant of 73 km s − Mpc − is assumed through-out. Unless otherwise stated, all errors are quoted at the 68 per centconfidence level. Our broad aim is to establish the relationship between the hot andcold gas in H I deficient groups, and explore the processes of gasremoval operating within them. Our sample is therefore based onthat studied by Verdes-Montenegro et al. (2001), from which weselected all HCGs which are highly deficient in H I ( ∆ HI > . )based on Very Large Array (VLA) H I data, contain four or moregroup galaxies, and lie at a distance D <
Mpc. This yielded aninitial sample of 11 groups, which will be discussed in its entiretyin Verdes-Montenegro et al. (in prep.). VLA H I mapping has beencompleted for all 11 systems, along with follow-up Green BankTelescope (GBT) observations for the groups discussed in this pa-per. Compared to the VLA, the single-dish GBT is more sensitiveto extended, smoothly distributed emission that may otherwise befiltered out by the VLA interferometer. Including the GBT data thusprovides a more complete census of the H I content in the groups.For HCG 48, included in the initial sample, the additional gas de-tected by GBT indicates that this group is not H I deficient afterall. Consequently, this group was omitted from the sample for thepurpose of the present study.Of the remaining systems, the eight included in this paper arethose for which Chandra or XMM-Newton
X-ray data are currentlyavailable. Four of these eight groups are exceptionally compact,with major galaxies (as catalogued by Hickson 1982) containedwithin a circle of radius ∼ arcmin, and thus requiring Chan-dra data to resolve their X-ray structure in the crucial region. Wenote that these groups are not generally mature, X-ray bright sys-tems dominated by early-type galaxies. Such groups may containlittle H I , but the H I content of their galaxies as predicted from theHaynes & Giovanelli (1984) results mentioned above is also small(zero for ellipticals and modest for lenticulars). Such groups aretherefore not usually H I deficient according to our definition inequation (1). Our chosen systems have H I content far below thatexpected for their galaxy contents, and are therefore those in whichthe processes which destroy H I should be in active, or very recent,operation.The observation log for the X-ray data considered in this paperis presented in Table 1, detailing the pointing coordinates for eachobservation (for archival observations not necessarily identical tothe optical group centre), group distance D , the observing instru-ment, date, and mode, along with cleaned exposure times t exp foreach camera, and the Galactic absorbing column density N H fromDickey & Lockman (1990) as adopted in the X-ray spectral analy-sis. c (cid:13) , 1–21 am pressure stripping in Hickson groups Table 1.
Log of available X-ray observations. Group luminosity distances D for the adopted value of H were taken fromthe NASA/IPAC Extragalactic Database (NED). Column 7 specifies the frame mode (full frame/extended full frame) andoptical blocking filter for XMM , and ACIS CCD aimpoint and telemetry mode (Faint/Very Faint) for
Chandra .Group RA Dec D Chandra / Obs. date Obs. mode t exp N H (J2000) (J2000) (Mpc) XMM (yyyy-mm-dd) (ks) ( cm − )HCG 7 00 39 13.5 +00
51 49.3 54
XMM pn 2004-12-26 FF Thin 25.9 2.24. . . . . . . . . . . .
XMM m1 . . . FF Thin 35.1 . . .. . . . . . . . . . . .
XMM m2 . . . FF Thin 35.9 . . .HCG 15 02 07 39.0 +02
08 18.0 92
XMM pn 2002-01-10 EFF Thin 23.3 3.20. . . . . . . . . . . .
XMM m1 . . . FF Thin 31.0 . . .. . . . . . . . . . . .
XMM m2 . . . FF Thin 31.1 . . .HCG 30 04 36 28.3 −
50 02.9 63
Chandra +30
01 25.5 97
Chandra −
50 55.9 98
Chandra +21
48 50.7 23
Chandra +21
41 17.0 . . .
XMM pn 2001-05-07 EFF Thick 8.1 . . .. . . . . . . . . . . .
XMM m1 . . . FF Thin 13.4 . . .. . . . . . . . . . . .
XMM m2 . . . FF Thin 13.1 . . .HCG 97 23 47 25.4 −
19 45.5 86
Chandra +13
06 30.4 69
Chandra
XMM-Newton data
The
XMM data were analysed using
XMMSAS v6.5.0, and cali-brated event lists were generated with the ‘emchain’ and ‘epchain’tasks. Event files were filtered using standard quality flags, whileretaining only patterns for pn and for MOS. Screeningfor background flares was first performed in the 10–15 keV bandfor MOS and 12–14 keV for pn. Following an initial removal ofobvious large flares, a 3 σ clipping of the resulting lightcurve wasapplied. Point sources were then identified by combining the re-sults of a sliding-cell search (‘eboxdetect’) and a maximum likeli-hood point spread function fitting (‘emldetect’), both performed infive separate energy bands spanning the total range 0.3–12 keV. Inorder to filter out any remaining soft protons in the data, a secondlightcurve ( σ ) cleaning was then done in the 0.4–10 keV band,within a 9–12 arcmin annulus which excluded the detected pointsources. Closed-filter data from the calibration database and blank-sky background data (Read & Ponman 2003) for the appropriateobserving mode were filtered similarly to source data, and screenedso as to contain only periods with count rates within σ from themean of the source data. All point sources were excised out to atleast 25 arcsec in spectral analysis.To aid the search for diffuse X-ray emission within the groups,smoothed exposure-corrected images were produced, with back-ground maps generated from blank-sky data. We allowed for adiffering contribution from the non-vignetted particle backgroundcomponent in source- and blank-sky data by adopting the followingapproach. First an EPIC mosaic image was smoothed adaptively( σ – σ significance range), and the particle background was sub-tracted. The latter was estimated from closed-filter data which werescaled to match source data count rates in the image energy bandin regions outside the field of view, and then smoothed at the samespatial scales as the source data. The resulting photon image in-cludes the X-ray background at the source position. To remove thiscomponent, a particle-subtracted blank-sky image was produced ina similar way, and scaled to match 0.3–2 keV source count ratesin a point-source–excised 10–12 arcmin annulus assumed to be free of IGM emission (this assumption is clearly justified in allcases, as will be shown). This image was then subtracted from thecorresponding source image, and the result was finally exposure-corrected. Chandra data
For all
Chandra data sets, calibrated event lists were regeneratedusing
CIAO v3.3. For Very Faint mode observations, the standardadditional background screening was carried out. Bad pixels werescreened out using the bad pixel map provided by the pipeline,and remaining events were grade filtered, excluding
ASCA grades1, 5, and 7. Periods of high background were filtered using σ clipping of full–chip lightcurves, binned in time bins of length259.8-s and extracted in off-source regions in the 2.5–7 keV bandfor back-illuminated chips and 0.3–12 keV for front-illuminatedchips. Blank-sky background data from the calibration databasewere screened and filtered as for source data, and reprojected tomatch the aspect solution of the latter. Point source searches werecarried out with the CIAO task ‘wavdetect’ using a range of scalesand detection thresholds, and results were combined. Source ex-tents were quantified using the σ detection ellipses from ‘wavde-tect’, and these regions were masked out in all spectral analysis.In order to produce smoothed images as for the XMM data,background maps were generated using blank-sky data, and scaledto match source count rates for each CCD. This scaling employedeither the full-chip 10–12 keV count rates, with point sources ex-cluded, or, where possible, count rates in the image energy bandwithin source-free regions on the relevant CCD. The backgroundmaps were then smoothed to the same spatial scales as the sourcedata and subtracted from the latter. The resulting images werefinally exposure corrected using similarly smoothed, spectrallyweighted exposure maps, with weights derived from spectral fits tothe integrated diffuse emission (where possible, otherwise the ex-posure maps were weighted by a T = 1 keV, Z = 0 . Z ⊙ thermalplasma model). c (cid:13) , 1–21 J. Rasmussen et al.
While the smoothed X-ray images described above are useful interms of establishing the presence and rough morphology of anyintragroup medium, we emphasize that they were used for illus-trative purposes only and not for quantitative analyses. WhereIGM emission was not immediately obvious from these images,we performed an additional source detection procedure based onVoronoi tessellation and percolation (‘vtpdetect’ in
CIAO ), whichcan be useful for detecting extended, low–surface brightness emis-sion missed by our standard detection algorithms. To reduce thefraction of spurious detections, a minimum of 50 net counts wererequired for a source to be considered real.As a second step in the search for group-scale diffuse emis-sion, we also extracted exposure-corrected 0.3–2 keV surfacebrightness profiles from the unsmoothed data, with all detectedpoint-like and extended galactic sources masked out. The profileswere extracted from the optical group centre defined by the prin-cipal members in the Hickson (1982) catalogue. For
XMM data,we used an EPIC mosaic image for this purpose, with the particlebackground removed using the method described above. The es-timated particle level shows a typical standard error of the meanof ≈ per cent, which should be representative of the uncertaintyassociated with particle subtraction if the ratio of particle eventsinside and outside the field of view is similar to that in the closed-filter data. We used the blank-sky background data to confirm thisassumption (since these have very little source contamination), buthave conservatively added a 10 per cent error in quadrature to our XMM surface brightness errors, to allow for any residual systematicuncertainties associated with the particle subtraction.For the spectral analysis of any extended emission, X-rayspectra were accumulated in energy bins of at least 20 net counts,and fitted in
XSPEC v11.3 assuming an APEC thermal plasmamodel with the solar abundance table of Grevesse & Sauval (1998).
XMM background spectra were extracted by means of the common’double-subtraction’ technique (Arnaud et al. 2002), using blank-sky background data for the on-chip background, and a large-radius(10–12 arcmin) annulus for determining the local soft X-ray back-ground. Owing to the smaller field of view, a similar approach wasnot generally possible or desirable for the
Chandra observationswhere source emission may completely fill the CCD under inves-tigation. The extraction of
Chandra background data products aretherefore described individually for each group in the next Section.Surface brightness profiles of the X-ray detected groups wereextracted from the peak of the diffuse X-ray emission when clearlyidentifiable (in HCG 37 and 97) and from the centroid otherwise(HCG 15 and 40). The profiles were fitted with standard β –modelsfor conversion into IGM density profiles under the assumption ofisothermality. Since X-ray emissivity is very nearly independent oftemperature for a T ∼ . –1 keV plasma of the relevant metal-licities (see Sutherland & Dopita 1993), this approach is entirelyadequate for our purposes, where the uncertainties of our analysiswill ultimately be dominated by those related to the modelling ofthe impact of the hot gas on the galaxies. The density profiles werenormalized using the spectral normalization A from XSPEC , A = 10 − πD (1 + z ) Z V n e n H d V cm − , (2)where D is the assumed group distance, and n e and n H are thenumber densities of electrons and hydrogen atoms, respectively.The integral represents the fitted emission integral I e over the cov-ered volume V , assumed to be spherically symmetric. Total gas masses within the region of interest were derived by simple vol-ume integration of n e ( r ) µ e m p , where m p is the proton mass andwe have assumed n e /n H = 1 . and µ e = 1 . , appropriate fora fully ionized Z = 0 . Z ⊙ plasma at the relevant temperatures(Sutherland & Dopita 1993).Since we have no knowledge of the density distribution ofany hot gas in the X-ray undetected groups, we have only derivedconstraints on their mean electron density h n e i within the regionconsidered, effectively assuming a uniform IGM distribution. Theadvantage of this approach is that it provides very conservative up-per limits to the IGM masses (and mean ram pressures) within therelevant region. The derived limits to h n e i and M IGM for thesegroups were obtained from the IGM count rate limits and thus de-pend on the depth of the X-ray data. These count rate limits wereestablished on the basis of the exposure-corrected (and, in the caseof
XMM , particle-subtracted) images, by comparing the emissionlevel within a region centred on the optical group centre with that ina surrounding annulus. The physical extent of the region of interestwas thus constrained by the need to evaluate the background locallyfrom our data, and varies from 50–150 kpc among the groups, asdetailed in the discussion of individual groups below. The derivedconstraints on IGM count rates were translated into constraints on I e , assuming the cooling function Λ( T, Z ) of Sutherland & Dopita(1993) and an IGM temperature taken from the σ V – T X relation ofOsmond & Ponman (2004),log σ V = (1 . ± . log T X + 2 . ± . , (3)with galaxy velocity dispersions σ V in km s − and T X in keV. Er-rors on T were derived from the dispersion of this relation, with σ V taken from P96 for HCG 7, 30, 37, 44, and 100, from Mahdavi et al.(2005) for HCG 97, and from Osmond & Ponman (2004) for the re-mainder. The resulting temperature range was then used to estimate σ upper limits on h n e i ∼ ( I e /V ) / inside the assumed spheri-cal volume V for any subsolar metallicity Z . The assumption of auniform IGM in these groups implies that we can identify the up-per limit to the central IGM density n with h n e i and constrain theIGM masses by simply multiplying h n e i and V .In order to briefly investigate the level of nuclear X-ray activ-ity among individual group members, 0.3–2 keV count rates of allgalactic central point sources were also extracted, adopting extrac-tion regions of 2 and 15 arcsec radius for Chandra and
XMM data,respectively. In the majority of cases, photon statistics were insuf-ficient to allow robust spectral fitting for individual sources. Forconsistency, all point source count rates were therefore convertedto luminosities assuming a power-law spectrum of photon index
Γ = 1 . , absorbed by the Galactic value of N H . The associateduncertainties were derived from the Poisson errors on the photoncount rates. In this section we discuss the results obtained for the IGM in eachgroup and for the X-ray emission associated with individual groupgalaxies. Figure 1 shows contours of the smoothed, background-subtracted X-ray emission of each group overlayed on DigitizedSky Survey (DSS) images. Diffuse X-ray emission associated withan intragroup medium is detected in four of the eight groups, as de-scribed for each group individually below. For the remaining four,we do not detect any extended group emission, neither inside agiven physical radius from the optical group centre when comparedto the emission level in surrounding regions, nor on the basis of the c (cid:13) , 1–21 am pressure stripping in Hickson groups Figure 1.
Contours of adaptively smoothed 0.3–2 keV emission overlayed on DSS images for all groups, with the principal group members labelled followingthe notation of Hickson (1982). Where relevant, dashed squares outline the coverage of the
Chandra /ACIS CCD’s. For HCG 44, dark (black) contours outlinethe
Chandra emission, and lighter (green) those of an overlapping
XMM pointing.
Voronoi source detection procedure. To corroborate these results,Figures 2 and 3 show the derived surface brightness profiles for thegroups with and without detectable intragroup emission, respec-tively.Table 2 summarizes the observed H I and X-ray properties ofthe groups, along with the adopted velocity dispersions from opticalspectroscopy. H I deficiencies in the Table are from our GBT mea-surements (Borthakur et al., in prep.), except for HCG 44, for whichwe have adopted the older VLA value (Verdes-Montenegro et al.2001) due to the GBT beam size only covering the central regionof this relatively nearby system. The listed H I deficiencies are basedon the integrated H I mass within the circular region covered by theradio data ( r HI in the Table). In many groups, a significant fractionof the detected H I is located outside the optical extent of individ-ual galaxies (i.e. is intergalactic) and cannot be clearly assigned toany individual group member. The derived values of ∆ HI should therefore generally be viewed as an average for the galaxies withinthe GBT or VLA field. As the fractional 1- σ uncertainty on themeasured H I masses from our GBT data is less than 1 per cent forall groups, uncertainties on the listed deficiencies are dominated bythose related to the predicted logarithmic H I mass, which we havetaken to be 0.2, adopting the standard estimate of error provided byHaynes & Giovanelli (1984).For reference, 0.3–2 keV X-ray luminosities are also listed inTable 2, corrected for Galactic absorption, and derived within theregion employed for the spectral analysis unless otherwise speci-fied in the subsection for the relevant group. The listed central hotIGM densities (or upper limits to the mean densities for the X-ray undetected groups) were computed as outlined in Section 3.3.IGM masses in the Table were derived within the same region asthe H I deficiencies (i.e. within r HI given in the Table), to enable adirect comparison between the two. Note, as indicated above, that c (cid:13) , 1–21 J. Rasmussen et al.
Table 2.
Summary of derived group properties. Except where indicated, H I deficiencies ∆ HI are from GBTdata (Borthakur et al., in prep.), obtained inside a radius r HI . Column 7 lists the derived constraints on centralhot IGM density, with upper limits for the X-ray undetected groups given at σ significance, and Column 8the corresponding hot IGM mass within r HI .Group ∆ HI r HI σ V T X L X n M IGM (kpc) (km s − ) (keV) ( erg s − ) ( − cm − ) ( M ⊙ )HCG 7 0.60 68 95 a . ± . ∗ < . < . < . HCG 15 0.46 113 404 b . +0 . − . ± . ± . . ± . HCG 30 1.37 79 72 a . ± . ∗ < . < . < . HCG 37 0.33 119 446 a . +0 . − . ± ± . ± . HCG 40 0.60 121 157 b . +0 . − . . ± . . ± . . ± . HCG 44 0.69 †
101 145 a . ± . ∗ < . < . < . HCG 97 0.35 106 383 c . +0 . − . +20 − ± . ± . HCG 100 0.27 86 100 a . ± . ∗ < . < . < . † From VLA data (Verdes-Montenegro et al. 2001). ∗ Obtained from the assumed σ V – T X relation, equation (3). a Ponman et al. (1996). b Osmond & Ponman (2004). c Mahdavi et al. (2005). -5 c t s / s / a r cs e c HCG 15
10 100r (arcsec)-40-2002040 R e s i d . ( % ) -5 -4 c t s / s / a r cs e c HCG 37
10 100r (arcsec)-60-3003060 R e s i d . ( % ) -5 c t s / s / a r cs e c HCG 40
10 100r (arcsec)-30-1501530 R e s i d . ( % ) -5 -4 c t s / s / a r cs e c HCG 97 R e s i d . ( % ) Figure 2. β –models (solid lines).Horizontal dotted lines mark the estimated background level in each case. For each plot, the bottom panel shows fit residuals relative to the best-fitting model.c (cid:13) , 1–21 am pressure stripping in Hickson groups
10 10010 -7 -6 c t s / s / a r cs e c HCG 710 10010 -5 c t s / s / a r cs e c HCG 3010 10010 -6 -5 c t s / s / a r cs e c HCG 44 (Chandra)10 10010 -6 c t s / s / a r cs e c HCG 44 (XMM)10 100r (arcsec)10 -5 c t s / s / a r cs e c HCG 100
Figure 3.
As Fig. 2, but for the groups without detectable diffuse emission. the upper limits to M IGM for the X-ray undetected groups conser-vatively assume a uniform IGM distribution. If instead assuming astandard β –model for the hot gas in these groups, with central den-sity equal to the inferred mean value h n e i and with typical groupvalues of, e.g., β = 0 . and r c = 20 kpc, the derived IGM masslimits would be reduced by factors of 4–7. This group remained X-ray undetected in shallow
ROSAT
All-SkySurvey (RASS) data. Despite the
XMM data of this target represent-ing the deepest X-ray observation within our sample, no diffuse X-ray emission is detected in the group when comparing the emissionlevel of the exposure-corrected and particle-subtracted 0.3–2 keVmosaic image inside r = 9 arcmin ( r ≈ kpc) with that mea-sured in a surrounding annulus. This is corroborated by the derivedsurface brightness profile shown in Fig. 3. Although this profiledoes seem to hint at a weak signal inside r ≈ arcmin, the com-bined signal inside this region is significant at less than . σ , isnot picked up by ‘vtpdetect’, and is not discernible in the smoothedimage presented in Fig. 1. Thus, we conservatively treat it as a non-detection. In fact, no extended emission unassociated with individualgalaxies is identified by ‘vtpdetect’ within the 9 arcmin radius, withthe exception of the source visible in Fig. 1 roughly ∼ arcminsouth of the optical group centre. There are no optical/infra-redcounterparts to this X-ray source listed in NED within a 2 arcmindiameter, despite the proximity of the group ( D ≈ Mpc). Theemission is detected out to 2.2 arcmin from the X-ray centroid at σ above the local background, confirming that it is clearly ex-tended. A thermal plasma model fit to the spectrum extracted fromthe pn data within r = 1 . arcmin of the centroid provides an ac-ceptable fit, with a reduced χ ν = 0 . for 24 degrees of freedom(d.o.f.). This yields a best-fitting temperature T = 2 . +0 . − . keVand redshift z = 0 . +0 . − . for an assumed abundance of 0.3 so-lar. A simple power-law model with Galactic absorption yields Γ = 1 . ± . for the power-law index, but the fit is not pre-ferred to a thermal model (red. χ ν = 1 . for 25 d.o.f, i.e. achange of ∆ χ = 4 . ). The estimated temperature can be com-pared to that expected for the IGM in HCG 7 on the basis of itsvery low galaxy velocity dispersion, σ V = 95 km s − , for whichthe σ V – T X relation of Osmond & Ponman (2004) would suggestonly T = 0 . ± . keV. Combined with the redshift estimate of z ≈ . , this strongly suggests that this emission is not associ-ated with HCG 7 itself. The X-ray centroid also coincides to within10 arcsec with an NVSS source with a 1.4 GHz flux of 19.4 mJy(corresponding to × W Hz − at z = 0 . ), so the X-rayemission is conceivably associated with a z ≈ . backgroundcluster harbouring a central radio-loud galaxy.For T anywhere in the range 0.2–0.4 keV and assuming anysubsolar metallicity, our failure to detect IGM emission inside r ≈ kpc translates into a σ upper limit to the unabsorbed 0.3–2 keV luminosity inside this region of L X < × erg s − ,with a corresponding limit to the mean gas density of h n e i < × − cm − . We note that HCG 7 is included in the group catalogueof Yang et al. (2007), with a total group mass, as estimated from itsoptical properties, ranging from 2.4–5.0 × M ⊙ depending onthe method assumed. This places HCG 7 at the very low-mass endof the group mass function, with a mass similar to that of the LocalGroup. Thus, it is perhaps not surprising that we fail to detect anyIGM emission in this system. This group has a relatively high velocity dispersion of σ V ≈ km s − , and hot intragroup gas was already detected in pointed ROSAT observations (P96). Significant IGM emission is seen in the
XMM data presented in Fig. 1, revealing a somewhat disturbed X-ray morphology. Despite this irregularity of the emission, an op-tically bright early-type galaxy is present roughly at centre of theX-ray emission, as is typical for fairly undisturbed X-ray brightgroups. The facts that this galaxy is a lenticular rather than an ellip-tical, and that the IGM emission is not (yet) strongly peaked on thisgalaxy, may suggest that the group is in the late stages of dynamicalrelaxation.Emission is detected in the imaging data out to r = 8 . arcmin( r ≈ kpc) at σ above the X-ray background level evaluatedfrom a surrounding annulus. A radial surface brightness profile isshown in Fig. 2, extracted from the centroid of emission in binscontaining a signal-to-noise ratio of S/N > . As expected fromthe irregular X-ray morphology, a standard β –model is not a satis-factory description of this profile, with χ ν of 3.5 for 52 d.o.f. Thedata show significant deviations from the best-fitting model (with β = 0 . ± . and r c = 21 . +5 . − . arcsec) at all radii. However, c (cid:13) , 1–21 J. Rasmussen et al. the fit residuals do not exhibit any systematic radial variation, sug-gesting they are caused by local fluctuations in the IGM distributionrather than large-scale inhomogeneities. Hence, despite the fact thatthe β –model fit is clearly statistically unacceptable, it remains use-ful for our purposes as a means of characterizing the global hot gasdistribution. From inspection of Fig. 2, it is also not clear that a dif-ferent, or more complex, model would be able to provide a betterdescription.The relatively broad point spread function (PSF) of XMM , notaccounted for in the surface brightness fitting, could potentially af-fect the observed profile at small radii, and hence the derived coreradius and central gas density. We do not expect this to be an im-portant effect, however, because even just the innermost radial binin Fig. 2 extends to r = 12 arcsec, roughly twice the full-width athalf maximum of the EPIC PSF. The fact that the best-fitting coreradius is another factor of two larger also suggests that PSF blurringdoes not have a significant impact on the derived results.Using the double-subtraction approach for extracting a back-ground spectrum, a fit to the global 0.3–5 keV spectrum extractedinside r = 6 arcmin ( r = 150 kpc) gives a temperature T =0 . +0 . − . keV and abundance Z = 0 . ± . Z ⊙ , thus confirm-ing the low abundance derived from ROSAT data within the sameregion (Osmond & Ponman 2004). These values are consistent withthose obtained using local background subtraction, but results arebetter constrained due to the superior statistics of the blank-skybackground data. The derived flux and surface brightness profileimply a central hot gas density n = 4 . ± . × − cm − . Despite this group representing the most H I deficient system withinour sample, the Chandra data do not reveal any clear evidencefor diffuse IGM emission. No extended sources outside individualgalaxies are detected by ‘vtpdetect’, thus corroborating the RASS-based result of P96. With a galaxy velocity dispersion of only72 km s − , the Osmond & Ponman (2004) scaling relations wouldsuggest a very low IGM temperature of T = 0 . ± . keV. In orderto test for the presence of any such gas, we generated 0.2–0.4 keVimages of the data on the S2 and S3 CCDs separately. These im-ages were exposure-corrected and smoothed but not background-subtracted, in an attempt to suppress any bias related to ACIS cal-ibration uncertainties at these low energies. The results reveal noclear spatial variations in the diffuse emission on either chip, sug-gesting emission at a level consistent with the local background.As a further test, we searched the unsmoothed data for a radialgradient in the exposure-corrected 0.3–2 keV emission level acrossthe S2 and S3 CCD’s, finding no significant variation with distancefrom the optical group centre (see Fig. 3 for a surface brightnessprofile extracted on the S3 CCD). This implies that any diffuse IGMemission would have to be near-uniformly distributed on scales of ∼ kpc, an unlikely scenario for this low- σ system, in which theangular extent of the region encompassing the four principal groupmembers is only ≈ arcmin ( ∼ kpc).The RASS 0.5–0.9 keV count rate in a 0.5–1 deg annulus cen-tred on the optical group centre is σ above the exposure-weightedmean of the appropriate Chandra blank-sky data, suggesting a con-siderable contribution from either background or (Galactic) fore-ground emission at this position. The high background rate relativeto blank-sky data requires us to evaluate the background from thesource data, thus restricting the source region under investigationon S3 to within r ∼ arcmin ( ∼ kpc) of the optical group cen-tre. Assuming that the background in the data can be safely evalu- ated from source-free regions on S3 outside this central region (anassumption supported by Fig. 3), the density of any hot IGM in thegroup can be constrained. Inside this region, and for T in the range0.1–0.3 keV and any subsolar metallicity, the σ upper limit to themean gas density is h n e i < . × − cm − . The possibilitythat T is very low in this group propagates into a relatively weakconstraint on h n e i . Irregular diffuse X-ray emission was detected in this group with
ROSAT out to r ∼ arcmin (Mulchaey et al. 2003), well beyondthe region covered by a single ACIS chip in our Chandra data.The latter clearly indicate that the group emission is sharply peakedon the early-type galaxy HCG 37a ( = NGC 2783), the nucleus ofwhich is also detected as a point-like source in the data. The as-sociation of the IGM X-ray peak with HCG 37a was not obviousfrom the earlier
ROSAT data, as the much broader
ROSAT
PSF re-quired Mulchaey et al. (2003) to exclude point-like emission outto r = 1 . arcmin from the peak, thus effectively masking outHCG 37a in the data. Despite the overall irregularity of the groupemission, the X-ray centroid (when masking out the HCG 37a nu-cleus) coincides to within 10 arcsec with the optical position ofHCG 37a as listed in NED.Since group emission covers the S3 CCD, we cannot reliablyuse any method relying on source-free regions on S3 to evalu-ate the background in the Chandra data. The situation is aggra-vated by the fact that RASS data indicate a σ soft backgrounddeficit at this position relative to the appropriate blank-sky data,so background subtraction by means of these is not straightfor-ward either. To circumvent these issues, we adopted the methodemployed by Vikhlinin et al. (2005). First, a source minus blank-sky spectrum was extracted on the back-illuminated S1 chip, toquantify the difference in the soft background between source- andblank-sky data. The spectrum was fitted with a T = 0 . keV Z = Z ⊙ mekal plasma in the 0.4–1 keV range, with the nor-malization allowed to be negative. The best-fitting model was thenadded to the model fit of the blank-sky subtracted source emissionon S3 inside r = 3 . arcmin ( r ≈ kpc) after scaling to thesource region area. The resulting background level was also usedfor the surface brightness analysis. We note that, at 90 per centconfidence, the best-fitting T and Z resulting from this approach, T = 0 . +0 . − . keV and Z = 0 . +0 . − . Z ⊙ , are just consistentwith the P96 values inside r = 150 kpc ( T = 0 . ± . keV, Z = 0 . ± . Z ⊙ ), lending some credibility to this approach.The surface brightness profile shown in Fig. 2 confirms thepresence of emission across the full S3 CCD in the Chandra data.The profile was centred on the X-ray peak and extracted in binscontaining at least 30 net counts. Despite the irregularity of theemission on large scales (Fig. 1), a β –model provides a good fitto the profile across the full radial range plotted in Fig. 2, yielding β = 0 . ± . and r c = 3 . +1 . − . arcsec, with χ ν = 0 . for19 d.o.f. The spatial and spectral results imply a gas density in thegroup core of . ± . cm − . The
Chandra observation of this group was split into two sepa-rate pointings, so a merged event file was produced for the imaginganalysis. Although undetected in a 3.6-ks
ROSAT pointing (P96),Fig. 1 suggests the presence of diffuse emission in this group. The c (cid:13) , 1–21 am pressure stripping in Hickson groups centroid of this emission, with the optical extent of the individualgroup members masked out, is located ∼ . arcmin to the NWof HCG 40c and so is not clearly associated with any individualgalaxy. To test that the emission seen in Fig. 1 is truly extended andnot simply due to the smoothing of point sources, a surface bright-ness profile of the unsmoothed, exposure-corrected emission wasextracted from the centroid in bins of at least 30 net counts, withindividual galaxies masked out. The result is shown in Fig. 2, withthe background evaluated from off–source regions on the S3 CCD.Emission is detected above this background out to r = 2 . arcmin( r = 65 kpc), suggesting group-scale extended emission, althoughthe detection is only significant at > σ for the innermost 30 kpc. A β –model provides an acceptable fit to this profile, with χ ν = 1 . for 7 d.o.f., yielding β = 0 . +0 . − . and r c = 46 . +35 . − . arcsec( +17 − kpc), in accordance with expectations for a typical X-raybright group.For the spectral analysis of this emission, spectra and responseproducts were extracted separately for each of the two observations.The spectra were then jointly fitted within the central r = 1 arcmin,within which the signal allows useful constraints to be obtained,using a surrounding . – . arcmin annulus for background esti-mation. With only ∼ net counts, the IGM abundance remainsunconstrained. Fixing Z at 0.3 solar yields T = 0 . +0 . − . , and T remains consistent with this for any subsolar Z . For these param-eters, the observed flux translates into a central electron density of . ± . × − cm − and implies a diffuse 0.3–2 keV luminosityinside r = 2 . arcmin of . ± . × erg s − .The extent of the emission, coupled with the fact that it isnot clearly centred on any group member, suggests that the emis-sion is not due to, for example, hot gas associated with an el-liptical but rather reflects the presence of a hot IGM. This in-terpretation would place HCG 40 among the relatively rare ex-amples of spiral-dominated groups showing intergalactic hot gas;within Hickson’s (1982) catalogue, only HCG 16, 57, and the well-studied HCG 92 (Stephan’s Quintet) share similar features (e.g.,Dos Santos & Mamon 1999; Fukazawa et al. 2002; Trinchieri et al.2003). Based on the B -band luminosities of the group mem-bers, and on the L X – L B relations for ellipticals and normal star-forming spirals from O’Sullivan, Ponman & Collins (2003) andRead & Ponman (2001) respectively, one would expect a totalgalactic diffuse L X ≈ × erg s − in the group, a factorof three larger than that found here for the intragroup emission. Al-though care has been taken in masking out emission from the groupmembers, the low S/N and the compactness of the galaxy config-uration implies that we cannot exclude a residual contribution tothe diffuse emission from individual galaxies. A conservative ap-proach would be to regard the association of the observed diffuseemission with an intragroup medium in HCG 40 as tentative ratherthan conclusive. While Fig. 1 does not indicate the presence of any IGM emissionin this system, this could simply be an artefact of the proximity ofthe group ( D ≈ Mpc) in combination with the limited
Chandra angular coverage, which furthermore renders quantitative analysisof the background level in the data non-trivial. Using local back-ground subtraction could potentially produce unreliable results, asIGM emission might cover the entire ACIS array. The situation isfurther complicated by an enhanced particle level in the cleaneddata compared to the blank-sky files, with the 10–12 keV countrate on the S3 CCD being 35 per cent higher than the correspond- ing blank-sky value. In addition, soft Galactic 0.5–0.9 keV emis-sion at a level of σ above the blank-sky data is also present atthis position, so it is not obvious that blank-sky data would be anappropriate choice for background estimates.Fortunately, the presence of an overlapping XMM pointingallows an independent test for the presence of diffuse emissionin the group. The
Chandra and
XMM surface brightness profilesshown in Fig. 3 suggest no detectable IGM emission close to theoptical group centre. Note that the extraction of these profiles ex-cluded different position angles due to the optical group centre be-ing close to the southern (northern) edge of the S3 (EPIC) CCDs,so the profiles extend in largely opposite directions on the sky. The
XMM profile shows no systematic variation out to r = 15 arcmin( r ≈ kpc), remaining largely consistent with the backgroundlevel evaluated outside this region in the source data. Consequently,the background level for the Chandra profile was estimated fromthe northern corners of the S3 CCD, the result suggesting no ex-cess diffuse emission extending northwards either.Furthermore, no extended sources that can be unambiguouslyassociated with group emission were detected by ‘vtpdetect’. In ad-dition to the two group galaxies HCG 44a and b, a third extendedX-ray source is seen on the S3 chip, clearly visible in Fig. 1 roughlyfour arcmin north of the spiral HCG 44a, and also seen in both the
XMM data and in pointed
ROSAT observations. The X-ray peak ofthis source coincides with a 2MASS source with m K = 14 . ,but there is no optical counterpart or redshift information availablein NED. If this source were at the group distance, the resulting K -band luminosity of . × L ⊙ ,K would place it at the ex-treme faint end of the dwarf galaxy luminosity function, with the ROSAT flux implying a ratio L X /L K ≈ . , two orders of mag-nitude above typical values seen even for dwarf starburst galaxies(Rasmussen, Stevens & Ponman 2004). Spectral fit results providefurther support for the idea that this source is unlikely to be associ-ated with HCG 44. A thermal plasma model fixed at the group red-shift returns an unacceptable fit ( χ ν = 1 . ), whereas a significantfit improvement results when leaving z as a free parameter, yield-ing χ ν = 1 . for 6 d.o.f, with T = 1 . +0 . − . and z = 0 . +0 . − . for an assumed abundance of Z = 0 . Z ⊙ .While the Chandra data are useful in terms of investigatingevidence for hot gas being stripped from individual galaxies in thisvery nearby system, it is not clear that these data enable significantimprovements on the hot IGM constraints over the existing 4.7-ks
ROSAT pointing (with its much larger field of view enabling amore reliable background subtraction in this case). Even the
XMM data can only probe emission from a quarter of the volume inside r = 100 kpc from the optical group centre, due to the latter beingclose to the edge of the XMM field of view. Using the adopted σ – T relation, which would suggest T = 0 . ± . keV, the ROSAT constraint of P96 on the X-ray luminosity inside r = 150 kpc( L X < . × erg s − for our adopted distance) translates into h n e i < . × − cm − for any subsolar metallicity. The cor-responding XMM constraint of h n e i < . × − cm − applieswithin a volume less than 10 per cent of that probed by ROSAT , sowe have adopted the stronger
ROSAT limit in Table 2.
This constitutes the most X-ray luminous system within the sam-ple. A two-dimensional analysis of a 21-ks
XMM observation ofthis group was performed by Mahdavi et al. (2005), along with op-tical spectroscopy identifying 37 members. Their
XMM data showa plume stretching to the southeast, beyond the region covered by c (cid:13) , 1–21 J. Rasmussen et al.
Fig. 1. Mahdavi et al. (2005) speculate that this plume could repre-sent gas stripped from one of the central galaxies, but the
XMM dataalone cannot establish this, and the
Chandra data on the S2 CCDcannot improve on the situation beyond confirming the presenceand overall morphology of this feature. One of the spectroscopi-cally identified member galaxies is located within the plume but isnot itself detected in either the
XMM or Chandra data. On the S3CCD, the emission appears fairly regular in Fig. 1, albeit with thecentral X-ray contours somewhat elongated towards the south-east.If masking out this elongated feature, the centroid of the IGM emis-sion coincides to within 10 arcsec with the position of the opticallybrightest group galaxy, HCG 97a, as listed in NED.As established already by
ROSAT observations (e,g.,Mulchaey et al. 2003), diffuse emission in this group extends wellbeyond the region covered by the S3 CCD, so blank-sky data wereused to evaluate the background for the
Chandra surface bright-ness analysis (the RASS 0.5–0.9 keV background count rate at thisposition is in good agreement with the corresponding exposure-weighted mean value of the blank-sky data). We note, however, thatthe
Chandra observation was somewhat affected by backgroundflares, reducing the useful exposure time from 57.9 to 36.3-ks. Asfor HCG 44, the background remains high after cleaning, with the10–12 keV count rate on S3 again being 35 per cent above theblank-sky value. While bearing this issue in mind, results indicatethat emission is detected above σ significance everywhere on theS3 chip. A fit to the exposure-corrected 0.3–2 keV surface bright-ness profile, extracted from the X-ray peak and shown in Fig. 2,yields β = 0 . ± . and r c = 6 . ± . arcsec, with χ ν = 1 . . However, while Figure 2 and the excellent agreementof these results with the best-fitting parameters of Mulchaey et al.(2003) (who find β = 0 . ± . and r c < . arcmin) sug-gest that our background estimate is not seriously in error, we willnevertheless base our normalization of the density profile on the ROSAT results of Mulchaey et al. (2003), given the concern aboutthe elevated particle background in the
Chandra data. Combin-ing their spectral results and X-ray luminosity (0.3–2 keV L X =1 . +0 . − . × erg s − for our adopted distance) with our sur-face brightness fit then implies a temperature T = 0 . +0 . − . keVand central density of . ± . cm − , which are the valueslisted in Table 2. This is a group in which the H I is clearly being stripped from thegalaxies at present, with much of it pulled into a 100 kpc longtidal tail extending to the southwest from the optical group centre(Borthakur et al., in prep.). In addition, the VLA data reveal a strik-ing H I trail extending to the east away from the group core, pro-truding from one of the galaxies in the field, Mrk 935. This galaxyis not included in the original Hickson (1982) catalogue, but is agroup member on the basis of its projected distance from the opticalgroup centre (6.7 arcmin ∼ kpc) and small radial velocity dif-ference of ∆ v ∼ km s − relative to the group mean, as listedin NED. The associated H I feature may therefore indicate ongoingstripping as the galaxy falls into the group. HCG 100 thus consti-tutes an excellent laboratory for the study of the processes wherebyH I is removed from individual galaxies and heated. Unlike the casefor the other Chandra observations presented here, this group wasobserved using the ACIS-I array, to allow the observed field to fullyencompass all of the interesting H I features mentioned above.The Chandra observation was split into two, so the imaginganalysis proceeded as for HCG 40. The combined imaging data, shown in Fig. 1, and the resulting surface brightness profile inFig. 3, do not reveal any clear indications of diffuse emission abovethe background level as evaluated outside r = 8 arcmin from thecorners of the ACIS-I array. Although Fig. 3 indicates a mild netexcess in the two innermost bins, the signal within this region issignificant at less than . σ . We also note that no extended sourcesare detected outside individual galaxies with ‘vtpdetect’, and thatthe group also remained undetected in RASS data (P96). With the σ – T relation suggesting T = 0 . ± . keV, the absence of anIGM detection inside r ≈ kpc ( r ≈ arcmin) from the opticalgroup centre implies a σ upper limit on the mean IGM density of h n e i < . × − cm − . The results presented so far demonstrate the presence of a de-tectable hot IGM within half of our sample only. However, even inthe absence of such gas, there could still be an intragroup mediumpresent with temperature or density below our detection limits (in-cluding any H I already stripped from individual galaxies, as evi-denced by the GBT detection of intergalactic H I in many of ourgroups).To explore the possibility that group galaxies could be inter-acting with such a medium, and to search for signs of galaxies be-ing stripped of any hot gas, we present in Fig. 4 a collage of allgroup members which were clearly identified as X-ray extendedsources by our source detection algorithms. These images wereadaptively smoothed following the procedure outlined in Section 3.Exceptions are HCG 30a and Mrk 935, for which a simple Gaussiansmoothing (with σ = 10 arcsec) was employed due to the very lowS/N. For the groups observed by XMM (HCG 7 and 15), the combi-nation of group distance and the broader EPIC PSF does not enablea clear distinction between point-like and diffuse emission, so noneof the relevant galaxies has been included in this figure. HCG 44b,lying on a chip gap in the
Chandra data and not covered by theoverlapping
XMM pointing, has also been excluded.The figure does not reveal any clear evidence for galaxies cur-rently being stripped of any hot gas. In particular, there are no in-dications of X-ray tails or bow-shock features indicating interac-tions with a surrounding medium. Such tails have been observedextending from galaxies in clusters (e.g., Wang, Owen & Ledlow2004; Sun & Vikhlinin 2005) but seem to be very rare in groups,with perhaps NGC 6872 and NGC 2276 the most prominent ex-amples (Machacek et al. 2005; Rasmussen et al. 2006). There is anindication of an asymmetric structure in HCG 97d even in the un-smoothed data, with a hint of a tail pointing south, but the sig-nal is too weak to exclude contamination by a faint point source.Mrk 935, with the remarkable H I tail extending to the east, alsopresents evidence of some X-ray asymmetry in this direction, butthe S/N is again too low to enable firm conclusions.In addition to diffuse galactic emission, X-ray point sourcesin individual galaxies are detected in all groups, with some of thesesources clearly associated with the galaxy nuclei. The relevanceof this is linked to the possibility that galaxies suffering strong(tidal) interactions could be showing enhanced nuclear activity, forexample associated with a nuclear starburst or strong AGN accre-tion fueled by a tidally induced gas inflow. For reference, Table 3lists detected X-ray sources whose position coincides with the op-tical centre of individual group members, with luminosities of anypoint-like component derived as described in Section 3. Note thatin some cases, such as HCG 97e and most of the XMM sources, wecannot clearly distinguish between nuclear and galaxy-wide diffuse c (cid:13) , 1–21 am pressure stripping in Hickson groups CHCG 44c HCG 37aD CB AHCG 40a-dHCG 30aB AHCG 100a,b HCG 44a E DAHCG 97a,d,eHCG 100: Mrk 935
Figure 4.
Smoothed 0.3–2 keV images of individual group galaxies show-ing diffuse X-ray emission. A horizontal bar marks a scale of 1 arcmin ineach case. emission, and the classification of these is followed by a ‘?’ in theTable 3. For HCG 7, however, the tentative identification of nuclearX-ray activity in three of the four principal galaxies agrees perfectlywith the
Spitzer far-infrared results of Gallagher et al. (2008), sug-gesting our identification is reasonably robust.Among the principal members in Hickson’s (1982) catalogue,we detect 22 candidate nuclear X-ray sources in 40 galaxies, withroughly two-thirds of these falling in the groups with a detectableIGM. The corresponding nuclear source fractions of ± and ± per cent in the groups with and without detectable hotgas, respectively, are statistically indistinguishable at the σ level.If instead splitting the sample according to H I deficiency, the cor-responding fractions are ± (high ∆ HI ) and ± per cent,a difference which is only just significant at σ . The median nu-clear X-ray luminosities for the two subsamples are also very sim- Table 3.
Overview of X-ray sources centred on individual group galaxiesas identified by our detection algorithms. Galaxy morphologies were takenfrom NED. Column 4 lists the unabsorbed 0.3–2 keV luminosity of anynuclear component.Galaxy Morph. Source L X , nucl (erg s − )HCG 7a Sa Nuclear? . ± . × HCG 7b SB0 Nuclear? . ± . × HCG 7c SBc Nuclear . ± . × HCG 15a S0 Nuclear? . ± . × HCG 15b S0 Nuclear? . ± . × HCG 15d S0 Nuclear? . ± . × HCG 15e S0 Nuclear? . ± . × HCG 30a SB0 Diffuse –HCG 30b SB0/a Nuclear . ± . × HCG 37a S0/E7 Diffuse + nuclear . ± . × HCG 37b Sbc Nuclear . ± . × HCG 40a E Nuclear . ± . × HCG 40b S0 Nuclear . ± . × HCG 40c SBb Diffuse? –HCG 40d SBa Diffuse? + nuclear . ± . × HCG 44a Sa Diffuse + nuclear . ± . × HCG 44b E Diffuse –HCG 44c SBa Diffuse –HCG 97a SB0 Diffuse + nuclear . ± . × HCG 97b Sc Nuclear . ± . × HCG 97c Sa Nuclear . ± . × HCG 97d E Diffuse? + nuclear . ± . × HCG 97e S0a Diffuse or nuclear . ± . × HCG 100a S0/a Diffuse + nuclear . ± . × HCG 100b S0/a Diffuse + nuclear . ± . × Mrk 935 S? Diffuse or nuclear . ± . × ilar, . × (high ∆ HI ) and . × erg s − , suggestingthat the above conclusions are not strongly biased by a system-atic difference in limiting X-ray flux between high– and low- ∆ HI groups. We also note that results from optical spectroscopy indi-cate that ∼ per cent of the principal members in HCGs in gen-eral show evidence for AGN activity, with a total of ∼ per centshowing emission lines from either AGN or star formation activity(Martinez et al. 2007). Our derived fractions are generally brack-eted by these values, suggesting that our results give a reasonablyreliable picture of the frequency of nuclear activity within our sam-ple. With the limited statistics available, there is thus no strongevidence from the X-ray data alone for enhanced nuclear activitywithin a certain kind of groups in our sample. Specifically, if inter-preting the X-ray bright or highly H I deficient systems as dynami-cally more evolved, we find no clear indication that the frequency orstrength of nuclear X-ray activity depends on the dynamical statusof the group. However, we note that this result applies to a smallsample and to the principal members only; a complete census ofgalaxy membership from optical spectroscopy would be requiredto extend this analysis to optically fainter group members and placethis conclusion on a more robust basis. The tentative lack of a clearenhancement in nuclear X-ray activity among the most H I defi-cient groups within our sample may tie in with the observation thatstar formation activity is not globally enhanced in HCG galaxiescompared to isolated ones (Verdes-Montenegro et al. 1998), as dis-cussed in more detail in Section 6. c (cid:13) , 1–21 J. Rasmussen et al. I deficiency and IGM properties The observed diversity in the diffuse X-ray properties of thesehighly H I deficient groups immediately suggests that galaxy–IGMinteractions are not the dominant mechanism for driving cold gasout of the galaxies within our sample and establishing the observedH I deficiencies. Of course, this conclusion neglects the fact that weare not uniformly sensitive to the presence of hot gas in the differ-ent groups. A quantitative comparison of observed H I deficienciesand derived IGM properties is therefore presented in Figure 5. Theleft panel shows ∆ HI and hot IGM mass as listed in Table 2, withboth quantities derived within the same region (inside r HI in theTable). Even when considering the X-ray detected systems alone,the obvious lack of a positive correlation between the two quanti-ties immediately suggests that the amount of hot gas in the groupcore is not a pivotal factor for H I removal. We note that an identicalconclusion is reached if replacing ∆ HI with ‘missing’ H I mass inthe plot. The strong σ upper limit ( M < . × M ⊙ ) on theIGM mass in the highly H I deficient HCG 7 (with the largest ‘miss-ing’ H I mass in the sample of ∼ . × M ⊙ ) only reinforcesthis conclusion. Also note, as pointed out in Section 4.1, that morerealistic assumptions about the IGM distribution in the X-ray unde-tected groups could reduce their upper limits to M IGM by perhapsan order of magnitude, but that this has no bearing on the aboveconclusions.Hence, neither is it surprising that ∆ HI does not show aclear dependence on the characteristic IGM ram pressure plottedin Fig. 5b and evaluated as the product of σ and the volume-weighted mean IGM density M IGM /V within the volume V =(4 / πr covered by the radio data, with all quantities taken fromTable 2. Note that the very compact HCG 40 – in which we cannotcompletely rule out a residual galactic contribution to the derivedIGM mass – stands out among the X-ray detected groups, witha characteristic ram pressure 1–2 orders of magnitude below thatseen in the X-ray bright systems. For the X-ray undetected groups,similar comments apply as for Fig. 5a.Finally, in Fig. 5c we investigate the dependence of ∆ HI on IGM temperature. Thermal evaporation of galactic H I throughheat conduction from the IGM is expected to proceed at a rate ˙ M ∝ T / if unsuppressed by, for example, magnetic fields. Giventhis strong temperature dependence, the lack of a positive ∆ HI – T correlation suggests that heat conduction is not an important effectwithin our sample. The location of the exceptionally H I deficientHCG 30 in Fig. 5c would seem to pose a particular challenge forthis mechanism.Overall, Fig. 5 therefore seems to confirm the notion thatH I deficiency is not tightly linked to the presence or nature of anIGM in these groups. There are some caveats to this interpreta-tion though. For example, it is worth emphasizing that it is the fourgroups with the highest velocity dispersion that are X-ray detected.If the remaining groups contain warm ( T . K) rather thanhot gas, and thus fall well below the σ – T relation for X-ray brightsystems, our constraints on the hot gas density could seriously un-derestimate the true IGM density in these systems. Unfortunately,this possibility cannot be directly tested with the present data. How-ever, the fact that our X-ray detected systems scatter fairly tightlyaround the Osmond & Ponman (2004) relation, as shown in the in-set in Fig. 5c, may support our use of this relation for predicting T also for the undetected systems. Furthermore, results for otherX-ray detected groups suggest, if anything, that the poorest sys-tems tend to have T on the high side for their velocity dispersion (Osmond & Ponman 2004) although the situation could, of course,be different for groups that remain X-ray undetected.Another concern is that σ V and hence P ram in Fig. 5b maynot be robustly determined, being based on just a handful of brightgalaxies in most cases. Large corrections to σ V would be needed,however, in order to affect our overall conclusions. A further pointis that any intergalactic H I already stripped from individual galax-ies could potentially contribute to the IGM mass and ram pressure.However, for the X-ray detected systems, the total H I mass insidethe GBT beam is of order 5 per cent of the corresponding hot gasmass (Borthakur et al., in prep.), suggesting that any cold gas canbe neglected for the present purposes.Despite the appearance of Fig. 5, it is premature to excludethe possibility that galaxy–IGM interactions could play a role forH I removal in some of our groups. For example, ram pressure strip-ping is expected to occur when the IGM ram pressure exceeds thegravitational restoring pressure of the galaxy. The efficiency of thisprocess therefore depends not only on the properties of the IGM,but also on those of the individual group galaxies. The process ofviscous stripping (Nulsen 1982) shares similar features, and couldbe operating even when ram pressure itself is insufficient to removeany H I . Finally, ram pressure may also indirectly affect the H I inthe disc. For example, many disc galaxy formation models pre-dict that massive spirals are surrounded by hot gaseous haloes fromwhich gas may cool out to provide fuel for ongoing star forma-tion in the disc (e.g., Toft et al. 2002). The removal of this coronalgas by external forces could contribute to H I deficiency, if the lim-ited supply of H I in the disc is consumed by star formation with-out being replenished from the hot halo (strangulation; see, e.g.,Kawata & Mulchaey 2008). We next seek to quantify the impor-tance of these various mechanisms. In an attempt to constrain the role of galaxy–IGM interactions fortypical disc galaxies in the individual groups, we constructed sim-ple analytical models of galaxies moving through the hot intragroupgas in the gravitational potential of each group. As explained inSection 4.1, the derived H I deficiency for each group should beviewed as an average for the group members, since it is often non-trivial to evaluate observed H I masses for the individual members.For our modelling purposes, we have therefore adopted a single,fiducial galaxy model, with overall properties broadly matched tothe fairly well-constrained mean properties of the late-type groupmembers in our sample.As described in detail below, the group potential and galaxyorbits are less well determined for each group. Consequently, weevolve the adopted galaxy model according to four different as-sumptions for each group, corresponding to two choices for thegravitational potential, and two for the galaxy orbits within the cho-sen potential. The variation among the resulting mass losses fromthe galaxy can serve as a means to gauge the uncertainties associ-ated with these assumptions. We will consider two different strip-ping processes for the cold gas in the disc, viz. classical ram pres-sure stripping and turbulent viscous stripping. For the gas in anyhot halo, we only consider ram pressure stripping for simplicity. c (cid:13) , 1–21 am pressure stripping in Hickson groups X-ray detectedX-ray undetected (a)
H15H37H40 H97 (b)
H15H37H40 H97 (c)
H15H37H40 H97
Figure 5. H I deficiency and (a) hot gas mass inside the region used for determining ∆ HI , (b) characteristic ram pressure, and (c) hot gas temperature for thevarious groups. Empty circles represent groups with no detectable hot gas. Inset in (c) shows velocity dispersion vs. T IGM for the X-ray detected groups, withthe Osmond & Ponman (2004) relation, i.e. equation (3), overplotted as a dashed line.
Our goal is to evaluate the efficiency of ram pressure strippingand related processes, without resorting to detailed numerical mod-elling, which is beyond the scope of this work. For this pur-pose, we would ideally adopt a single value of the ram pressurefor each group. The simplest approach is to assume the classi-cal analytical Gunn & Gott (1972) stripping criterion along witha constant ram pressure equal to its peak value. Hydrodynami-cal simulations involving a constant ram pressure have shown theGunn & Gott (1972) criterion to be remarkably accurate in termsof predicting the disc stripping radius and the mass of gas lost(Abadi, Moore & Bower 1999), and it remains a reasonable ap-proximation even when allowing for orbital variations in ram pres-sure (J´achym et al. 2007; Roediger & Br¨uggen 2007).However, recent hydrodynamical simulations of disc galaxiesmoving in radial (J´achym et al. 2007) and two-dimensional orbits(Roediger & Br¨uggen 2007) within a non-uniform gas distributionhave revealed an important exception to this rule. If the ram pres-sure changes faster than the characteristic stripping time-scale, aswill be the case for galaxies moving through a highly concentratedIGM, the Gunn & Gott (1972) criterion tends to overestimates thestripping efficiency. Since stripping is not instantaneous, an ISMelement may not always be accelerated to galactic escape velocitybefore the peak of the ram pressure is over. In such cases the gaswill eventually re-accrete, a possibility not taken into account bythe Gunn & Gott (1972) criterion. In the present study, these con-siderations could potentially be relevant for several of our groups,given the fairly small core radii resulting from the surface bright-ness fits. Instead of using the peak value of the ram pressure inour modelling, it therefore seems more sensible to adopt an orbit-averaged mean ram pressure by evaluating the time spent by thegalaxy at a given velocity and IGM density.To this end, we first derived total mass profiles M tot ( r ) foreach X-ray detected group, assuming a spherically symmetric gasdistribution in hydrostatic equilibrium, M tot ( < r ) = − kT ( r ) rGµm p (cid:18) d ln n ( r ) d ln r + d ln T ( r ) d ln r (cid:19) . (4)Since in general we have neither the statistics nor the spatial cov-erage to constrain T ( r ) to large radii, we made two assumptionsabout the temperature profile which are likely to bracket the actual temperature distribution in these somewhat disturbed groups. Weeither simply assumed T ( r ) equal to a constant mean value h T i , or T ( r ) = − . h T i log ( r/r ) + 0 . , appropriate for reasonablyundisturbed groups outside any cool core (Rasmussen & Ponman2007). In the latter case, r , the radius enclosing a mean den-sity of 500 times the critical value ρ c , was evaluated iteratively un-til convergence was reached. For h T i we used the measured valuelisted in Table 2.The resulting mass profiles were characterized analytically byfitting them with standard ‘NFW’ models (Navarro, Frenk & White1997), in which the dark matter distribution is described by ρ ( r ) = ρ c δ c ( r/r s )(1 + r/r s ) , (5)where δ c is a dimensionless parameter related to total group mass M , and r s is a scale radius which reflects the more commonlyused concentration parameter c = r /r s (see Navarro et al. 1997for details). With M , and hence δ c , fixed from the measured(extrapolated) mass profile, equation (5) was fitted to this profileto derive values of c under the two assumptions about T ( r ) de-scribed above. The results, summarized in Table 4, imply typicalvalues of M ≈ − × M ⊙ with c ∼ − . For thesegroup masses, derived concentration parameters are thus in goodagreement with expectations from cosmological N -body simula-tions (Bullock et al. 2001). Note that the assumption of a declin-ing temperature profile typically reduces the derived group massby ∼ per cent while increasing the halo concentration by a fac-tor of ∼ .For the galaxy orbital configuration within the derived gravi-tational NFW potential, we assume two different orbits, both radialand hence going through the group core. The galaxy is assumedto be experiencing a face-on IGM encounter in either case. Thesemaximizing assumptions allow us to estimate how important rampressure stripping can ideally be in our groups. The two orbits dif-fer only in terms of the assumed initial position and velocity of thegalaxy, with the galaxy initially at rest at a small clustercentric ra-dius in the first scenario, and falling towards the group core from alarge radius and with a high initial velocity in the second. Specifi-cally, the following two scenarios are considered:(i) For each group, we determine ¯ r , the observed (projected)mean clustercentric distance of the principal galaxies in each group. c (cid:13) , 1–21 J. Rasmussen et al.
Table 4.
Results of NFW fits to the derived group mass profiles under thetwo assumptions for T ( r ) described in the text. T isothermal
Group r r M c (kpc) (kpc) ( M ⊙ )HCG 15 342 541 2.0 4.5HCG 37 362 572 2.3 4.8HCG 40 357 565 2.2 4.0HCG 97 379 500 2.7 4.7 T declining
Group r r M c (kpc) (kpc) ( M ⊙ )HCG 15 317 474 1.3 10.0HCG 37 334 499 1.6 11.0HCG 40 318 474 1.3 9.0HCG 97 349 523 1.8 10.9 The model galaxy is assumed to be initially at rest at a larger clus-tercentric distance r , from which it falls freely towards the groupcentre. r is chosen such that when the galaxy reaches r = ¯ r , it hasattained a velocity corresponding to the observed group velocitydispersion. Typical values are r ∼ kpc and ¯ r ∼ kpc.(ii) The galaxy enters the group halo at r = r with a radialvelocity v r corresponding to the halo circular velocity at this radius, v r = ( GM/r ) / . Typical values are r ∼ kpc and v r ∼ km s − .In both cases we follow the galaxy until it turns around, havingcompleted one passage through the group core. The first scenario ischosen in an effort to match the observed average position (moduloprojection effects) and galaxy velocity in each group at present. Inpractice, it represents galaxy motion fairly close to the group core,with a mildly varying ram pressure, and so is somewhat reminis-cent of the ‘classical’ ram pressure scenario involving a constant,high ram pressure. The assumption underlying this orbit is extreme,however, in the sense that the true clustercentric distances will gen-erally be larger than the observed (projected) ones, which impliesthat galaxies will generally spend a larger fraction of their time atlarge distance than implied by this assumption. Therefore we alsoconsider scenario (ii) as a kind of opposite extreme. In this, galaxiesexperience a considerably higher peak ram pressure, but this occursonly relatively briefly. In the following, these two scenarios will bereferred to as orbit (i) and (ii), respectively.Although these two orbits do not necessarily encompass theextreme orbital solutions for the group members, they do representtwo rather different cases, thus offering a handle on the uncertaintyin the predicted mass loss related to orbital assumptions. Further-more, while completely radial orbits may not be very common, wenote that H I deficient spirals in clusters tend to have more eccentricorbits than non-deficient ones (Solanes et al. 2001), and that cos-mological infall along filaments would proceed in fairly eccentricorbits, thus lending some support to our simplifying assumption.For a detailed discussion of the impact of orbital parameters on thestripping efficiency, we refer to Hester (2006). For the galaxy model, needed to estimate the gravitational restor-ing force and hot halo thermal pressure of a ‘typical’ late-typegroup galaxy within our sample, we follow the general approachdescribed in Rasmussen et al. (2006) which is repeated here forcompleteness. The model consists of a spherical dark matter (DM)halo with density profile ρ h ( r ) = M h π / ηr t r exp ( − r /r )(1 + r /r ) , (6)a hot gaseous halo of the same form, a spherical bulge with ρ b ( r ) = M b πr r (1 + r/r b ) , (7)and exponential stellar and gaseous discs, each of the form ρ d ( R, z ) = M d πR z d exp ( − R/R d ) sech ( z/z d ) . (8)Here M b , M h , and M d are the total masses of each component, r b and r h are the scalelengths of the bulge and halo, respectively, r t isthe DM halo ‘truncation’ radius, R d is the cylindrical scalelengthof the disc components and z d the corresponding thickness, andfinally η = { − π / q exp ( q )[1 − erf ( q )] } − , (9)where q = r h /r t and erf is the error function. With this model, therestoring gravitational acceleration ∂ Φ ∂z ( R, z ) in the direction z per-pendicular to the disc can be evaluated analytically for each modelcomponent using the equations of Abadi et al. (1999), to whom werefer for more details.In order to constrain model parameters, stellar masses of theHCG members were evaluated from their B - and K -band mag-nitudes as listed in NED, following the prescription adopted byMannucci et al. (2005). For this purpose, only the principal mem-bers in our eight groups, as listed by Hickson (1982), with mor-phological types later than S0 were included, yielding a mean stel-lar mass of . × M ⊙ . For a subset of these galaxies (8out of 27), maximum disc rotational velocities, indicative of totalgalaxy masses, are also available in the Hyperleda database, witha mean value of 135 km s − . We note that, in terms of mean stel-lar mass, this subset is representative of the full sample, showing h M ∗ i = 4 . × M ⊙ . We therefore assume a stellar mass of × M ⊙ for the galaxy model, distributed such as to yield abulge-to-disc mass ratio of 1/4, appropriate for an Sb/c spiral. Usingthe relation of Haynes & Giovanelli (1984), we further assume anH I mass of . × M ⊙ to ensure that our model galaxy initiallyhas a ‘normal’ H I mass for its sample-averaged blue luminosity of L B = 1 . × L ⊙ .The scalelengths of the stellar and gaseous disc componentsin the model were chosen to ensure that at least 85 per cent of thestellar and H I mass resides within the average optical disc radius of r D ≃ kpc, as derived from the size of the D ellipse for eachspiral member in our groups. For simplicity, the gas distribution inthe hot gaseous halo is assumed to follow that of the underlyingdark matter, but with a smaller value of r t corresponding to twicethe ‘size’ r D of the stellar disc, and with a total mass correspondingto the H I mass in the disc. The hot halo is assumed to be isothermalat a temperature T ∼ . keV, the virial temperature correspond-ing to the maximum allowed disc rotational velocity in the model,taken to be ∼ km s − . Once the baryonic model componentshave been specified, the parameters of the DM halo are effectively c (cid:13) , 1–21 am pressure stripping in Hickson groups Table 5.
Adopted parameters of the galaxy model. L is the characteristicscalelength for each component, i.e. r h and r t for the haloes, r b for thebulge, and R d and z d for the disc components.Component M total L ( M ⊙ ) (kpc)DM halo 28 5, 100Hot gas halo 0.7 5, 20Stellar bulge 0.8 0.5Stellar disc 3.2 3, 0.25Cold gas disc 0.67 3, 0.25 set by this maximum allowed rotation velocity, and by the require-ment that the model baryon fraction match the universal value of ∼ per cent. Table 5 summarises the adopted model galaxy pa-rameters. Note that the cold gas component in the model refers tothe distribution of H I only, as we do not consider any moleculargas here. In summary, this model roughly reproduces the averagestellar mass, bulge-to-disc ratio, disc size, and maximum disc rota-tional velocity seen for the spirals in our groups, with a total baryonfraction consistent with the universal value, and an initial H I massas expected for a non-stripped isolated galaxy with these properties. Combining equation (5) with the measured density distribution ofintragroup gas, time–averaged values of IGM density h ρ i and thesquare of the orbital velocity h v i experienced by the galaxy inits orbit can be evaluated. For the averaging time-scale, we onlyconsider the segment of the orbit for which the ram pressure ex-erts a significant influence on the ISM, taken to be from the time atwhich the instantaneous ram pressure, if sustained, would removeat least 1 per cent of the cold gas. This is to avoid the artificial sup-pression of h ρ i and h v i that would otherwise result from includingthe time spent by the galaxy at large radius where ram pressure iscompletely negligible. Note that the corresponding characteristicram pressure is independent of group mass and orbital initial con-ditions, as it depends only on the assumed galaxy model.Assuming a face-on IGM encounter, the condition for ram-pressure stripping is then evaluated as Σ g (cid:16) ∂ Φ b ∂z + ∂ Φ h ∂z + ∂ Φ g ∂z + ∂ Φ ∗ ∂z (cid:17) < h ρ ih v i , (10)where ∂ Φ ∂z ( R, z ) is the restoring gravitational acceleration in thedirection z perpendicular to the disc, originating from the stellarbulge (subscript ’b’), dark matter halo (’h’), gaseous disc (’g’) andstellar disc (’ ∗ ’), respectively, and Σ g is the surface density of coldgas. Equation (10) is similar to the classical Gunn & Gott (1972)stripping criterion, but takes into account the mass distribution inthe galaxy rather than simply assuming a homogeneous disc thinenough to be described solely by its surface density. Its solutionalso provides us with the ‘stripping region’, the region in the ( R , z )–plane from which gas is permanently lost by the galaxy.Apart from conventional ram pressure stripping, transportprocesses such as viscous stripping (Nulsen 1982), caused byKelvin–Helmholtz instabilities arising at the ISM–IGM interface,could also play a role even when the ram pressure itself is in-sufficient to remove galactic gas (Quilis, Moore & Bower 2000;Rasmussen et al. 2006). Turbulent viscous stripping of a gas disc of radius r D is expected to operate at Reynolds numbers Re & ,where Re = M r D /λ, (11) M is the Mach number of the IGM flow past the galaxy, and λ ∝ T n − is the ion mean free path in the IGM. The expected mass-loss rate due to this process (Nulsen 1982), ˙ M vs ≈ . πr D ρv gal , (12)scales only linearly with galaxy velocity and so could be importantin a wider range of environments than ram pressure itself. We in-clude this process in the stripping calculations by evaluating equa-tions (11) and (12) at each point in the orbit and adding up thetotal mass loss. Dealing with disc galaxies rather than spheroids,we have used only half the Nulsen (1982) mass loss rate in equa-tion (12) because of the correspondingly smaller galaxy surfacearea for a given r = r D . For r D itself, we use the smaller ofthe sample-averaged value of r D = 10 kpc and the stripping ra-dius predicted by equation (10). Note that M in equation (11)even if the galaxy is moving supersonically, as the post-shock IGMflow past the galaxy will always be subsonic. Also note that thestripping efficiency of both ram pressure and viscous strippingshould be largely unaffected by the presence of a shock front (seeRasmussen et al. 2006).Having dealt with the stripping of cold gas from the disc, wenow turn to the removal of any gas situated in a hot galactic halo.It seems plausible that, at the very least, the cooling and gradualinflow of any such gas will be disrupted once the external rampressure P ram exceeds its thermal pressure P th . The simulationsof Mori & Burkert (2000) confirm that this condition provides areasonable estimate of the mass of a galaxy that will have its hothalo completely stripped by ram pressure. For the purpose of alsoassessing the importance of strangulation for the model galaxy, wetherefore use the derived values of h ρ i and h v i to evaluate the‘strangulation region’ where P ram > P th , and the correspondingmass of affected coronal gas. The aim here is only to develop arough picture of the potential importance of IGM interactions forstrangulation, so the effects of viscous stripping are not consideredfor the hot halo.We readily acknowledge that our modelling approach is infe-rior to detailed numerical models (e.g., Hester 2006), and hydrody-namical simulations in particular. It is our hope that one can nev-ertheless have some confidence in the results, given that we haveemployed a reasonably detailed galaxy model and are making someallowance for orbital variations in ram pressure. The advantage ofthe adopted method is that it is easily tailored to the specific condi-tions in individual groups at little computational cost. We re-iterate that we are considering four scenarios for each group,consisting of two separate assumptions about the galaxy orbit, andtwo about the IGM temperature (and hence total mass) distributionin the groups, as specified in Section 5.1. The results of the H I stripping calculations for each of these four cases are summarizedin Table 6, which lists the derived r.m.s. orbital velocity v rms , theorbit-averaged IGM number density, the mass of H I lost due to rampressure and viscous stripping, and the total fraction of H I strippedfrom the initial model reservoir of . × M ⊙ . Recall that v rms is computed only from the point in the orbit where ram pressurebecomes significant and so depends not only on orbital parametersbut also on the IGM distribution. c (cid:13) , 1–21 J. Rasmussen et al.
Table 6.
Orbit–averaged galaxy velocities and IGM densities, along withpredicted H I mass loss ∆ M due to ram pressure (‘rp’) and viscous stripping(‘vs’) after one passage through the group core for the assumed orbits andgroup mass profiles. The final column lists the total fraction f of H I lost. Orbit (i), T isothermal
Group v rms h n i ∆ M rp ∆ M vs f (km s − ) (cm − ) ( M ⊙ ) ( M ⊙ )HCG 15 354 . × − . × − . × − . × − Orbit (i), T declining
Group v rms h n i ∆ M rp ∆ M vs f (km s − ) (cm − ) ( M ⊙ ) ( M ⊙ )HCG 15 404 . × − . × − . × − . × − Orbit (ii), T isothermal
Group v rms h n i ∆ M rp ∆ M vs f (km s − ) (cm − ) ( M ⊙ ) ( M ⊙ )HCG 15 498 . × − . × − . × − . × − Orbit (ii), T declining
Group v rms h n i ∆ M rp ∆ M vs f (km s − ) (cm − ) ( M ⊙ ) ( M ⊙ )HCG 15 448 . × − . × − . × − . × − As can be seen from the Table, the amount of gas lost througheither stripping process is generally an appreciable fraction of theinitial H I mass. Viscous stripping in particular can remove a sub-stantial fraction of the cold disc gas for all model assumptions.HCG 40 is the exception, with the relatively tenuous IGM in thisgroup removing at most 10–15 per cent of the H I . The steeper IGMdensity profile in this group also implies that a larger fraction of thetotal IGM mass is encountered at high velocity, leading to a higher ∆ M rp / ∆ M vs ratio than for the other groups, and a higher valueof v rms in orbit (ii), because the ram pressure becomes significantcloser to the core in this system.The Table further shows that the main variation in the com-puted mass loss for a given group and stripping mechanism derivesfrom the choice of orbit rather than the assumed group mass pro-file. Orbit (i) is generally more efficient at removing gas throughram pressure stripping than orbit (ii), because although the peakram pressure is considerably higher in the latter case, the galaxyspends a comparatively shorter time in regions corresponding tohigh values of P ram . Conversely, viscous stripping is more efficientin orbit (ii), as this mechanism acts even at relatively low v gal , al-lowing the associated mass loss to build up significantly over themuch longer crossing time-scale relevant for this orbit. Note, how- | z | ( k p c ) HCG 15HCG 37 HCG 40HCG 97
Figure 6.
Isobars of gravitational restoring pressure (dashed) and hot halothermal pressure (dotted) for our fiducial model galaxy in the differentgroups. Dashed lines outline the galactic regions outside which H I in thedisc can be stripped by ram pressure, corresponding to equality in equa-tion (10). Dotted lines show the corresponding regions for the hot halo gas.The order of the contours is the same in both cases. ever, that for a given group, the outcome in terms of total H I massloss, ∆ M tot = ∆ M rp +∆ M vs , is largely insensitive to the variousorbit and mass profile assumptions, despite clear variations in h n i and v rms among the four scenarios. The only exception to this isthe significantly more massive HCG 97, for which the deeper grav-itational potential and higher IGM mass implies that orbit (ii) isrelatively more efficient at removing galactic gas than for the othergroups.Figure 6 outlines the stripping region for a model galaxy ineach of the X-ray detected groups, based on solving equation (10).The figure shows the one of our four scenarios in which the effect ofram pressure is generally most pronounced, i.e., orbit (i) with T ( r ) declining. In the outer disc, the gravitational restoring pressure, andhence the stripping region for the cold gas, is seen to be nearly in-dependent of vertical disc height for interesting values of | z | , andthe gravitational restoring force at a given R peaks well above thedisc. Compared to the ‘size’ of the stellar disc ( r D = 10 kpc), itis clear that the gas disc becomes mildly truncated by ram pressurein HCG 97 and 37 but remains largely unaffected in the other twogroups. Note that this truncation reduces the viscous stripping effi-ciency by reducing the surface area of the gas disc exposed to theIGM. Viscous stripping is therefore slightly more efficient for lowvalues of P ram , adding to the explanation of the higher values of ∆ M vs for orbit (ii) in Table 6.Predicted H I deficiencies corresponding to the mass losses inTable 6 are compared to the observed values in Fig. 7. Errors onpredicted deficiencies correspond to the full range in predicted H I mass loss for each group under the various orbit and mass pro-file assumptions. Shown are the expectations from ram pressurestripping alone, as well as the full H I mass loss from combiningequations (10) and (12). The dashed line in the figure representsequality between modelled and observed H I deficiencies; under ourmodel assumptions, anything below this line cannot be explainedby galaxy–IGM interactions alone. The plot clearly suggests thatram pressure stripping on its own is not sufficient to cause the ob- c (cid:13) , 1–21 am pressure stripping in Hickson groups r.p. + visc. strippingr.p. stripping only H97 H97H37H37 H15H15 H40H40
Figure 7.
Observed H I deficiencies compared to the model predictions ofTable 6 for the stripping of H I by ram pressure alone (empty diamonds),and by ram pressure plus viscous stripping (shaded). Dashed line representsequality between observed and predicted ∆ HI . Table 7.
Fraction of hot halo gas lost due to ram pressure in the variousorbital scenarios.Group Orbit (i) Orbit (i) Orbit (ii) Orbit (ii) T ( r ) const. T ( r ) decl. T ( r ) const T ( r ) decl.HCG 15 0.53 0.64 0.31 0.27HCG 37 0.74 0.85 0.39 0.35HCG 40 0.31 0.46 0.21 0.18HCG 97 0.84 0.94 0.52 0.48 served H I deficiencies, even in this X-ray bright subsample of ourgroups. When including viscous stripping, however, if as efficientin removing H I as assumed here, galaxy–IGM interactions can cer-tainly help explain observed values of ∆ HI , but they can potentiallyfully account for the H I loss only in HCG 37 and HCG 97, i.e. injust two out of our eight groups.Regarding the issue of strangulation, Fig. 6 also illustrates thederived stripping region for hot halo gas according to the adopted P ram > P th criterion. The figure suggests that ram pressure alonecould remove a sizable fraction of the halo gas in the model. Ta-ble 7 lists the fractions of stripped halo gas in the different scenar-ios, showing that these can be substantial for orbit (i) in particular,peaking at ∼ per cent in HCG 97. Viscous stripping, not in-cluded here, could potentially contribute beyond these estimates.This suggests that galaxy–IGM interactions in X-ray bright groupscould play an important role in removing the gas supply that mayultimately fuel star formation in spirals, in qualitative agreementwith the simulation results of Kawata & Mulchaey (2008). For ourspecific model setup, there is a large dispersion in the fraction ofgas affected, however, and it is not clear that the effect would beimportant in groups such as HCG 40. It is important to stress that the calculations presented here are onlyintended to provide a rough picture of the impact of galaxy–IGMinteractions in our sample, and we do not claim that these results areanything but indicative. Caveats include the fact that the adoptedgalaxy model parameters provide a plausible, but not necessarilyunique, representation of the spirals in our sample. It is also apossibility, albeit one we cannot easily evaluate, that some groupmembers may be individually less H I deficient than the ‘group-averaged’ value of ∆ HI , and so could have H I deficiencies con-sistent with removal by ICM interactions alone, even if our modelresults suggest otherwise for the group members as a whole.As regards the model calculations, an important limitation isthe fact that the model is completely static. In practice, the gasdistributions both in the groups and their galaxies will be evolv-ing, which could be particularly relevant for orbit (ii) where orbitaltime-scales are several Gyr. Also, the estimate of viscous strippingmass loss does not take into account that this process itself will re-duce the size of the gas disc and hence the mass loss rate accordingto equation (12). Furthermore, this stripping process may saturate(Nulsen 1982) which would reduce the mass loss, the presence ofmagnetic fields could also suppress hydrodynamical instabilities,and the presence of a hot halo could perhaps to some extent shieldthe cold gas disc from such instabilities. The latter could be of par-ticular relevance for orbit (ii), in which ∆ M vs is high but where acomparatively smaller fraction of halo gas is lost due to ram pres-sure. These considerations suggest that the estimated contributionby viscous stripping to the H I mass loss should be regarded as anupper limit.With their compact galaxy configurations and low velocitydispersions, our groups could also represent environments in whichgalactic dark matter haloes are subject to significant tidal trunca-tion. Hence, another issue is to what extent our results are affectedby the adopted assumptions on dark matter in the galaxy model.For the adopted disc and bulge parameters, our freedom to modifythe assumed DM distribution is mainly constrained by the require-ment that the maximum disc rotational velocity v max of the modelshould not exceed the allowed ∼ km − . A full exploration ofmodel parameter space is beyond the scope of this work, and werefer to Hester (2006) for a more thorough investigation of these is-sues. Nevertheless, to provide a rough picture of the impact of DMfor our results, we repeated the stripping calculations of Section 5.3with three different modifications, subject to the above v max crite-rion:(i) Assuming no DM [or larger DM halo scalelengths r t and r h in equation (6)]. This must be regarded as an extreme assumption,which should facilitate stripping.(ii) Increasing the DM mass by a factor of two (for which the v max criterion then requires at least one of the DM scalelengths r t and r h to increase by a similar factor). This should suppress stripping.(iii) Assuming a higher DM concentration (i.e. lower r h or r t ) bya factor of three (suppresses stripping), which in turn requires alower total DM mass by a factor of five (facilitates stripping). Thesechoices comply with the typical tidal truncation of DM halos in-ferred for galaxies in massive clusters (Limousin et al. 2007).We find that, in all three cases, the amount of H I lost by the modelgalaxy is at most modified by 10–15 per cent, regardless of thegroup and orbit considered. This perhaps somewhat surprising re-sult has its origin in the fact that the vast majority of cold gasin the model is located at low galactocentric distances, where the c (cid:13) , 1–21 J. Rasmussen et al. restoring gravitational acceleration – which determines the strip-ping region in accordance with equation (10) – is dominated bythe baryonic disc components and not the dark matter halo. Thevariations in the stripped gas mass for the hot halo can be larger,up to 45 per cent for HCG 40, but they can generally easily be ac-commodated by the variations associated with the different orbitalassumptions. We therefore conclude that our results are reasonablyrobust to changes in the assumed amount and distribution of darkmatter in the galaxy model.
The groups in our sample are all H I deficient even when account-ing for any intergalactic H I not clearly associated with individualgalaxies. Removal of H I from the group members is by itself insuf-ficient to explain this situation, as the removed gas must also be pre-vented from staying neutral. Thus, unless H I is somehow destroyed in situ within the galaxies, it must go through a two-stage processwhereby it is first removed from the galaxies and then ionized bya possibly unrelated mechanism. Here we discuss these differentpossibilities in light of our X-ray and modelling results. In situ destruction of H I The H I could potentially have been destroyed within the galaxiesthemselves, either by evaporation though thermal conduction fromthe IGM, or through consumption by star formation, provided thereis no continuous replenishment of cold material. A third possiblemechanism involves heating and possibly ejection from the galax-ies by starburst winds. While the first possibility, direct heating bythe IGM, seems implausible in light of the absence of a positive cor-relation between ∆ HI and T IGM (cf. our discussion of Fig. 5c), thetwo other scenarios deserve some further attention. H I consumptionby star formation could help establish observed deficiencies, but aprerequisite is that the consumed gas is not simply replenished ata similar rate, for example through the cooling out of hot, coro-nal material on to the disc. Our model results for the X-ray brightgroups (Table 7) suggest that such a strangulation scenario could atleast be greatly facilitated by the removal of coronal gas due to rampressure, but a more detailed exploration of model parameter spacewould be required to assess the general validity of this conclusion.In this context, it is instructive to compare the resultsfrom our simple analytical model to the simulation results ofKawata & Mulchaey (2008). These authors investigate the effi-ciency of ram pressure stripping and strangulation for a galaxyin a small galaxy group on the basis of a cosmological smoothedparticle hydrodynamics (SPH) simulation. They do not specificallyconsider a ‘compact’ group, but this has the advantage that theirtarget galaxy is not subject to noticeable gas loss from tidal interac-tions (D. Kawata, priv. comm.), so in this sense a comparison to ourmodel calculations is justified. Their target galaxy has initial prop-erties broadly similar to those of our galaxy model in terms of stel-lar, gaseous, and total mass. It is followed for one passage throughthe group as it enters a group of virial mass M ≈ × M ⊙ from an initial position roughly corresponding to twice the virialradius. The galaxy orbit thus shows some similarities to our orbit(ii), but, with a pericentre at r ∼ kpc, does not take the galaxythrough the very core of the IGM distribution. Kawata & Mulchaey(2008) find that the resulting ram pressure induces strangulation toa significant degree. Star formation and thus H I consumption be-comes mildly enhanced during infall, and the hot halo gas is almost completely removed during the group passage. The combination ofthese processes has effectively consumed the H I and quenched starformation by the time the galaxy re-emerges at the virial radius.Our model conclusion that ram pressure stripping maystrongly affect any coronal gas in the X-ray bright groups is thus inencouraging agreement with these simulation results, and it is evenpossible that our model underestimates the importance of ram pres-sure in this context. This is made more relevant still if ram pressure– or other galaxy interactions with the group environment – indi-rectly accelerate strangulation by enhancing the disc star formationrate and hence the consumption of H I . This possibility draws ob-servational support from the strong star formation activity seen inthe ram-pressure affected group spiral NGC 2276 (Rasmussen et al.2006), as well as from recent hydro-simulations (Kronberger et al.2008). However, Verdes-Montenegro et al. (1998) noted on the ba-sis of star formation rates derived from IRAS far-infrared luminosi-ties that there is no indication of enhanced star formation amongthe HCG galaxies compared to the level seen in isolated galaxies.In fact, current star formation rates in these groups generally seemtoo low to explain the missing H I by consumption through star for-mation. For the spirals in our sample, a mean star formation rate of ˙ M ∗ ≈ . M ⊙ yr − can be derived for the IRAS -detected galaxies.However, many of our galaxies remain undetected in one or more
IRAS bands and so have only upper limits to ˙ M ∗ (if including theseupper limits in the mean, the result is ˙ M ∗ < . M ⊙ yr − ). Atthese rates, the time-scale for star formation to exhaust an initial H I supply of . × M ⊙ to current levels is at least 5 Gyr, even ifneglecting the return of unprocessed material to the ISM and anycosmological accretion of gas by the galaxy.Thus, while our model calculations and the results ofKawata & Mulchaey (2008) suggest that hot halo gas can be re-moved fairly efficiently in several of our groups, potentially inhibit-ing or at least suppressing any replenishment of disc H I from a hothalo, strangulation does not seem to have played an important rolein establishing current H I levels within our sample. At the observedstar formation rates, the remaining H I can continue to fuel star for-mation for many Gyr without drastically reducing the H I supply.Of course, star formation rates could have been much higher inthe past, which could also indirectly have contributed to exhaustingthe gas supply through the ejection of gas from the disc by star-burst winds. However, unless ˙ M ∗ in these galaxies has generallydeclined to current levels only fairly recently or has been affectedby the group environment over cosmological time-scales, the impli-cation seems to be that destruction of neutral hydrogen within thegalaxies themselves cannot explain the H I deficiencies. The pres-ence of intergalactic H I in some of our groups also shows that thisscenario cannot provide an exhaustive explanation. We are there-fore compelled to also consider the alternative, externally–drivenremoval and destruction of H I . I removal by external forces Focussing first on the removal of H I , both galaxy–IGM and galaxy–galaxy interactions could be envisaged to play a role. Our resultsdemonstrate that although earlier studies indicated a link betweensignificant H I deficiency in Hickson groups and the presence ofhot intragroup gas (Verdes-Montenegro et al. 2001), there is clearlyno one-to-one correspondence between the two. The lack of adetectable intragroup medium inferred here for half of the eightmost H I deficient compact groups specifically appears to rule outIGM dynamical interactions as generally dominant for H I removalwithin these environments. Ram pressure and viscous stripping c (cid:13) , 1–21 am pressure stripping in Hickson groups could nevertheless still have played a role for H I removal in our X-ray detected systems. Our modelling results indicate, however, thatthe efficiency of these processes is generally insufficient to fully ac-count for the H I missing from the individual group members, espe-cially if our simplistic treatment of viscous stripping overestimatesthe predicted H I loss, as hypothesized in Section 5.5.Overall, the simulations of Kawata & Mulchaey (2008) seemto support these conclusions. In these, ram pressure affects theamount of cold gas in the disc to an even lesser degree than inour model, although this could perhaps be attributed to the factthat our groups are at least twice as massive, and perhaps also tothe maximizing orbital assumptions adopted in our model. Viscousstripping does not appear to be important in the simulations, but asnoted by Kawata & Mulchaey (2008), such processes are not nec-essarily well treated by SPH schemes, so a direct comparison to ourresults may not be meaningful.The limited impact of galaxy–IGM interactions inferred forour groups commands an alternative explanation for the H I re-moval. Tidal stripping constitutes an obvious candidate in thesedense, low- σ environments, particularly in light of the fact thatthe four X-ray undetected groups within our sample exhibit thelowest velocity dispersions and so could be expected to representenvironments where tidal interactions should be most important.However, even if for now ignoring the problem of subsequentlyheating the removed H I , it is not immediately clear that tidal in-teractions would necessarily result in increased H I deficiency , asthey would also affect the stellar component in the galaxies. If starsand gas are removed in roughly equal proportion, and if the re-moved stars become part of any undetected intracluster light (cf.Gonzalez, Zaritsky & Zabludoff 2007) while the H I remains de-tectable, this could potentially even result in an H I excess. Recall-ing that H I deficiency is defined here on the basis of the observed B -band galaxy luminosity, increasing ∆ HI through tidal strippingtherefore requires either the preferential removal of cold gas com-pared to stars (and its subsequent heating), or that L B is simultane-ously boosted relative to the H I mass for the non-stripped compo-nents.The latter possibility gains support from the observation thatspecific star formation rates generally tend to be higher in galaxieswith close neighbours (Li et al. 2008), with enhanced nuclear starformation plausibly arising as a consequence of tidally induced gasinflow. It is perhaps curious then, as mentioned above, that there isno evidence for enhanced star formation in Hickson groups com-pared to the level in isolated galaxies. This runs contrary to expec-tations for interacting galaxies, and would seem to argue against L B being significantly boosted in the HCG members relative totheir remaining stellar or H I mass. The other possibility mentionedabove, that H I is more easily tidally stripped than the stellar com-ponent, is perhaps more promising. In many of our groups, such asHCG 100 (cf. Section 4.1.8), significant amounts of intergalactic H I is detected with the GBT, whereas the stellar components are notnoticeably affected. A possible explanation is that a relatively largerfraction of H I compared to stars initially resided at large galacto-centric radii, where tidal removal would be most efficient. The factthat the H I disc in typical spirals and late-type dwarfs is often moreextended than its optical counterpart, with the radial distribution ofH I declining less steeply with radius than that of the B -band light(Broeils & Rhee 1997; Swaters et al. 2002), seems to support suchan explanation.Tidal stripping clearly is taking place in some of our groups,notably in the X-ray undetected groups HCG 100 and HCG 44. Inthe latter, both optical (Fig. 1) and H I data (Borthakur et al., in prep.) show strong evidence for the SBc galaxy HCG 44d beingtidally stripped. This may indicate that tidal stripping is the primarymechanism by which H I is removed from the galaxies in these twogroups. Although it is tempting to extend this conclusion to all theX-ray undetected groups in our sample, and perhaps even beyond,the inconspicuous star formation rates in Hickson groups in gen-eral, and the tentative lack of enhanced nuclear X-ray activity inthe highly H I deficient systems in particular (Section 4.2), may notargue in favour of such an explanation. I As emphasized in the beginning of this Section, H I deficienciescan only be explained if the hydrogen, once removed from thegroup members, is also transformed from its neutral phase. Irre-spective of the processes accomplishing its removal, a mechanismmust therefore also be invoked for ionizing the H I during or follow-ing its transfer to intergalactic space. In our X-ray bright groups, acandidate process is readily available, since any removed H I is ex-pected to evaporate due to heating by the ambient IGM at a rate ˙ M ∝ T / , almost independently of realistic values of the IGMdensity (Spitzer 1962). A detailed comparison of the X-ray and H I properties of these groups should help test this explanation and willbe presented elsewhere (Verdes-Montenegro et al., in prep.). In theX-ray undetected groups, where any IGM is expected to be rela-tively cool (cf. Fig. 5c) if at all present, the picture is less clear-cut. Even if these groups do contain an IGM, the lower predictedIGM temperatures in this subsample imply average IGM heatingtime-scales an order of magnitude above those in the X-ray brightsystems, casting doubt on whether this mechanism would be suffi-cient.If gas is predominantly removed by tidal interactions in theX-ray undetected groups, it is instead conceivable that some of theH I has been heated by tidal shocks, although the presence of inter-galactic H I in some of the groups would imply heating time-scaleswell in excess of those associated with the H I removal itself. An-other possibility is that the column density of any removed H I be-comes too low for the gas to be self-shielding against ionization bycosmic UV radiation. If so, much of the undetected hydrogen inthese groups could potentially be in the form of a tenuous, photo-ionized intergalactic plasma. A quantitative investigation of thesepossibilities is the subject of future work, but at present a clear pic-ture of the fate of the removed H I in the X-ray undetected groupsremains elusive. Based on a sample of eight Hickson compact groups selectedfor their high H I deficiencies, we have used Chandra and
XMM-Newton data to assess the properties of any hot intragroup medium(IGM) and constrain the role of galaxy–IGM interactions in remov-ing H I from the galaxies in these groups. The X-ray analysis re-veals a detectable IGM in four of the eight groups. We have tenta-tively identified the detected diffuse emission in HCG 40, a spiral-dominated group, as associated with an intragroup medium, butthe combination of a low signal-to-noise ratio and an exceptionallycompact galaxy configuration precludes a highly robust conclusionfor this particular system. The remaining three groups are all fairlyX-ray luminous, showing substantial amounts of intergalactic hotgas with a somewhat disturbed morphology in all cases. c (cid:13) , 1–21 J. Rasmussen et al.
The remarkable X-ray diversity seen across the sample imme-diately suggests that the presence of a significant IGM is not a dom-inant factor in establishing observed H I deficiencies, despite ear-lier results indicating such a connection (Verdes-Montenegro et al.2001). It is particularly notable that some of the most H I deficientgroups show no detectable hot IGM, including HCG 30 which onlycontains a few per cent of the expected H I mass for its galaxy con-tent. A comparison of H I deficiency with either hot IGM mass orcharacteristic ‘mean’ ram pressure confirms the lack of a clear cor-relation even for the X-ray bright systems, although statistics arenaturally too limited to enable firm conclusions on this basis alone.The H I deficiency does not seem to depend on IGM temperature ei-ther, suggesting that heat conduction from the IGM does not play animportant role in destroying galactic H I (although once removed,this gas is likely to evaporate on fairly short time-scales in the X-ray bright groups).From fitting analytical models to the derived mass profiles ofthe X-ray detected groups, we have constructed plausible modelsof the gravitational potential and associated radial galaxy orbits foreach of these groups. Combined with the inferred IGM distributionsand a numerical model of a late-type galaxy with properties broadlymatching those of our observed spirals, this has enabled estimatesof the importance of ram pressure stripping and viscous stripping inremoving H I from the late-type galaxies in each group. The resultsindicate that, even under maximizing assumptions about the galaxyorbit, ram pressure stripping will remove only small amounts ofcold gas from the group members, peaking at 10–25 per cent in theX-ray bright HCG 97. We find that viscous stripping is generallymore efficient, with the combination of the two processes capableof removing more than half of the cold ISM in HCG 97 and po-tentially fully accounting for the missing H I mass in both HCG 37and HCG 97. However, the efficiency of viscous stripping is likelyoverestimated with our simple analytical approach, and yet theseprocesses are insufficient in terms of explaining the H I deficiencyof the X-ray detected HCG 15 and HCG 40.The model results also indicate that ram pressure can effi-ciently remove a large fraction of any hot galactic halo gas thatmay otherwise act as a supply of fresh material for star forma-tion in the disc. However, even if the gas supply to the disc canbe completely cut off, gas consumption at the typical star forma-tion rates in the groups would proceed far too slowly to explainthe observed shortfall of H I by itself. Much higher star formationrates in the recent past are required for this process to have had anysignificant impact. This may suggest that the observed H I deficien-cies are not caused by in situ destruction of H I within the galaxiesthemselves. It remains a possibility that even modest star forma-tion activity could have heated some of the H I and lifted it abovethe disc midplane where it would be more susceptible to removalby ram pressure, similar to the situation proposed for NGC 2276(Rasmussen et al. 2006). The absence of observational signaturesof this process, e.g., in the form of a hot gas tail extending fromany of the group members, may suggest that such a mechanism isnot generally very important within our sample though.By the process of elimination, it seems plausible that tidal in-teractions have played a key role for H I removal in the groups, par-ticularly in those systems containing no detectable IGM. In orderto explain the H I deficiencies , this scenario would likely requirepreferential removal of H I over stars, as perhaps facilitated by themore extended distribution of cold gas relative to stars in typicallate-type galaxies. While it is perhaps not surprising that tidal in-teractions are affecting the gas content of galaxies in these compactgroups, the tidal stripping explanation still faces some outstanding issues. Among these is the expectation that such interactions wouldgenerate enhanced star formation or nuclear activity, but there isno indication that the X-ray faint or highly H I deficient systems inour sample show evidence for increased such activity (although thefrequency of nuclear activity in galaxies in compact groups in gen-eral may be rather high; Martinez et al. 2007). If interpreting theX-ray bright or highly H I deficient systems within our sample asdynamically more evolved, we thus find no clear evidence that thefrequency or strength of nuclear X-ray activity in the group mem-bers depends on the dynamical status of the group. It also remainsunclear whether tidal interactions themselves can destroy the H I during or following its removal from galactic discs and so fully ex-plain the H I deficiency in any of our groups.In closing, our results suggest that galaxy–IGM interactionscan have played a role for the removal and destruction of H I insome of our groups, but a complete understanding of the originof the observed H I deficiencies and the processes causing it isstill lacking. Strangulation or thermal evaporation do not emergeas important contenders, and typical indirect signatures of tidalinteractions, such as enhanced star formation or nuclear X-rayactivity, are not more pronounced within the more H I deficienthalf of our sample. The latter seems in line with previous results(Verdes-Montenegro et al. 1998) which indicate that star formationrates in Hickson compact groups are not globally enhanced relativeto the field. We note here that this result could potentially be mis-leading, however, perhaps masking an evolutionary trend in whichgalaxies initially experience enhanced star formation which is thenfollowed by an environment–driven suppression. A detailed corre-lation of H I and X-ray morphology in the groups, coupled with abroad comparison of individual galaxy properties such as specificstar formation rates, may therefore shed further light on the fateof the missing H I in these compact systems. This is the subject offuture work. ACKNOWLEDGMENTS
We thank the referee for useful comments which helped to clar-ify the presentation of our results. This work made use of theNASA/IPAC Extragalactic Database (NED) and the Two MicronAll Sky Survey (2MASS) database. Support for this work wasprovided by the National Aeronautics and Space Administrationthrough Chandra Postdoctoral Fellowship Award Number PF7-80050 and Chandra Award Number GO5-6127X and GO6-7128Xissued by the Chandra X-ray Observatory Center, which is operatedby the Smithsonian Astrophysical Observatory for and on behalfof the National Aeronautics and Space Administration under con-tract NAS8-03060. LVM is partially supported by DGI Grant AYA2005-07516-C02-01 and Junta de Andaluc´ıa (Spain).
REFERENCES
Abadi M. G., Moore B., Bower R. G., 1999, MNRAS, 308, 947Arnaud M., et al., 2002, A&A, 390, 27Broeils A. H., Rhee M.-H., 1997, A&A, 324, 877Bullock J. S., Kolatt T. S., Sigad Y., Somerville R. S., KravtsovA. V., Klypin A. A., Primack J. R., Dekel A., 2001, MNRAS,321, 559Cortese L., Gavazzi G., Boselli A., Franzetti P., Kennicutt R.C.,O’Neil K., Sakai S., 2006, A&A, 453, 847Coziol R., Plauchu-Frayn I., 2007, AJ, 133, 2630 c (cid:13) , 1–21 am pressure stripping in Hickson groups Dickey J.M., Lockman F.J., 1990, ARA&A, 28, 215Dos Santos S., Mamon G. A., 1999, A&A, 352, 1Fukazawa Y., Kawano N., Ohta A., Mizusawa H. 2002, PASJ, 54,527Gallagher S. C., Johnson K. E., Hornschemeier A. E., CharltonJ. C., Hibbard J. E., 2008, ApJ, 673, 730Giovanelli R., Haynes M.P., 1985, ApJ, 292, 404Gonzalez A. H., Zaritsky D., Zabludoff A. I., 2007, ApJ, 666, 14Grevesse N., Sauval A.J., 1998, Space Sci. Rev., 85, 161Gunn J.E., Gott J.R. III, 1972, ApJ, 176, 1Haynes M.P., Giovanelli R., 1984, AJ, 89, 758Helsdon S.F., Ponman T.J., 2003, MNRAS, 339, L29Hester J. A., 2006, ApJ, 647, 910Hickson P., 1982, ApJ, 255, 382J´achym P., Palouˇs J., K¨oppen J., Combes F., 2007, A&A, 472, 5Kawata D., Mulchaey J.S., ApJL, 672, L103Kronberger T., Kapferer W., Ferrari C., Unterguggenberger S.,Schindler S., 2008, A&A, in press (arXiv:0801.3759)Lewis I., et al., 2002, MNRAS, 334, 673Li C., Kauffmann G., Heckman T., Jing Y. P., White S. D. M.,2008, MNRAS in press (arXiv:0711.3792)Machacek M. E., Nulsen P., Stirbat L., Jones C., Forman W. R.,2005, ApJ, 630, 280Limousin M., Kneib J.P., Bardeau S., Natarajan P., Czoske O.,Smail I., Ebeling, H., Smith G.P., 2007, A&A, 461, 881Mahdavi A., Finoguenov A., B¨ohringer H., Geller M.J., HenryJ.P., 2005, ApJ, 622, 187Mannucci F., Della Valle M., Panagia N., Cappellaro E., CresciG., Maiolino R., Petrosian A., Turatto M., 2005, A&A, 433, 807Martinez M. A., Del Olmo A., Perea J., Coziol R., 2007, in Sa-viane I., Ivanov V.D., Borissova J., eds, Groups of Galaxies inthe Nearby Universe, Springer-Verlag, Berlin, p. 163Mori M., Burkert A., 2000, ApJ, 538, 559Mulchaey J.S., Davis D.S., Mushotzky R.F., Burstein D., 2003,ApJS, 145, 39Navarro J. F., Frenk C. S., White, S. D. M., 1997, ApJ, 490, 493Nulsen P.E.J., 1982, MNRAS, 198, 1007Osmond J. P. F., Ponman T. J., 2004, MNRAS, 350, 1511O’Sullivan E., Ponman T. J., Collins R. S., 2003, MNRAS, 340,1375Ponman T.J., Bourner P.D.J., Ebeling H., B¨ohringer H., 1996,MNRAS, 283, 690 (P96)Quilis V., Moore B., Bower R., 2000, Science, 288, 1617Rasmussen J., Ponman T. J., 2007, MNRAS, 380, 1554Rasmussen J., Ponman T.J., Mulchaey J.S., 2006, MNRAS, 370,453Rasmussen J., Stevens I.R., Ponman T.J., 2004, MNRAS, 354,259Read A.M., Ponman T.J., 2001, MNRAS, 328, 127Read A.M., Ponman T.J., 2003, A&A, 409, 395Roediger E., Br¨uggen M., 2007, MNRAS, 380, 1399Sengupta C., Balasubramanyam R., 2006, MNRAS, 369, 360Solanes J.M., Manrique A., Garc´ıa-G´omez C., Gonz´alez-CasadoG., Giovanelli R., Haynes M.P., 2001, ApJ, 548, 97Spitzer L., 1962, Physics of Fully Ionized Gases, Interscience,New York, NYSun M., Vikhlinin A., 2005, ApJ, 621, 718Sutherland R.S., Dopita M.A., 1993, ApJS, 88, 253Swaters R. A., van Albada T. S., van der Hulst J. M., Sancisi R.,2002, A&A, 390, 829Toft S., Rasmussen J., Sommer-Larsen J., Pedersen K., 2002, MN-RAS, 335, 799 Trinchieri G., Sulentic J., Breitschwerdt D., Pietsch W., 2003,A&A, 401, 173Verdes-Montenegro L., Yun M.S., Perea J., del Olmo A., HoP.T.P., 1998, ApJ, 497, 89Verdes-Montenegro L., Yun M.S., Williams B.A., HuchtmeierW.K., Del Olmo A., Perea J., 2001, A&A, 377, 812Vikhlinin A., Markevitch M., Murray S. S., Jones C., Forman W.,Van Speybroeck L., 2005, ApJ, 628, 655Wang Q. D., Owen F., Ledlow M., 2004, ApJ, 611, 821Williams B.A., Rood H.J., 1987, ApJS, 63, 265Yang X., Mo H. J., van den Bosch F. C., Pasquali A., Li C., BardenM., 2007, ApJ, 671, 153This paper has been typeset from a TEX/ L A TEX file prepared by theauthor. c (cid:13)000