THE ASTROPHYSICAL JOURNAL, 523 :575È584, 1999 October 11999. The American Astronomical Society. All rights reserved. Printed in U.S.A.(

VERY EXTENDED X-RAY AND Ha EMISSION IN M82: IMPLICATIONSFOR THE SUPERWIND PHENOMENON

MATTHEW D. LEHNERT1Max-Plank-Institut extraterrestrische Physik, Postfach 1603, D-85740 Garching, Germanyfu� r

TIMOTHY M. HECKMAN

Department of Physics and Astronomy, Johns Hopkins University, Baltimore, MD 21218

AND

KIMBERLY A. WEAVER

NASA Goddard Space Flight Center, Greenbelt, MD 20771 ; and Department of Physics and Astronomy, Johns Hopkins University, Baltimore, MD 21218Received 1999 January 29 ; accepted 1999 May 11

ABSTRACTWe discuss the properties and implications of the recently discovered (3.7] 0.9 kpc) region3@.5] 0@.9

of spatially coincident X-ray and Ha emission about 11@ (11.6 kpc) to the north of the prototypicalstarburst/superwind galaxy M82. The total Ha Ñux from this ridge of emission is 1.5 ] 10~13 ergs s~1cm ~2, or about 0.3% of the total M82 Ha Ñux. The implied Ha luminosity of this region is 2.4 ] 1038ergs s~1. Di†use soft X-ray emission is seen over the same region by the ROSAT PSPC and HRI. ThePSPC X-ray spectrum is Ðtted by thermal plasma absorbed by only the Galactic foreground columndensity cm~2) and having a temperature of kT \ 0.80^ 0.17 keV. The total unab-(NH \ 3.7] 1020sorbed Ñux from the ridge is 1.4 ] 10~13 ergs cm~2 s~1 (D2.2] 1038 ergs s~1), comprising about 0.7%of the total X-ray emission from M82. We evaluate the relationship of the X-ray/Ha ridge to the M82superwind. The Ha emission could be excited by ionizing radiation from the starburst that propagatesinto the galactic halo along the cavity carved by the superwind. However, the main properties of theX-ray emission can all be explained as being due to shock heating driven as the superwind encounters amassive ionized cloud in the halo of M82 (possibly related to the tidal debris seen in H I in the inter-acting M81/M82/NGC 3077 system). This encounter drives a slow shock into the cloud, which contrib-utes to the excitation of the observed Ha emission. At the same time, a fast bow shock develops in thesuperwind just upstream of the cloud, and this produces the observed X-ray emission. This interpretationwould imply that the superwind has an outÑow speed of roughly 800 km s~1, consistent with indirectestimates based on its general X-ray properties and the kinematics of the inner kiloparsec-scale region ofHa Ðlaments. The alternative, in which a much faster and more tenuous wind drives a fast radiativeshock into a cloud that then produces both the X-ray and Ha emission, is ruled out by the long radiativecooling times and the relatively quiescent Ha kinematics in this region. We suggest that wind-cloudinteractions may be an important mechanism for generating X-ray and optical line emission in the halosof starbursts. Such interactions can establish that the wind has propagated out to substantially greaterradii than could otherwise be surmised. This has potentially interesting implications for the fate of theoutÑowing metal-enriched material, and bears on the role of superwinds in the metal enrichment andheating of galactic halos and the intergalactic medium. In particular, the gas in the M82 ridge is roughly2 orders of magnitude hotter than the minimum ““ escape temperature ÏÏ at this radius, so this gas will notbe retained by M82.Subject headings : galaxies : active È galaxies : individual (M82) È galaxies : ISM È galaxies : nuclei È

galaxies : starburst

1. INTRODUCTION

The history of the study of M82 is a long and rich one.M82 is the prototypical starburst galaxy, and since it is oneof the nearest (D\ 3.63 Mpc ; Freedman et al. 1994) andbrightest members of that class, it has been studied atalmost every wavelength ranging from c-rays through low-frequency radio waves (e.g., Bhattacharya et al. 1994 ;Bregman, Schulman, & Tomisaka 1995 ; Courvoisier et al.1990 ; Lynds & Sandage 1963 ; Shopbell & Bland-Hawthorn1998 ; Larkin et al. 1994 ; Petuchowski et al. 1994 ; Seaquist,Kerton, & Bell 1994 ; Hughes, Gear, & Robson 1994 ;

1 Visiting observer at Kitt Peak National Observatory of the NationalOptical Astronomy Observatories, operated by AURA under contractwith the National Science Foundation.

Reuter et al. 1994 ; Moran & Lehnert 1997 ; Ptak et al. 1997 ;Strickland, Ponman, & Stevens 1997). It is perhaps impossi-ble for any paper, let alone one this brief, to do justice tothis rich history. We refer the reader to excellent overviewson M82 in, for example, Bland & Tully 1988 and Telesco(1988).

M82 is also the prototype of the ““ superwind ÏÏ phenome-non (Axon & Taylor 1978 ; McCarthy, Heckman, & vanBreugel 1987 ; Bland & Tully 1988 ; Heckman, Armus, &Miley 1990 ; Bregman et al. 1995 ; Strickland et al. 1997 ;Shopbell & Bland-Hawthorn 1998). These Ñows arepowered by the collective kinetic energy and momentuminjected by stellar winds and supernovae in starburst gal-axies. While recent studies have established the ubiquity ofsuperwinds in local starburst galaxies (e.g., Lehnert &Heckman 1996 ; Dahlem, Weaver, & Heckman 1998, here-

575

576 LEHNERT, HECKMAN, & WEAVER Vol. 523

after DWH), there is still only a rough understanding ofboth their dynamics and the processes responsible for theobserved optical and X-ray emission (e.g., Strickland 1998and references therein). Without such understanding, it isdifficult to quantitatively assess the role that superwindshave played in the formation and evolution of galaxies(Kau†mann & Charlot 1998 ; Somerville, Primack, & Faber1998) and in the chemical enrichment, heating, and possiblemagnetization of the intergalactic medium (Gibson, Loe-wenstein, & Mushotzky 1997 ; Ponman, Cannon, &Navarro 1999 ; Kronberg, Lesch, & Hopp 1999).

In this paper we discuss the optical and X-ray propertiesof the spatially correlated region of X-ray and Ha line emis-sion recently reported by Devine & Bally (1999, hereafterDB) to be located roughly 11 kpc above the disk of M82, farbeyond the region in which the superwind has been pre-viously investigated. We are motivated by the importantimplications for the superwind theory of understandingsuch a region. If this region could be shown to be physicallyrelated to the superwind that M82 is driving, then it impliesthat winds driven even by energetically modest starburstssuch as M82 are able to reach large distances into the halo,making them more likely to be able to escape the gravita-tional potential of the galaxy and thereby supply mass,metal, and energy to the intergalactic medium (IGM). Sucha Ðnding would have implications for our understanding ofthe evolution of the IGM, the halos of galaxies, and quasarabsorption-line systems. Our study of this emission regionalso provides some unique clues to the physics that under-lies the observed X-ray and optical emission from super-winds, and strongly suggests that wind-cloud collisions instarburst galaxy halos are an important emission process.

2. OBSERVATIONS AND DATA REDUCTION

2.1. X-Ray DataThe ROSAT PSPC data were obtained during the period

of 1991 March through October and had a total integrationtime of about 25 ks. M82 was placed approximately in thecenter of the Ðeld for all the observations, and the sourcesize is small enough so that there is no obscuration of theX-ray emission from M82 due to the mirror ribs.

We extracted a spectrum from a region of di†use X-rayemission located about 11@ north of M82, and coincidentwith the region of Ha emission reported by DB. We willsubsequently refer to this region as the ““ ridge ÏÏ of emission.This area of X-ray and optical emission is oriented roughlyparallel to the major axis of M82, and was denoted assources 8 and 9 in DWH. We extracted source counts in anelliptical region with major and minor axes of and3@.3 1@.8,respectively, oriented in a position angle of 43¡. This regionis just sufficient to encompass all of the obvious X-ray emis-sion, but excludes the soft point source at the east-northeastend of the ridge of emission that is discussed in ° 3.2. Wehave also extracted source counts for the portion of theridge that is contaminated by this point source using acircular aperture with a diameter of The background1@.75.for these spectra was taken from a large region in whichthere were no apparent background sources to the southand west of M82. The relative distance of the backgroundregion from the center of the PSPC Ðeld was approximatelythe same as that for the ridge of emission. The spectrum wasthen grouped such that each channel contained at least 25counts. The extracted background-subtracted data were

subsequently Ðtted with various models using the packageXSPEC.

We have also inspected the archival ROSAT HRI datafor M82. These data have lower sensitivity than the PSPCdata, and provide no spectral information. However, theydo provide signiÐcantly higher spatial resolution (D6AFWHM on-axis and D15A at the position of the X-rayridge). The data we have used comprise a 53 ks exposuretaken in 1995 April and May. We have used these data onlyto study the morphology of the X-ray ridge.

2.2. Optical ImagingOur optical images of M82 were taken on the night of UT

1995 February 27, using the KPNO 0.9 m telescope. Weobserved with a Tek 20482 CCD and through the optics of0.9 m, yielding a scale of pixel~1 and a total Ðeld of0A.68view of over 23@. The night was not photometric, with cirrusaround all night and intermittent patches of thicker clouds.M82 was observed through two Ðlters, a narrowband Ðlterwith a FWHM of 29 and a central wavelength of 6562A� A� ,and a wider Ðlter with a FWHM of 85 and a centralA�wavelength of 6658 The Ðrst Ðlter e†ectively isolated HaA� .and the underlying continuum and did not include emissionfrom either [N II] lines at 6548 and 6584 The second ÐlterA� .provided a line-free measure of the continuum and was usedfor continuum subtraction of the Ðrst narrowband Ðlter toproduce a Ha lineÈonly image of M82. The total integrationtime in each Ðlter was 1200 s.

The images were reduced in the usual way (bias-sub-tracted and Ñat-Ðelded, and then the narrowband contin-uum image was aligned and scaled to subtract thecontinuum from the narrowband Ha image) using thereduction package IRAF.2

Since the data were taken under nonphotometric condi-tions, Ñuxing the data directly using observations ofspectrophotometric standards is impossible. However,obtaining Ñuxes and hence luminosities is very important toour analysis. Therefore, to estimate the Ha Ñux and contin-uum Ñux density of the o†-band Ðlter of the continuum andemission-line nebulae of M82, we bootstrapped from theÑux given in a circular aperture centered on knot C5A.8from OÏConnell & Mangano (1978). Knot C was chosenbecause it was easily identiÐable in our images and appearsin a relatively uncomplicated region of line emission. Thisprocedure yields a Ha Ñux from M82 of about 4.6] 10~11ergs cm~2 s~1, which compares very well with the Ha Ñuxfrom M82 estimated by McCarthy et al. (1987) of4.5] 10~11 ergs cm~2 s~1 over a similar region 90@@] 90@@.All subsequent results for the Ha Ñux and luminosity ofvarious features within the Ha image of M82 will rely onthis bootstrapping procedure.

To estimate the continuum Ñux density, we used the Ñuxdensity of OÏConnell & Mangano (1978) taken at 6560 A� .Since the center of our narrowband continuum emission is6658 we are thus assuming that the continuum has aA� ,constant Ñux density between 6560 and 6658 To see ifA� .this extrapolation procedure is accurate, we compared theÑux in R of M82 in a 35A aperture of 1.97 ] 10~13 ergscm~2 s~1 (from NED, based on data originallyA� ~1

2 IRAF is distributed by the National Optical Astronomy Observa-tories, which are operated by the Association of Universities for Researchin Astronomy, Inc., under cooperative agreement with the NationalScience Foundation.

DE

CL

INA

TIO

N

RIGHT ASCENSION09 57 00 56 30 00 55 30 00 54 30

69 52

50

48

46

44

42

40

38

36

No. 2, 1999 VERY EXTENDED EMISSION IN M82 577

obtained by Johnson 1966). If we also take the total countsin the 6658 Ðlter and determine the Ñux density byA�extrapolating the conversion factor from measuring knot Cof OÏConnell & Mangano (1978), we Ðnd a Ñux density of2.11] 10~13 ergs cm~2 s~1 in a 35A aperture. OurA� ~1extrapolated Ñux is within 10% (0.1 mag), and thus thisextrapolation is reasonably accurate.

3. RESULTS

3.1. Optical ImagingThe Ha and narrowband continuum images of M82 are

quite spectacular (Figs. 1 and 2). We Ðnd that the line emis-sion is very extended, with discernible Ha line emissionextending to about 11 kpc above the plane of M82. Thisvery faint emission can be seen even more clearly in thedeeper Ha image obtained by DB. Like previous authors(e.g., Bland & Tully 1988 ; McCarthy et al. 1987), we Ðnd

that the Ha line emission has a very complex morphologythat is highly suggestive of outÑowing gas, with loops, ten-drils, and Ðlaments of line emission pointing out of theplane of the galaxy.

An especially interesting feature reported by DB is thefaint ridge of emission about 11@ north of M82 at a positionangle of about [20¡ from the nucleus. This ridge is thenabout 11.6 kpc (in projection) north of M82 at the adopteddistance of 3.63 Mpc (Freedman et al. 1994). The dimen-sions of the ridge down to surface brightnesses of about3.5] 10~17 ergs s~1 cm ~2 arcsec~2 are about long and3@.5

wide, with the long axis being approximately parallel to0@.9the major axis of M82. At the assumed distance of M82, thissize translates into approximately 3.7 kpc ] 0.9 kpc, andthe total Ha Ñux from this ridge of emission is 1.5 ] 10~13ergs s~1 cm ~2. This Ñux is reasonably close to the estimatein DB of 1.2] 10~13 ergs s~1 cm~2 and yields an Ha lumi-nosity of 2.4 ] 1038 ergs s~1. The total Ha Ñux for the

FIG. 1.ÈGray scale showing the Ha image and contours showing the ROSAT PSPC image of M82. The lowest surface brightness emission visible in theHa corresponds to a Ñux of 3.5 ] 10~17 ergs s~1 cm ~2 arcsec~2. The spatial coincidence of the Ha and X-ray emission is quite good. The images werealigned using the alignment determined for the continuum image. Coordinates are shown for J2000.

578 LEHNERT, HECKMAN, & WEAVER Vol. 523

FIG. 2.ÈGray scale showing a narrowband continuum image (central wavelength approximate 6658 and contours showing the smoothed (p \ 10@@A� )Gaussian) ROSAT PSPC image of M82. The images were aligned using what is likely to be a background active galactic nucleus and galaxy that were bothvisible in the optical continuum and X-ray images. Coordinates are shown for J2000.

whole of M82 is approximately 4.6] 10~11 ergs s~1 cm ~2.Thus, the ridge of emission constitutes approximately 0.3%of the total Ha emission from M82 (not corrected for inter-nal extinction). While the ridge appears rather structurelessin our Ha image, the deeper image published by DB showsa collection of bright knots and loops or Ðlaments con-nected to and immersed in the more di†use and pervasiveemission.

The narrowband line-free continuum image reveals acomplex morphology that is as interesting as the Ha image.In addition to the obvious complex dust absorption that isseen in the image (with some dark features oriented perpen-dicular to the plane of the galaxy), we also see Ðlaments ofcontinuum emission extending out of the plane of thegalaxy (although the Ðlaments are not entirely obvious inFig. 2). These Ðlaments have twists and turns as seen pro-jected on the sky, but generally point perpendicular to theplane of the galaxy. They extend outward almost 3@ to thenorthwest and almost to the southeast, corresponding to2@.5a projected distance out of the plane of about 3 kpc. Eventhough the continuum emission is in many ways as compli-cated as that seen in the Ha image, we do not observe anycontinuum emission from the ridge of spatially coincidentHaÈX-ray emission. Using the counts-toÈÑux density con-version for the narrowband line-free continuum image andthe same aperture as used to measure the Ha Ñux (but withthe point sources removed from the area), we Ðnd an upperlimit for the possible continuum emission from the ridge offainter than 15.8 R mag. This then implies an absolute Rmagnitude fainter than [12.0.

While we Ðnd no evidence for any di†use, spatiallyextended optical continuum associated with the Ha ridge,we do Ðnd a relatively bright point source in the line-freeimage that is coincident with the X-ray point source seen atthe east-northeast end of the ridge in the ROSAT HRI data.

This point source has an R magnitude of 12.2. We willdiscuss its likely nature in the next section.

3.2. X-Ray Imaging and SpectroscopyThe optical images described above were aligned with the

X-ray images, starting with the coordinates given for theX-ray data from ROSAT and using the positions of starswithin the frame from the Guide Star Catalog and coordi-nates given in the header from the digital sky survey. Wethen slightly adjusted the X-ray pointing by identifying andaligning X-ray sources on the digital sky survey image ofthe region. The close spatial correspondence between theridge of optical and X-ray emission about 11@ to the north ofM82 is obvious (Fig. 1).

We do not display the ROSAT HRI image of the ridge,since these data show only that the ridge is largely di†useand structureless with the 15A (260 pc) o†-axis spatialresolution of the HRI. The only substructure is a pointsource at the easternmost end of the ridge. This sourceappears to coincide with a stellar object visible in Figure 2(the relatively bright starlike object located about 40A east-northeast of knot I in Fig. 2 of DB). This is most likely aforeground star, as we will brieÑy discuss below.

Examination of the 0.25, 0.75, and 1.5 keV PSPC maps ofM82 presented by DWH) shows that the X-ray morphologyof the ridge is energy dependent. The east-northeast portionof the ridge (source 9 in DWH) is relatively much moreconspicuous in the 0.25 keV map compared to the rest ofthe ridge (source 8 in DWH). This reÑects the contami-nation of the easternmost ridge emission by the(presumably) unrelated point source visible in the HRIimage.

As described in ° 2.1 above, we have extracted a PSPCspectrum of the main body of the ridge from a region thatexcluded the point source. The relatively small number of

No. 2, 1999 VERY EXTENDED EMISSION IN M82 579

detected photons precludes Ðtting complex multi-component models, and we Ðt a model of a single thermal(““MEKAL ÏÏ) component to the spectrum. We are moti-vated to use a thermal model by the detailed X-ray spec-troscopy of the brighter portions of the M82 outÑow(DWH; Moran & Lehnert 1997 ; Strickland et al. 1997). TheÐts strongly preferred a low-absorbing column, so wesimply froze the column at the Galactic value of NH \ 3.7] 1020 cm~2. The data quality were not adequate to Ðt forthe metal abundance, so we froze this at the solar value.Changing the assumed metal abundance did not signiÐ-cantly a†ect the best-Ðt temperature, but does directly a†ectthe normalization of the Ðt. The best-Ðt temperature waskT \ 0.76^ 0.13 keV (90% conÐdence interval with areduced s2B 1.3 ; Fig. 3). The relatively high temperature issimilar to that measured in the ROSAT PSPC spectroscopyof the innermost region of the M82 outÑow (D0.7 keV;DWH; Strickland et al. 1997). The total unabsorbed Ñux inthe 0.1È2.4 keV band from this part of the ridge of emissionis 8.5] 10~14 ergs cm~2 s~1.

For the east-northeast part of the ridge, the emission ofwhich is contaminated by the point source described above,we have Ðtted a two-component MEKAL model (Fig. 3).Both components were frozen at solar abundances ; onecomponent (the ridge emission) had an absorbing columnfrozen at cm~2, and the other (the stellarNH \ 3.7] 1020point source) had an absorbing column frozen at 0. This Ðtyields kT \ 0.85^ 0.17 keV for the ridge and kT D 0.1 keVfor the point source. The corresponding unabsorbed Ñuxesare 5.8 ] 10~14 ergs cm~2 s~1 and 3.5 ] 10~15 ergs cm~2s~1 for the ridge and point source, respectively. Combiningthe Ðts for the two regions of the ridge, the total X-ray Ñuxis then 1.4 ] 10~13 ergs cm~2 s~1. This comprises about0.7% of the total unabsorbed X-ray emission from M82 inthe PSPC band (e.g., Moran & Lehnert 1997 ; DWH). Thetotal X-ray luminosity of the ridge of emission is 2.2 ] 1038ergs s~1 (in the 0.1È2.4 keV band).

As stated above, the east-northeast point source is identi-Ðed with a stellar object having an R magnitude of 12.2

ergs cm~2 s~1). The X-ray/optical Ñux(jFj\ 1.6] 10~10ratio suggests that the most likely nature of this source is a

FIG. 3.ÈROSAT PSPC spectrum of the di†use X-ray emission fromthe ridge to the right of the point source (circles) and superimposed on thepoint source (crosses). Data below 0.6 keV are ignored in the latter case toeliminate the point source. Also shown are the best-Ðt MEKAL plasmamodels (see text) folded through the instrumental response.

late-type dwarf star, and this would also be consistent withits soft ROSAT PSPC spectrum (e.g., Rosner, Golub, &Vaiana 1985).

4. DISCUSSION

4.1. Is the Ridge of Emission Associated with M82?Being able to isolate the ridge of emission to the north of

M82 allows us to investigate its nature separately from therest of M82. Could this ridge of emission be due to anothergalaxy near M82, or is it instead gas excited by the M82starburst ?

We have found that the ridge of X-ray and Ha emissionhas fairly high X-ray and Ha luminositiesÈboth lumi-nosities are in the range observed for star-forming dwarfgalaxies (e.g., Hunter, Hawley, & Gallagher 1993 ; Heckmanet al. 1995). However, we see no evidence for any opticalcontinuum emission from this ridge. The limit for the levelof continuum emission implies that any putative galaxymust be of very low luminosity, comparable to that of thelocal dwarf spheroidal galaxies (Mateo 1998). Moreover,the limit on the continuum emission implies an Ha equiva-lent width of greater than 180 Examining the sample ofA� .galaxies with fainter than [15 in the Ha survey ofM

Bnearby dwarf irregular galaxies by Hunter et al. (1993), weÐnd that these have typical global Ha equivalent widths of afew to a few tens of The only such galaxy with an HaA� .equivalent width comparable to our lower limit for the M82ridge is DDO 53. This galaxy has a ratio of H I mass to Haluminosity (in solar units) of 420. The corresponding H I

mass for the M82 ridge would have to be 2.5 ] 107 M_

.The H I map in Yun et al. (1993) implies a maximum H I

mass in the M82 ridge that is only about 106 More-M_

.over, while the M82 ridge has an X-ray luminosity similarto that of the dwarf starburst galaxy NGC 1569 (Heckmanet al. 1995), the M82 ridge would have a ratio of X-ray tooptical continuum luminosity that would be over 40 timeshigher than NGC 1569. Thus, it is clear that the ridge inM82 has properties that are very di†erent from those ofstar-forming dwarf galaxies. This argument, together withthe strong morphological connection between the M82ridge and the M82 outÑow (see DB), leads us to concludethat this ridge is not a dwarf star-forming galaxy.

Other possibilities for this region, such as a cloud in thehalo of the Milky Way, are also highly unlikely. The H I

detected in the data of Yun et al. (1993) and the Ha emissionfrom the spectrum of DB have velocities such that they areboth likely to be associated with M82 and not the Galaxy(Burton & Hartmann 1997 ; Ha†ner et al. 1998 ; Reynolds etal. 1995). Moreover, while Galactic high-velocity H I cloudsoften show soft X-ray excesses (Kerp et al. 1996, 1999), theassociated X-ray emission is very soft (B0.1È0.2 keV; seediscussion in Kerp et al. 1996).

4.2. Photoionization of the Ridge by M82The ridge is evidently related to, and excited by, M82. In

the case of the Ha emission, two possibilities are obvious.Either the superwind in M82 has worked its way out toB12 kpc above the plane of the galaxy and is shocking halomaterial that appears in plentiful supply around M82 (Yunet al. 1993, 1994), or Lyman continuum radiation from thestarburst is ionizing gas at very large distances from thedisk. Only the Ðrst possibility could apply to the X-rayemission from the ridge.

580 LEHNERT, HECKMAN, & WEAVER Vol. 523

Let us suppose that ambient gas is being photoionized bythe stellar population of M82. The rate of ionizing photonsfrom M82 is approximately 1054 s~1 (McLeod et al. 1993and references therein). Using the projected size and dis-tance from M82 estimated for the ridge of emission given inthe previous section, and assuming that the ridge is really a““ box ÏÏ with a depth equal to its projected length (i.e., 3.7kpc] 3.7 kpc), we Ðnd that the ridge will intercept about8 ] 1051 ionizing photons s~1. This of course is an upperlimit, since some (most?) of the ionizing photons will beabsorbed or dust-scattered before they reach distances asgreat as 11.6 kpc out of the plane of M82. We can estimatethe rate of photons necessary to give the Ha luminosity ofthe ridge (2.4 ] 1038 ergs s~1) by assuming that the gas is inphotoionization equilibrium, i.e., the number of recombi-nations equals the number of absorptions. Using thenumbers for the case B e†ective recombination rate of Hafrom Osterbrock (1989), we Ðnd that it is necessary to havea rate of D 2 ] 1050 ionizing photons s~1. This is D40times smaller than the upper limit for the number ofphotons that M82 is able to provide, so only about 2% to3% of the ionizing radiation initially emitted in the direc-tion of the ridge needs to make it to this distance. Such an““ escape fraction ÏÏ is consistent with Hopkins UltravioletTelescope (HUT) spectra of starbursts below the Lymanedge (Leitherer et al. 1995 ; Hurwitz, Jelinsky, & Dixon1997), and it seems plausible that the optimal path alongwhich ionizing radiation could escape from a starburst isalong the cavity carved out by the superwind.

DB have made similar arguments in favor of a starburstphotoionization origin for the Ha emission. However,photoionization by stars does not produce the hot X-rayemission that is coincident with the Ha emission. Instead,shock heating by the superwind may be an efficient mecha-nism for producing coincident X-ray and line-emittingplasma. We now explore the energetic plausibility of thismodel, and revisit the issue of the role of photoionization in° 4.5.

4.3. Shock Energetics and the Basic Physical Properties ofthe Ridge

The ridge of emission has an X-ray luminosity of2.2] 1038 ergs s~1 and a temperature of 9 ] 106 K.Assuming that the shock is strong and adiabatic, we have

where k is the mass per particle,Tpostshock\ (3/16)[(kvs2)/k],

is the shock velocity, and k is BoltzmannÏs constant.vsThus, to produce gas with T \ 9 ] 106 K requires a shock

speed of 820 km s~1. Such a high shock speed is quitereasonable considering that M82 has been estimated tohave a deprojected outÑow velocity in optical line emittingÐlaments observed along the minor axis of roughly 660 kms~1 (Shopbell & Bland-Hawthorn 1998), and the outÑowingX-ray gas could in principle have a velocity as high as 3000km s~1 (Chevalier & Clegg 1985). It would take materialtraveling at 820 km s~1, B1.4] 107 yr to travel the 11.6kpc from M82 to the ridge of emission. Since this is approx-imately the estimated lifetime of the starburst in M82 (Riekeet al. 1993 ; Bernloehr 1992 ; Schreiber 1998 ; Satyapal et al.1997), it is very plausible that the starburst could drive ashock over its lifetime this far out of the galactic plane.

If we assume that we are observing a ““ box ÏÏ of emissionwith dimensions of 3.7] 3.7] 0.9 kpc, the total volume is3.7] 1065 cm~3. Assuming that this region is occupied byhot gas Ðlling a fraction of this volume, the measuredfX

luminosity in the ROSAT bandpass of 0.1È2.4 keV impliesthat the X-rayÈemitting gas has an electron density of

cm~3, a pressurene,X\ 5.4] 10~3f X~1@2 PX B 2n

e,X kT \dynes cm~2, a cooling time1.3] 10~11f X~1@2 tcool Byr, mass3kT /n

e,X"\ 2.7] 108f X1@2 MX \ 2.0 ] 106f X1@2and a total thermal energy of ergs. Simi-M_

, 8 ] 1054f X1@2larly, assuming that conditions of case B recombination anda temperature of 104 K holds in the region producing theHa emission, and that this region also occupies a volume of

cm~3, then the implied density and mass of3.7] 1065fHathis gas are cm~3 andne,Ha \ 4.3] 10~2f Ha~1@2 MHa\ 8.0

respectively. If we make the physically rea-] 106f Ha1@2 M_

,sonable assumption that this cooler gas is in rough pressurebalance with the hot X-ray gas, then the minimum pressureof 1.3 ] 10~11 dyne cm~2 would require that the Ha-emitting gas is highly clumped This in( fHa D 8 ] 10~5).turn implies that cm~3 andn

e,Ha \ 5 MHa \ 1.1] 105 M_

.We note that a small volume-Ðlling factor for this gas isconsistent with the complex knotty, Ðlamentary structureseen in the deep Ha image in DB.

The total bolometric luminosity and ionizing luminosityof M82 (McLeod et al. 1993 and references therein) can beused together with the starburst models of Leitherer &Heckman (1995) to predict the rate at which the starburstinjects mechanical energy and momentum. While these pre-dictions will depend to some degree on the evolutionarystate of the starburst (cf. Figs. 66È69 in Leitherer &Heckman 1995), we estimate that the mechanical luminosityand momentum Ñux provided by M82 are D2.5] 1042 ergss~1 and D2 ] 1034 dyne, respectively. For this windmechanical luminosity, the rate at which the X-ray ridgeintercepts the energy is ergs s~1, where is2.5] 1041)

w~1 )

wthe total solid angle into which the wind Ñows ()w

¹ 4n).Comparing this heating rate to the X-ray luminosity of theridge (2.2 ] 1038 ergs s~1) shows that the wind could easilypower the emission. It could also supply the ridgeÏs esti-mated thermal energy in a timescale of 106 yr (muchf X1@2)wless than the wind/starburst lifetime).

Matching the observed high pressure in the X-ray ridge ismore challenging. A wind momentum Ñux of 2.5] 1034dyne leads to a wind ram pressure at a radius of 11.6 kpc of2 ] 10~11 dyne cm~2. Consistency with the X-ray)

w~1

pressure estimated above would require that the X-ray gashas a large volume-Ðlling factor and that the( fX D unity)M82 wind is rather well collimated sr). This latter()

wD 1.5

constraint is consistent with the morphology of the outÑow(e.g., Goetz et al. 1990 ; McKeith et al. 1995 ; Shopbell &Bland-Hawthorn 1998). Alternatively, we can use theempirical measurements of the gas pressure in the innerregion of the M82 nebula (Heckman et al. 1990) andextrapolate these outward, assuming that the wind rampressure falls as r~2. The pressure at a radius of 1 kpc wasmeasured to be D5 ] 10~10 dyne cm~2, and so the predict-ed wind ram pressure at a radius of 11.6 kpc would beD4 ] 10~12 dyne cm~2, about a factor of 3 smaller thanthe estimated value (for In view of the uncertaintiesfX \ 1).in the estimated physical parameters for the X-ray ridge, weregard this level of disagreement as acceptable.

If the optical line emission is also shock-excited by theoutÑowing wind, then the luminosity observed at Ha shouldbe consistent with the wind energetics. We can estimate theamount of emission from a shock as follows. The totalenergy dissipated in a shock is L shock \ 12oAv

s3\ 1.0 ] 1034

ergs s~1, where is the shock surface area inn0Akpc vs3 Akpc

No. 2, 1999 VERY EXTENDED EMISSION IN M82 581

kpc2, is the preshock density (cm~3), and is in units ofn0 vskm s~1. From Binette, Dopita, & Tuohy (1985), using the

equation above, we have L (Ha)shockB 3L (Hb)shock Bergs s~1, if 90 \3L shock/100 \ 1.7] 1041 n0Akpc(vs/820)3

km s~1. Adopting the geometry above, the totalvs\ 1000

surface area that the ridge would present to the wind wouldbe 3.7 kpc] 3.7 kpc. The observed Ha luminosity of theridge (2.4 ] 1038 ergs s~1) then requires that n0(vs/820)3 \1.0] 10~4, while the pressure estimated in the X-ray ridge(1.3] 10~11 dynes cm~2) requires n0(vs/820)2 \ 8 ] 10~4.Thus, provided that the shock speed is greater than 100 kms~2, it can both produce the observed Ha emission-lineluminosity and be consistent with the pressure estimated forthe hot shocked gas observed in the X-rays.

4.4. T he Relationship Between the Ha and X-Ray EmissionWhat is the physical relation between the Ha and X-ray

emission in this type of model, in which the superwinddrives a shock into an ambient cloud in the halo of M82?One possibility is that the shock driven into the cloud bythe wind is fast enough (i.e., D800 km s~1) to produce theX-ray emission, and that the shocked gas then cools radi-atively down to temperatures of the order of 104 K andproduces the observed Ha emission via recombination. Thedifficulty with this simple model is that the estimated radi-ative cooling time in the X-ray gas (D3 ] 108 yr) is verylong compared to the characteristic dynamical time for theX-ray ridge, kpc/820 km s~1)D 1 ] 106 yr.tdynD (0.9Simply put, this model would require the X-ray ridge to beD300 kpc thick, and the Ha emission would be at the backedge of this very thick cooling zone. An additional per-suasive argument against this model is provided by therather quiescent kinematics of the Ha-emitting materialreported by DB. Their optical spectra show that the Haemission in the ridge spans a velocity range of only about102 km s~1, which is inconsistent with material coolingbehind a shock with a velocity nearly an order of magnitudelarger. While the relatively small di†erence DB reportbetween the mean velocity of the ridge and M82 proper(blueshift of 50È200 km s~1) could perhaps be explained asa projection e†ect in an outÑow seen nearly in the plane ofthe sky, this e†ect could not explain the narrowness of theHa emission line in the ridge (D102 km s~1) withoutappealing to an unreasonably idealized shock geometry(which is bellied by the morphological complexity of the Haemission in the image published by DB).

A much more plausible alternative is that we are actuallyobserving two physically related shocks in the M82 ridge.We propose that as the superwind encounters a large cloudin the halo of M82, it drives a relatively slow (of order 102km s~1) radiative shock into the cloud. At the same time,this encounter also results in a fast (820 km s~1) stand-o†bow shock in the superwind Ñuid upstream from the cloud.The slow cloud shock (aided and abetted by ionizing radi-ation from M82) produces the Ha emission. This is consis-tent with the quiescent Ha kinematics. The fast bow shockin the wind produces the X-ray emission and is predicted toonly be o†set from front edge of the cloud by a fraction ofthe size of the cloud (Klein, McKee, & Colella 1994 ;Bedogni & Woodward 1990). Pressure balance between thetwo shocks requires that whereowind vbow2 \ ocloud vshock,cloud2 ,

and are the preshock densities in the wind andowind ocloudcloud, respectively. The observed X-ray pressure and tem-perature imply that km s~1 andvbowD 820 nwind,ionD 1

] 10~3 cm~3. Thus, for example, if the shock speed in thecloud is km s~1, then cm~3. Wev

s,cloud \ 100 ncloudD 0.07note further that the velocity of the bow shock is just thedi†erence between the outÑow velocity of the superwinditself and that of the shocked cloud. Assuming that lattervelocity is much smaller than the former (as suggested bythe kinematic information provided by DB), the wind veloc-ity implied by the X-ray temperature is D820 km s~1. Todrive a wind with a terminal velocity of 820 km s~1, hot gaswould need to be injected into the starburst at a meanmass-weighted temperature of 1 ] 107 K (Chevalier &Clegg 1985). This temperature is consistent with X-ray spec-troscopy of the hot gas in the core of M82 (DWH; Moran &Lehnert 1997 ; Strickland et al. 1997).

We have seen above that the radiative cooling time in theX-ray ridge is D300 times longer than the characteristicdynamical time for the ridge. This implies that the windbow shock is adiabatic. This can also be seen by comparingthe total rate of energy dissipation in the shock (L shock\

ergs s~1 for the parameters12owindAridge vbow3 \ 7 ] 1040estimated above) to the observed X-ray luminosity of theridge (2.2] 1038 ergs s~1). Thus, the thickness of the X-rayridge in the direction of the outÑow is set by the dimensionsof the region in which adiabatic expansion and coolingcauses the X-ray surface brightness to drop precipitously.This region will be of the order the size of the obstacle in theÑow that has caused the bow shock to develop (the cloud).This is consistent with the observed X-ray morphology andthe good spatial correlation seen between the Ha and X-rayemission. We have also estimated previously that the super-wind can supply the ridge with its estimated thermal energycontent in a timescale of the order of 106 yr. This is roughlyequal to the dynamical time in the ridge (the timescale foradiabatic cooling), as required.

We conclude that it is energetically feasible for the cloudto be shock heated by a superwind driven by the starburstin M82. A wind/cloud shock interpretation leads to nouncomfortable estimates of the luminosity, covering frac-tion, or timescales. In addition, as noted previously, there isample evidence that M82 lies in an ““H I halo ÏÏ (Yun et al.1993, 1994). It is likely that the outÑowing gas has inter-cepted and shock heated some of this halo material, whichis observed to be clumpy over large scales (see Yun et al.1993). We also note that perhaps one might have expectedto Ðnd stronger evidence a priori for such interactions onthe north side of the plane of M82 compared to the southside ; Yun et al. (1994) Ðnd that the velocity gradient in theH I is much higher on the northern side (12 km s~1 kpc~1)of the plane of M82 compared to that on the southern side(3È6 km s~1 kpc~1). This higher gradient may be related toour observation of cospatial Ha and thermal X-ray emis-sion at large distances from the plane of M82.

4.5. W here is the Cloud?We have shown that a simple model of the interaction

between a fast (D800 km s~1) wind and a massive cloud canquantitatively explain the physical and dynamical proper-ties of the ridge. However, it a fair question to ask why theputative cloud is not observed in the sensitive H I mappublished by Yun et al. (1993). More speciÐcally, we haveestimated above that the high pressures in the X-ray gasand the low velocity of the shock driven into the cloud bythe wind together mean that the preshock density of thecloud must be of the order of 10~1 cm~3. For a cloud with a

582 LEHNERT, HECKMAN, & WEAVER Vol. 523

line-of-sight dimension of 3.7 kpc, the implied cloud columndensity is then D1021 cm~2. The H I map in Yun et al.(1993) constrains the H I column density of the ridge to be\2.7] 1019 cm~2. This implies that the cloud must nowbe fully ionized.

This could be accomplished by either the wind-drivenshock or the ionizing radiation from M82. The time for theshock to traverse the cloud will be D107 yr for vshock,cloud\100 km s~1, comparable to the estimated age of the star-burst and its superwind. Similarly, the velocity of a photo-ionization front moving though the cloud is just v

iB

(e.g., Osterbrock 1989), where is the incident'LyC ncloud~1 'LyCÑux of ionizing radiation. The observed Ha luminosity andthe assumed cross sectional area that the cloud presents tothe starburst (3.7 ] 3.7 kpc) imply that the average value of

is s~1 cm~2, where is the fraction of'LyC 1.5 ] 106f abs~1 fabsthe incident ionizing photons absorbed by the cloud (i.e.,allowing the cloud to be optically thin). For a preshockdensity of 0.07 cm~3 in the cloud, the implied velocity is

km s~1, and the time for the front to cross theviB 210f abs~1

cloud is about yr. Thus, it is plausible that an4 ] 106f abs~1initially neutral cloud such as those seen in the H I maps ofYun et al. (1993, 1994) has indeed been fully ionized in thetime since the starburst and its superwind turned on.

Our rough estimates suggest that the timescale for thecloud to be fully photoionized is likely to be somewhatshorter than the timescale for it to be fully shock ionized. Ifcorrect, this would mean that the Ha emission produced bythe gas cooling and recombining behind the shock wouldhave a high density and pressure, while the rest of theionized cloud (the as-yet unshocked material) would be at amuch lower pressure and density. We speculate that thebright knots and loops of Ha visible in the image in DB maycorrespond to shock-excited emission from gas with a highpressure, while the fainter and more di†use emission maycome from the low-pressure starburst-photoionized gas thathas not yet been shocked.

4.6. Implications for the ““ Superwind ÏÏ PhenomenonIt is often difficult to associate discrete X-ray sources seen

in the halos of starburst galaxies with activity in the star-burst galaxy itself. Often, it is assumed that most discreteobjects observed in the halos of starbursts are more distantobjects (usually QSOs) seen in projection. Finding a spatialassociation between Ha emission at the redshift of the star-burst galaxy and X-ray emission is an irrefutable way ofbeing able to associate the X-ray emission directly withactivity within the starburst galaxy.

Having argued that this cospatial region of soft X-rayand Ha emission must be due to the outÑowing superwind,it is now evident that at least some starburst galaxies areable to drive material to much larger distances from theirnuclei than would be inferred using the relatively highsurface brightness X-ray or line emission that is generallyobserved along their minor axes (see, e.g., Lehnert &Heckman 1996 ; Heckman et al. 1990 ; DWH). In M82, forexample, the inner ““ continuous ÏÏ region of bright Ha andX-ray emission extends only about 5@ from the nucleusalong the minor axis. The detached ““ ridge ÏÏ of emission thatwe have discussed is at a projected angular separation fromthe nucleus of over twice this distance (about 11@). Thus,without making the association between the Ha and X-rayemission, a variety of interpretations for the source of theX-ray emission from this region would be possible. This

obviously has important implications for our understand-ing of superwinds. It emphasizes that one must be careful inusing the observed spatial scales of the extended emission instarbursts to either access whether or not the wind materialis able to escape the potential of the host galaxy or toestimate the dynamical age of the outÑow.

It is interesting in this vein to consider the long-term fateof the hot gas in the X-ray ridge of M82. To do so, we willcompare the observed temperature of the gas to the ““ escapetemperature ÏÏ at that location in the M82 halo. FollowingWang (1995), the escape temperature for hot gas in a galaxypotential with an escape velocity is given byves Tes^ 1.1] 105 km s~1)2 K. The combined CO] H I rota-(ves/100tion curve of M82 shows a roughly Keplerian decrease withradius for radii greater than 0.4 kpc, and has dropped tovrot60 km s~1 at a radius of 3.5 kpc (Sofue 1998). If thisKeplerian fallo† continues to larger radii, the impliedescape velocity at a radius of 11.6 kpc is only 45 km s~1,and the corresponding escape temperature is Tes ^ 2 ] 104K. This is 450 times cooler than the observed X-ray emis-sion. Even if we assume instead that M82 has an isothermaldark matter halo with a depth corresponding to a circularvelocity of 60 km s~1 (the maximum permitted by thedirectly measured rotation curve) and further assume thatthis halo extends to a radius of 100 kpc, the escape velocityat r \ 11.6 kpc is about 150 km s~1 and K.Tes \ 2.5 ] 105This is still a factor of 36 below the observed temperature ofthe gas in the ridge. This implies that the observed hot gas(and by inference, the superwind itself) will be able to escapethe gravitational potential of M82.

The X-ray/Ha ridge has particularly interesting implica-tions for the physics of superwind emission. While the exis-tence of superwinds is now well established, the process(es)by which they produce X-ray emission is not clear. Thistopic has been nicely reviewed by Strickland (1998). If theoutÑowing wind consists purely of the thermalized ejectafrom supernovae and stellar winds, then the wind Ñuid ishot (D108 K prior to adiabatic cooling) and is so tenuousthat it will be an insigniÐcant source of X-ray emission(Chevalier & Clegg 1985 ; Suchkov et al. 1994). Variousalternatives have been proposed to boost the X-ray lumi-nosity produced by a superwind : (1) Substantial quantitiesof material could be mixed into the wind in or near thestarburst as the stellar ejecta interact with the starburstISM (the wind could be centrally ““ mass loaded ; ÏÏ Suchkovet al. 1996 ; Hartquist, Dyson, & Williams 1997). (2) Largequantities of gas could be added in situ as the wind over-takes and then evaporates or shreds gas clouds in the galac-tic halo (e.g., Suchkov et al. 1994). These clouds could eitherbe preexisting clouds in the halo or material from the diskof the starburst galaxy that has been carried into the haloby the wind. (3) The tenuous wind Ñuid could drive a shockinto a denser volume-Ðlling galactic halo, with the observedX-ray emission arising from the shocked-halo materialrather than the wind Ñuid itself.

In the case of the ridge of emission in the halo of M82, wehave argued that the observable X-ray emission arises asthe wind Ñuid, which, due to adiabatic expansion andcooling, would otherwise have an X-ray surface brightnessbelow the level detectable in the existing ROSAT image,encounters a few-kiloparsec sized cloud in the halo. Wehave argued that the observed X-ray emission is producedin the bow shock created in the superwind just upstream ofthe cloud. This process might be particularly relevant to the

No. 2, 1999 VERY EXTENDED EMISSION IN M82 583

case of M82, which is immersed in an extensive system oftidally liberated gas clouds, shared with its interaction part-ners M81 and NGC 3077. Such material would also providea potential source of optical line emission either by inter-cepting a small portion of the ionizing radiation producedby the central starburst or as it is shock heated by thesuperwind. Given the strong connection between galaxyinteractions/mergers and the starburst phenomenon (e.g.,Sanders & Mirabel 1996 and references therein), thismechanism may be generally applicable to superwinds.

5. SUMMARY AND CONCLUSIONS

We report an analysis of a newly discovered (Devine &Bally 1999) region of spatially coincident X-ray and Haemission to the north of the prototypical starburst/superwind galaxy M82. At the distance of M82 (3.63 Mpc)the dimensions of this correlated ridge of emission are3.7] 0.9 kpc, and it lies at a projected distance of about11.6 kpc above the plane directly along the minor axis ofM82 (and hence along the axis of the superwind). Wemeasure a total Ha Ñux from the ridge of 1.5 ] 10~13 ergss~1 cm ~2. This Ñux yields a total Ha luminosity of2.4] 1038 ergs s~1. The total Ha Ñux for the whole of M82is approximately 4.6 ] 10~11 ergs s~1 cm ~2. Thus, the Haridge constitutes approximately 0.3% of the total observedHa Ñux from the galaxy. X-ray emission is seen over thesame region as the Ha emission by the ROSAT PSPC andHRI. With the 15A (260 pc) o†-axis spatial resolution of theHRI, the ridge is di†use and featureless except for a(probably unrelated) soft point source at its east-northeastend. The best Ðt to the X-ray emission from the ridge is asingle thermal component with a temperature ofkT \ 0.80^ 0.17 keV, absorbed by a column density NH \3.7] 1020 cm~2. The absorption is consistent with the fore-ground Galactic column. The total unabsorbed Ñux fromthe ridge of emission is 1.4] 10~13 ergs cm~2 s~1, and itsluminosity of 2.2] 1038 ergs s~1 comprises about 0.7% ofthe total X-ray emission from M82 in the 0.lÈ2.4 keVROSAT band.

We Ðnd that the observed Ha emission from the ridgecould be produced via photoionization by the dilute radi-ation from the starburst propagating into the M82 haloalong the path cleared by the superwind. On the otherhand, the X-ray emission from the ridge can be most natu-rally explained as the result of shock heating associatedwith an encounter between the starburst-driven galactic““ superwind ÏÏ and a large photoionized cloud in the halo ofM82. Shock heating may also contribute signiÐcantly to theexcitation of the Ha emission, especially the regions of thehighest surface brightness. We Ðnd that the Ha and X-rayluminosities and the X-ray temperature and estimated gaspressure can all be quantitatively understood in this

context. The speciÐc model we favor is one in which theencounter between the superwind and the cloud drives aslow radiative shock into the cloud (producing some of theHa emission) and also leads to an adiabatic X-rayÈemittingbow shock in the superwind Ñuid upstream from the cloud.The cloud is now fully ionized, and so is not detected in theH I maps of Yun et al. (1993, 1994). The wind-cloud inter-action process may be especially relevant to M82, the haloof which contains tidally liberated gas clouds. However, itmay well occur generally in powerful starbursts (which areusually triggered by galaxy interactions or mergers).

The existence of a region of spatially coincident X-rayand Ha emission with strong evidence for wind/cloud inter-action is important for several reasons. First, our analysissheds new light on the physical processes by which galacticwinds produce X-ray emission and implies that wind/cloudcollisions are an important emission-producing process.The emission ridge in M82 highlights the potential pitfallsin using the observed X-ray or Ha emission to trace the sizeand infer the ages or fates of galactic winds ; the winds mayextend to radii far larger than the region over which theyhave densities and temperatures high enough to produceobservable emission. Wind-cloud encounters then o†er usthe possibility of ““ lighting up ÏÏ the wind at radii that areotherwise inaccessible. The X-ray ridge in M82 implies thatthe wind is able to reach large heights above the plane of thegalaxy, thereby increasing the likelihood that it will be ableto eventually escape the galaxy potential well. Indeed, theobserved gas temperature exceeds the local ““ escapetemperature ÏÏ from the M82 gravitational potential well byabout 2 orders of magnitude. This result strengthens theargument that starburst-driven galactic winds are impor-tant sources of the metal enrichment and heating of theintergalactic medium.

The authors wish to thank Kitt Peak National Observa-tory for their generous allocation of observing time andtheir sta† for making sure that the observations werecarried out efficiently and e†ectively. We thank Ed Moran,David Strickland, and Crystal Martin for illuminating dis-cussions about the physics of superwinds, and DavidDevine for communicating his discovery paper in advanceof its publication. We would like to express our sincerestthank you to the referee for carefully reading the manu-script and for providing us with detailed comments thathave substantially improved this presentation. This workwas supported in part by NASA LTSA grants NAGW-3138to T. H. and NAGW-4025 to K. W. This research has madeuse of the IRAF package provided by NOAO and NASA/IPAC extragalactic database (NED), which is operated bythe Jet Propulsion Laboratory, Caltech, under contractwith the National Aeronautics and Space Administration.

REFERENCES

Axon, D. J., & Taylor, K. 1978, Nature, 274, 37Bedogni, R., & Woodward, P. R. 1990, A&A, 231, 481Bernloehr, K. 1992, A&A, 263, 54Bhattacharya, D., The, L. S., Kurfess, J. D., & Clayton, D. D. 1994, ApJ,

437, 173Binette, L., Dopita, M., & Tuohy, I. 1985, ApJ, 297, 476Bland, J., & Tully, R. B. 1988, Nature, 334, 43Bregman, J. N., Schulman, E., & Tomisaka, K. 1995, ApJ, 439, 155Burton, W. B., & Hartmann, D. 1997, Atlas of Galactic Neutral Hydrogen

(Cambridge : Cambridge Univ. Press)Chevalier, R., & Clegg, A. 1985, Nature, 317, 44Courvoisier, T. J. L., Reichen, M., Blecha, A., Golay, M., & Huguenin, D.

1990, A&A, 238, 63

Dahlem, M., Weaver, K. A., & Heckman, T. M. 1998, ApJS, 118, 401(DWH)

Devine, D., & Bally, J. 1999, ApJ, 510, 197 (DB)Freedman, W. L., et al. 1994, ApJ, 427, 628Gibson, B., Loewenstein, M., & Mushotzky, R. 1997, MNRAS, 290, 623Goetz, M., Downes, D., Greve, A., & McKeith, C. 1990, A&A, 240, 52Ha†ner, L. M., Reynolds, R. J., & Tufte, S. L. 1998, ApJ, 501, 83Hartquist, T., Dyson, J., & Williams, R. 1997, ApJ, 482, 182Heckman, T. M., Armus, L., & Miley, G. K. 1990, ApJS, 74, 833Heckman, T., Dahlem, M., Lehnert, M., Fabbiano, G., Gilmore, D., &

Waller, W. 1995, ApJ, 448, 98Hughes, D. H., Gear, W. K., & Robson, E. I. 1994, MNRAS, 270, 641Hunter, D., Hawley, W., & Gallagher, J. 1993, AJ, 106, 1797

584 LEHNERT, HECKMAN, & WEAVER

Hurwitz, M., Jelinsky, P., & Dixon, W. V. D. 1997, ApJ, 481, L31Johnson, H. L. 1966, ApJ, 143, 187Kau†mann, G., & Charlot, S. 1998, MNRAS, 294, 705Kerp, J., Burton, W. B., Egger, R., Freyberg, M. J., Hartmenn, D., Kalberla,

P. M. W., Mebold, U., & Pietz, J. 1999, A&A, 342, 213Kerp, J., Mack, K.-H., Egger, R., Pietz, J., Zimmer, F., Mebold, U., Burton,

W. B., & Hartmenn, D. 1996, A&A, 312, 67Klein, R. I., McKee, C. F., & Colella, P. 1994, ApJ, 420, 213Kronberg, P., Lesch, H., & Hopp, U. 1999, ApJ, 511, 56Larkin, J. E., Graham, J. R., Matthews, K., Soifer, B. T., Beckwith, S.,

Herbst, T., & Quillen, A. 1994, ApJ, 420, 159Lehnert, M. D., & Heckman, T. M. 1996, ApJ, 462, 651Leitherer, C., & Heckman, T. M. 1995, ApJS, 96, 9Leitherer, C., Ferguson, H., Heckman, T., & Lowenthal, J. 1995, ApJ, 454,

L19Lynds, R., & Sandage, A. 1963, ApJ, 137, 1005Mateo, M. 1998, ARA&A, 36, 435McCarthy, P., Heckman, T. M., & van Breugel, W. 1987, AJ, 93, 264McKeith, C., Greve, A., Downes, D., & Prada, F. 1995, A&A, 293, 703McLeod, K., Rieke, G., Rieke, M., & Kelly, D. 1993, ApJ, 412, 111Moran, E. C., & Lehnert, M. D. 1997, ApJ, 478, 172OÏConnell, R. W., & Mangano, J. J. 1978, ApJ, 221, 62Osterbrock, D. 1989, Astrophysics of Gaseous Nebulae and Active Galac-

tic Nuclei (Mill Valley : Univ. Science Books)Petuchowski, S. J., Bennett, C. L., Haas, M. R., Erickson, E. F., Lord, S.,

Rubin, R., Colgan, S., & Hollenbach, D. 1994, ApJ, 427, L17Ponman, T., Cannon, D., & Navarro, J. 1999, Nature, 397, 135

Ptak, A., Serlemitsos, P., Yaqoob, T., Mushotzky, R., & Tsuru, T. 1997, AJ,113, 1286

Reuter, H. P., Klein, U., Lesch, H., Wielebinski R., & Kronberg, P. 1994,A&A, 282, 724

Reynolds, R. J., Tufte, S. L., Kung, D. T., McCullough, P. R., & Heiles, C.1995, ApJ, 448, 715

Rieke, G., Loken, K., Rieke, M., & Tamblyn, P. 1993, ApJ, 412, 99Rosner, R., Golub, L., & Vaiana, G. 1985, ARA&A, 23, 413Sanders, D., & Mirabel, I. F. 1996, ARA&A, 34, 749Satyapal, S., Watson, D. M., Pipher, J. L., Forrest, W. J., Greenhouse,

M. A., Smith, H. A., Fischer, J. & Woodward, C. E. 1997, ApJ, 483, 148Schreiber, N. M. 1998, Ph.D. thesis, Ludwig-Maximilians-Universita� t

(Mu� nchen)Seaquist, E. R., Kerton, C. R., & Bell M. B. 1994, ApJ, 429, 612Shopbell, P. L., & Bland-Hawthorn, J. 1998, ApJ, 493, 129Sofue, Y. 1998, PASJ, 50, 227Somerville, R., Primack, J., & Faber, S. 1998, preprint astro-ph/9806228Strickland, D. K. 1998, Ph.D. thesis, Univ. BirminghamStrickland, D. K., Ponman, T. J., & Stevens, I. R. 1997, A&A, 320, 378Suchkov, A. A., Balsara, D. S., Heckman, T. M., & Leitherer, C. 1994, ApJ,

430, 511Suchkov, A. A., Berman, V., Heckman, T. M., & Balsara, D. 1996, ApJ, 463,

528Telesco, C. M. 1988, ARA&A, 26, 343Wang, B. 1995, ApJ, 444, 590Yun, M. S., Ho, P. T. P., & Lo, K. Y. 1993, ApJ, 411, L17ÈÈÈ. 1994, Nature, 372, 530