Interstellar Weather Vanes: GLIMPSE Mid-Infrared Stellar-Wind Bowshocks in M17 and RCW49
Matthew S. Povich, Robert A. Benjamin, Barbara A. Whitney, Brian L. Babler, Remy Indebetouw, Marilyn R. Meade, Ed Churchwell
aa r X i v : . [ a s t r o - ph ] A ug Accepted for publication in the Astrophysical Journal
Interstellar Weather Vanes: GLIMPSE Mid-Infrared Stellar-WindBowshocks in M17 and RCW49
Matthew S. Povich , , Robert A. Benjamin , , Barbara A. Whitney , Brian L. Babler ,R´emy Indebetouw , Marilyn R. Meade , and Ed Churchwell ABSTRACT
We report the discovery of six infrared stellar-wind bowshocks in the Galacticmassive star formation regions M17 and RCW49 from
Spitzer
GLIMPSE (Galac-tic Legacy Infrared Mid-Plane Survey Extraordinaire) images. The InfraRed Ar-ray Camera (IRAC) on the
Spitzer Space Telescope clearly resolves the arc-shapedemission produced by the bowshocks. We combine
Two Micron All-Sky Survey(2MASS), Spitzer, MSX, and
IRAS observations to obtain the spectral energydistributions (SEDs) of the bowshocks and their individual driving stars. Weuse the stellar SEDs to estimate the spectral types of the three newly-identifiedO stars in RCW49 and one previously undiscovered O star in M17. One of thebowshocks in RCW49 reveals the presence of a large-scale flow of gas escapingthe H II region at a few 10 km s − . Radiation-transfer modeling of the steep risein the SED of this bowshock toward longer mid-infrared wavelengths indicatesthat the emission is coming principally from dust heated by the star driving theshock. The other 5 bowshocks occur where the stellar winds of O stars sweep updust in the expanding H II regions. Subject headings: infrared: ISM — shock waves — stars: winds — H II regions:individual(RCW49, M17) Dept. of Astronomy, University of Wisconsin-Madison, 475 N. Charter St., Madison, WI 53706 Dept. of Physics, University of Wisconsin-Whitewater, 800 W. Main St, Whitewater, WI 53190 Space Science Institute, 3100 Marine Street, Suite A353, Boulder, CO 80303-1058 Dept. of Astronomy, University of Virginia, Charlottesville, VA 22903-0818 email: [email protected] email: [email protected]
1. Introduction
The Solar wind ends in a termination shock (e.g. Decker et al. 2005), where the pressureof the heliosphere balances the ram pressure of the surrounding interstellar medium (ISM).Massive stars with more energetic winds generate much stronger shocks. In cases wherethe relative motion between the star driving the wind and the ambient ISM is large, theshock will be bent back around the star. If the relative velocity is supersonic, the ambientISM gas is swept into a second shock, forming an arc-shaped “bowshock” that is separatedfrom the termination shock by a contact discontinuity. Stellar-wind bowshocks have been re-ported for a variety of sources, including nearby runaway O stars (van Buren & McCray 1988;van Buren, Noriega-Crespo, & Dgani 1995; Noriega-Crespo et al. 1997; Brown & Bomans 2005;Comer´on & Pasquali 2007; France et al. 2007), high-mass X-ray binaries (Churchwell et al.1992; Kaper et al. 1997; Huthoff & Kaper 2002), LL Ori-type stars (Bally et al. 2000), radiopulsars (Gaensler & Slane 2006), Galactic center O stars (Geballe et al. 2004, 2006), andmass-losing red giants (Martin et al. 2007). Recently, an infrared (IR) bowshock has beenobserved around the young A-type star δ Vel (G´asp´ar et al. 2008). Cometary H II regionsalso resemble bowshocks, due either to density gradients in the ambient gas or to motionof the ionizing source with respect to the interstellar surroundings (van Buren et al. 1990;Arthur & Hoare 2006). Both the direction of a bowshock and its “standoff distance” fromthe star generating the wind are determined by the velocity of the star with respect to thesurrounding medium. In the case of runaway O stars, this is dominated by the high spacemotion of the star.We report the detection of three mid-IR bowshocks in each of two massive star formationregions: M17 and RCW49. Two of the bowshocks in M17 are around known O stars. We willdemonstrate that the other bowshocks are also around likely O stars. Since these stars arein or near expanding H II regions, we find that the direction of the bowshock is determinedprincipally by the flow of the ISM rather than the space motion of the star.
2. Observations and Interpretation
The Galactic Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE; Benjamin et al.2003) programs have mapped the inner Galactic midplane ( | l | ≤ ◦ ) using the IRAC in-strument on the Spitzer Space Telescope (2 ′′ resolution; Fazio et al. 2004). RCW49, locatedat ( l, b ) = (284 . , − . ′′ pixels) ofthis region in the four IRAC bands: 3.6, 4.5, 5.8, and 8.0 µ m. An overview of these obser- 3 – b ( d e g r ee s ) Fig. 1.— GLIMPSE full-color image of M17 ( blue: µ m, green: µ m, orange: µ m, red: µ m). The region containing the bowshocks M17-S1, -S2, and -S3 is enlarged(scalebar shows 30 ′′ = 0 .
23 pc at 1.6 kpc). The central ring of O stars in the NGC 6618cluster is circled. The bowshocks, along with the prominent pillar structure near M17-S2,all appear to be oriented in the general direction of CEN 1, the O4-O4 binary system in thecenter of the ionizing cluster.vations was given by Churchwell et al. (2004). M17, at ( l, b ) = (15 . , − . ′′ pixels. Povich et al. (2007) studied the diffuse emission morphology of M17 at multiple wave-lengths from IR to radio, constraining the distance to 1.4–1.9 kpc. We will assume thewidely-adopted value of 1.6 kpc for the M17 distance, following Nielbock et al. (2001). Thewinds and radiation of the O stars in the NGC 6618 cluster have excavated a cavity in thecenter of the M17 H II region. The cavity is filled with hot, X-ray-emitting gas from shockedstellar winds (Townsley et al. 2003). Povich et al. (2007) noted the presence of 3 appar- 4 –ent stellar-wind bowshocks along the northern wall of the cavity, along with a prominent“elephant trunk,” or pillar, all oriented in the direction of the central ring of 7 O stars inNGC 6618. These structures are highlighted in Figure 1, a GLIMPSE image of M17 with anenlargement of the region containing the bowshocks. The bowshocks stand out as yellow-orange features in the image because they are faint at 3.6 µ m (colored blue ) compared tothe other 3 IRAC bands.We designate these stellar-wind bowshocks in Figure 1 with the name of the regionfollowed by an identification number, in order of increasing Galactic longitude (for example,M17-S1). IR fluxes for all of the bowshock driving stars are given in Table 1, and background-subtracted IR fluxes from aperture photometry of the bowshocks are listed in Table 2.Spectral types of two of the driving stars have been determined photometrically (Bumgardner1992) and spectroscopically (Hanson, Howarth, & Conti 1997). M17-S1 is associated withCEN 16, an O9–B2 star. The larger bowshock, M17-S2, is driven by CEN 18, an earlier-typestar (O7–O8). Both bowshocks were detected at 10.5 and 20 µ m by Nielbock et al. (2001).These observations did not resolve the arc shapes of the bowshocks, and Nielbock et al.(2001) attributed the excess IR emission to circumstellar disks and classified CEN 16 andIRS 9 (the star visible just to the right of the arrowhead in Figure 1) as massive protostars.M17-S3 lies outside of the field analyzed by Hanson, Howarth, & Conti (1997), and thedriving star does not appear in any catalog of the region. It is not found in any GLIMPSEsourcelist, because the bright, spatially variable diffuse background prevented the automaticextraction of the point source. We measured the flux this source manually using a 5 ′′ aper-ture. Using the spectral energy distribution (SED) fitting tool of Robitaille et al. (2007),we fit the IR fluxes of this star (Table 1) with Kurucz (1993) stellar atmosphere models.Following the method described by Watson et al. (2008), we scale the models to the 1.6-kpcdistance of M17 and estimate a spectral type of O7 V for the star. Carrying out the sameanalysis on 7 other O stars in M17 with independently known spectral types (Povich et al.2008), we estimate that our spectral typing is accurate to within 2 subclasses. RCW49 presents a more complicated morphology than M17. Churchwell et al. (2004)discussed the structure and spectrum of the diffuse emission. Like M17, RCW49 is filled withX-ray gas (Townsley et al. 2005). The interstellar structures are dominated by two largecavities. The first, blown out to the West, contains the massive young cluster Westerlund 2,and the second is an enclosed bubble around the Wolf-Rayet star WR 20b (Figure 2). The 5 –Table 1. 2MASS and IRAC a IR Fluxes for Bowshock Driving Stars (mJy)ID l b F [ J ] F [ H ] F [ K ] F [3 . F [4 . b F [5 . b M17-S1 15.07486 -0.64607 196 214 186 96 67 · · ·
M17-S2 15.08126 -0.65699 427 534 495 257 181 120M17-S3 c ≤ ≤ ≤ ≤ a None of these stars was detected in the IRAC [8.0] band. b In cases where bowshock emission appears to cause a mid-IR excess over the stellar spec-trum, the [4.5] and [5.8] fluxes are reported as upper limits. Due to suspected contaminationfrom the bowshock or other diffuse emission, these fluxes are treated as upper limits to thestellar flux. c Because the star driving SWB M17-3 is surrounded by bright, complex diffuse backgroundemission, it was not extracted as part of the GLIMPSE point-source catalog. The fluxesreported here were measured using aperture photometry. 6 –Table 2. IR Fluxes for Bowshocks (mJy)IRAC Fluxes a ID l apex b apex F [3 . b F [4 . F [5 . F [8 . ≤
34 71 ± ±
42 1110 ± ±
17 245 ±
48 1240 ±
159 8700 ± ±
16 240 ±
19 620 ±
100 1800 ± . ± . ± ± ± ≤ . ± . ± ± ≤
21 50 ± ±
20 825 ± MSX
Fluxes c MSXC6 Name F [8 . F [12 . F [14 . F [21 . . × IRAS
Fluxes c IRAS
Name F [12] F [25] F [60] F [100]RCW49-S1 IRAS 10205-5729 3810 4 . × ≤ . × . × The IRAC fluxes are background-subtracted. Fluxes were measured using irregularapertures drawn to enclose all of the visible bowshock structure. Separate apertures wereused to estimate the background flux. For each bowshock, the same set of apertures wasused for all IRAC wavelengths. b In cases where stellar emission appears to be confused with bowshock emission, the[3.6] bowshock fluxes reported are upper limits only. c RCW49-S1 is a point source in both the
MSX and
IRAS catalogs. All of the otherbowshocks are confused with bright diffuse background IR emission features at the res-olutions of
MSX and
IRAS . 7 – b ( d e g r ee s ) Fig. 2.— GLIMPSE full-color image of RCW49 ( blue: [3.6], green: [4.5], orange: [5.8], red: [8.0]). The bowshocks RCW49-S1, -S2, and -S3, are enlarged in 3 separate insets (scalebarsare 30 ′′ ≈ . ± II region, but we believe it is associated with RCW49 for the following reasons: (1) TheRCW49 distance is consistent with the luminosity of the driving star being an O star (seebelow); and (2) the bowshock points (approximately) toward the central cluster, Westerlund2. Because it is far from any other bright IR source, RCW49-S1 appears as a point source inboth the Midcourse Space Experiment ( MSX ; Price et al. 2001) and
Infrared AstronomicalSatellite ( IRAS ; Beichman et al. 1988) point source catalogs, and these fluxes are also given 8 –in Table 2.RCW49-S2 is oriented away from Westerlund 2 and appears to be influenced primarilyby the nearby WR 20b. RCW49-S3 is oriented in the general direction of Westerlund 2.None of the RCW49 bowshocks points directly back toward the central cluster, while allthree M17 bowshocks do. Perhaps the expanding bubbles driven by Westerlund 2 and theWolf-Rayet stars are interacting turbulently, producing non-radial components to the flows.It is also possible that the bowshock driving stars have large orbital motions relative to thedynamic interstellar medium.The three bowshock driving stars in RCW49 are of previously undetermined spectraltype, so we again estimate the spectral type by fitting model SEDs to the broadband fluxesof Table 1, scaled to both the 4.2 kpc and 6 kpc distances. The results are given in Table 3.[ht] All three stars are plausibly O stars. Assuming 4.2 kpc, the driving star of RCW49-S1is fit as O6 V, RCW49-S2 as O 9 V, and RCW49-S3 as O5 V (or O9 III). If we increasethe distance to 6 kpc, the fits become O 5 III, O6 V, and O5 V (or O6.5 III), respectively(see Watson et al. 2008, for an explanation of the degeneracy between luminosity classes).The spectral types at 6 kpc seem improbably luminous. Highly luminous and windy starsdominate the dynamics of their local ISM. We do not observe IR bowshocks around any ofthe earliest-type stars in either RCW49 or M17, because they have blown large, evacuatedcavities in the centers of the H II regions. Little or no ambient material remains close to thestars to produce a bowshock.Apart from the Wolf-Rayet systems, one of the earliest stars in RCW49 is G284.2642-00.3156. Using optical spectroscopy, Uzpen et al. (2005) classified this star as O4 V(f) andderived a spectrophotometric distance of 3 . ± .
3. Bowshock Properties
The standoff distance d w of a bowshock from its driving star is the point where themomentum flux of the stellar wind balances the momentum flux of the ambient medium: n w v w = n v . Following van Buren & McCray (1988), we normalize the stellar wind mass 9 –Table 3. Bowshock Standoff Distances and Estimated Stellar Wind PropertiesID M w, − v w, d w cos i d cl cos i v n / , (cos i ) − type a (pc) (pc) (km s − )M17 Distance = 1 . ≤ . ≤ . ∼ . ∼ . . ∼
16 0.23 23.2 26RCW49-S2 O6 V ∼ . > . > a Spectral types for the stars driving M17-S1 and -S2 are taken from CEN andHanson, Howarth, & Conti (1997). All others were estimated by fitting Kurucz(1993) stellar atmosphere models to the broadband IR fluxes (Table 1) and scal-ing to the distance of M17 or RCW49. Spectral types given for RCW49 are highlyuncertain due to the disputed distance to that region. b Estimates of stellar mass-loss rates ˙ M w, − = ˙ M w / (10 − M ⊙ yr − ) and wind veloc-ities v w, = v w / (10 cm s − ) are based upon Vink, de Koter, & Lamers (2001) andFullerton, Massa, & Prinja (2006). 10 –loss rate, ˙ M w, − = 10 − M ⊙ yr − , stellar wind velocity v w, = 10 cm s − , ambient hydrogenparticle density of n , = 10 cm − , and use a mean ISM gas mass per hydrogen atom µ = 2 . × − g. Assuming a spherically symmetric stellar wind with a mass-loss rategiven by ˙ M = 4 πd w µn w v w , the velocity of the star with respect to the ambient ISM can bewritten as v = 1 . (cid:18) d w pc (cid:19) − ( ˙ M w, − v w, ) / n − / , [km s − ] , (1)where v w is the terminal velocity of the stellar wind. Values of v n / , and d w for eachbowshock are presented in Table 3. The standoff distance can be measured only as d w cos i on the sky, where i is the inclination (or viewing angle) made by the line connecting the starwith the apex of the bowshock against the plane of the sky ( i = 0 if the line joining the starto the bowshock apex lies in the plane of the sky). Because a bowshock oriented at high i will not produce an arc morphology, it is likely that i . ◦ , and hence cos i will not differgreatly from unity in our measurements. The distance from each bowshock to the likelysource of the large-scale ISM flow, measured on the sky as d cl cos i , are also presented forreference in Table 3.The mass-loss rates and stellar wind velocities in Equation 1 suffer from high dispersionas a function of spectral type (Fullerton, Massa, & Prinja 2006), a factor of 2 or even greater,and this is compounded by a comparable level of uncertainty in the spectral types. Theuncertainty on our measurements of d w ranges from ∼
20% for the largest bowshocks (M17-S2and RCW49-S1) to ∼
40% for the smaller bowshocks that are barely resolved by IRAC. Weestimate that v n / is uncertain by a factor of 2 in M17 and up to a factor of 3 in RCW49,where the spectral types of the driving stars are less constrained. These uncertainties,reflected in the range of values for v n / in Table 3, are dominated by the uncertainty inthe stellar wind properties.We have neglected the effects of turbulent pressure in our calculation of the momentumflux balance of the bowshocks. This is a potentially significant contributor to the totalISM pressure held off by the bowshocks. The effect of turbulent pressure would be tosystematically decrease the standoff distance d w , causing us to overestimate v .In reality, for the fast winds of early-type stars, the observed bowshock is displaced fromthe standoff distance by a significant amount. This happens because the cooling timescale ofthe shocked stellar wind is very long. The result is a thick layer of hot gas intervening betweenthe wind terminal shock at the standoff distance and the thin, dense layer of interstellar gasand dust forming the observed bowshock. The numerical simulations of Comer´on & Kaper The average value of cos i for 0 ≤ i ≤ ◦ is 0.9.
11 –(1998) predict that the bowshock should be located at twice the standoff distance from thedriving star. In this case, our derived values of v n / in Table 3 would be underestimated bythe same factor of 2. This systematic correction is comparable to the intrinsic uncertaintiesin our estimates of v n / , and it partially compensates for the effects of neglecting turbulentpressure in the ambient ISM. Therefore, the assumption that the observed distance of thebowshocks from the driving stars corresponds to the standoff distance d w should not have alarge impact on our results, and we find that the cautious application of Equation 1 yieldsreasonable results.Orbital velocities of O stars in massive clusters are typically <
10 km s − . The expansionspeed of ionized gas in H II regions is generally comparable to the sound speed of ∼
10 km s − ,and this appears to be true in M17 (Pellegrini et al. 2007). Most of the bowshocks (theexception being RCW49-S1) are apparently located within the ionized gas of the radio H II regions. The likely explanation for the bowshock emission in the IR is that dust in the H II regions (Povich et al. 2007) is swept-up by the bowshocks. The observed average electrondensity in the Northern bar of the M17 H II region is ∼ cm − (Felli, Churchwell, & Massi1984), so the values of v n / , listed in Table 3 are likely to be close to the actual relativevelocities of the stars and the ISM for most of the bowshocks.The values of v n / , calculated for M17-S1 and -S2 are in good agreement. M17-S3,however, presents a different picture. M17-S3 appears to be associated with a “teardrop”structure (Figure 1) in the photodissociation region (PDR), near the ionization front. Theambient density surrounding this star could be significantly higher than the density withinthe H II region. If v for this bowshock is comparable to that of the other 2 bowshocks inM17, then n , ∼ .
25 cm − in this location. This value agrees with the measurements ofelectron density in the dense clumps of ionized gas in M17 (Felli, Churchwell, & Massi 1984).The unusual morphology of the diffuse IR emission associated with M17-S3 suggests thatthe star may have recently emerged from an evaporating globule on the edge of the PDR; alarger, more evolved analog of the nearby pillar structure seen in Figure 1.In RCW49, all 3 bowshocks are found in very different locations, but we note that 2of the bowshocks appear to be similar in size, color, and ambient environment (Figure 2).The largest bowshock, RCW49-S1, is different, since it is located relatively far (16.2 pc at4.2 kpc) from Westerlund 2, outside the H II region. The presence of RCW49-S1, alongwith its orientation, indicates that RCW49 vents a large-scale flow of gas through the cavityopening to the West. This flow likely originates in the combined winds of the Westerlund 2cluster and thus should be much more diffuse than the H II region gas (Townsley et al. 2003,2005). Assuming a density of 1 cm − in the flow from Westerlund 2, the standoff distanceof RCW49-S1 at 4.2 kpc gives a flow velocity of ∼
350 km s − . Such a high value of v is 12 –reasonable, given that the gas must move supersonically relative to the star to produce ashock, and the sound speed in the hot, rarefied gas of the flow streaming away from the H II region is ∼
100 km s − .Fig. 3.— Radiation transfer model of RCW49-S1. Top:
Model SEDs plotted with the fluxesof the bowshock and driving star from Tables 1 & 2 ( triangles: crosses: MSX ; diamonds: IRAS ). The green curve includes low-density material in a shell2–3 pc from the star in order to match the IRAS
60 and 100 µ m fluxes. Bottom:
Image ofthe model bowshock at GLIMPSE wavelengths (compare to Fig. 2).Because RCW49-S1 was detected by
MSX and
IRAS in addition to GLIMPSE, we canconstruct the SED of the bowshock from 4.5 µ m to 100 µ m (Table 2). We computed SEDmodels of RCW49-S1 using a 3-D radiative equilibrium code (Whitney et al. 2003) modifiedto include very small grain (VSG) and PAH emission (Wood et al. 2008). We used thecanonical mass fraction for VSG/PAH grains of 5% (Draine & Li 2007). We modeled thebowshock geometry as a paraboloid with the apex offset by d w = 0 .
16 pc from the O6 V star 13 –(assuming i = 0 and the 4.2 kpc distance). For models that reproduced the observed images,the SED shape was insensitive both to the thickness of the shock and to the radial densityprofile of the dust (with the total mass scaled to match the observed SED). The black line inFigure 3 shows the SED for a model with a density varying as r − from the 0.16 pc standoffdistance out to a radius of 1.5 pc. The observed SED shortward of 30 µ m is well-matchedby the model. For radial density exponents from − ∼ . ⊙ to 2 M ⊙ , respectively, assuming a dust-to-gas mass ratio of 0.01. The opticaldepth in all models is low, with A V < .
02 within 1 pc. To match the
IRAS data at 60 and100 µ m requires low-density material farther from the star. The green line in Figure 3 is amodel including a shell 2–3 pc from the star. This material added mid-IR PAH emission,so we lowered the VSG/PAH mass fraction to 3% to continue to match the mid-IR SED,and the image still matches the data well. The A V through the bowshock in this model is0.25, most of it due to the outer shell. Along the line of sight to the star, A V = 10, sothe bowshock and shell contribute a negligible fraction of the interstellar extinction. Thesemodels show that dust distributed in a bowshock geometry matches both the images and theSED reasonably well. The IR emission from the bowshock can be explained by reprocessedstellar radiation without any additional dust heating by the shock.The structure of RCW-S1 is probably significantly different from that of the other 5bowshocks in our sample, because it is located in a distinct interstellar environment. Becauseof the high temperature and low density in the flow outside the H II region, the shockedinterstellar gas cannot cool quickly enough to form a dense layer behind the bowshock, aslikely happens in the other 5 cases. The shock is approximately adiabatic, remaining veryhot and only moderately compressed (by a factor of ∼
4) as it flows past the star, forming arelatively thick layer. Hence, while RCW-S1 is the only bowshock in our sample observed atenough different mid-IR wavelengths to allow us to create a meaningful model of the emission,it may not be appropriate to draw strong conclusions about the remaining bowshocks basedupon this model.
4. Summary
We have observed 6 prominent IR bowshocks in M17 and RCW49. These objects appearto be produced by the winds of individual O stars colliding with large-scale interstellar gasflows in the H II regions. One bowshock, M17-S3, may be the leading edge of an evaporatingglobule containing a newly-formed and previously undiscovered O star in the well-studiedM17 region. All three bowshocks associated with RCW49 lead us to identify new candidateO stars. Our stellar classifications also suggest that the true distance to RCW49 is less than 14 –the kinematic distance of 6 kpc.The bowshocks are bright at IR wavelengths due to emission from dust swept up fromthe ambient ISM and heated by radiation from the bowshock driving stars. As G´asp´ar et al.(2008) note, IR excess emission from a bowshock could be attributed to the presence of acircumstellar disk, particularly when the bowshock morphology is not spatially well-resolved.This can be a pitfall for observational studies of accreting massive protostars.The collective winds of the most luminous stars in young, massive clusters produceoverlapping large-scale flows that hollow out thermally hot cavities in the parent molecularcloud (Townsley et al. 2003). The largest bowshock presented here, RCW49-S1, is evidencethat the combined winds of the ionizing stars in Westerlund 2 have escaped the H II region,creating a flow of hot gas moving at a few 10 km s − that extends at least 16 pc away fromRCW49.The driving stars of the other 5 bowshocks are surrounded by ionized gas and dustof their natal H II regions, where the density of the ambient medium ( n ∼ cm − ) issufficiently high to produce the observed bowshocks with a relative velocity of only 10–20km s − . The winds of the bowshock driving stars do not directly encounter the > − winds from the most massive stars in the cluster. Instead, the bowshocks are shaped by theexpansion of the ionized gas in the H II regions relative to the orbital motions of the stars.Eventually, supernova explosions will produce high velocity shock waves that heat anddisperse the original gas cloud. In star forming regions like M17 and RCW49 that havenot yet been disrupted by supernovae, IR bowshocks serve as interstellar “weather vanes,”indicating the speed and direction of large-scale gas flows at points within and around giantH II regions.We thank the anonymous referee for incisive and very useful suggestions that helpedus improve this work. We are grateful to Joseph Cassinelli, Eric Pellegrini, John Raymond,Ellen Zweibel, Heidi Gneiser, and Don Cox for useful conversations while preparing thispaper. M. S. P. thanks the members of the “dissertator club,” Kathryn Devine, K. TabethaHole, and Nicholas Murphy, for their helpful comments. This work was supported by NSFgrant AST-030368 (E. B. C.) and NASA/JPL Contracts 1282620 and 1298148. Additionalsupport was provided by the NASA Theory Program (NNG05GH35G; B. A. W.). R. I.acknowledges support from a Spitzer Fellowship at the time that these data were analyzed,and from JPL RSA1275467. 15 – REFERENCES
Arthur, S. J., & Hoare, M. G. 2006, ApJS, 165, 283Ascenso, J., Alves, J., Beletsky, Y., & Lago, M. T. V. T. 2007, A&A, 466, 137Bally, J., O’Dell, C. R., & McCaughrean, M. J. 2000, AJ, 119, 2919Beichman, C. A., Neugebauer, G., Habing, H. J., Clegg, P. E., & Chester, T. J., eds. 1988,IRAS Catalogs and Atlases: Explanatory Supplement, NASA RP-1190Benjamin, R.A. et al. 2003, PASP, 115, 953Brown, D., & Bomans, D. J. 2005, A&A, 439, 183Bumgardner, T. E. 1992, M.S. thesis, Ohio State Univ., ColumbusChini, R., Els¨asser, H., & Neckel, T. 1980, A&A, 91, 186 (CEN)Churchwell, E., Bieging, J. H., van der Hucht, K. A., Williams, P. M., Spoelstra, T. A. Th.,& Abbott, D. C. 1992, ApJ, 393, 329Churchwell, E. et al. 2004, ApJS, 154, 322Comer´on, F. & Kaper, L. 1998, A&A, 338, 273Comer´on, F. & Pasquali, A. 2007, A&A, 467, L23Dame, T. M. 2007, ApJ, 665, L163Decker, R. B. et al. 2005, Science, 309, 2020Draine, B. T. & Li, A. 2007, ApJ, 657, 810Fazio, G. G. et al. 2004, ApJS, 154, 10Felli, M., Churchwell, E., & Massi, M. 1984, A&A, 136, 53France, K., McCandliss, S. R., & Lupu, R. E. 2007, ApJ, 655, 920Fullerton, A. W., Massa, D. L., & Prinja, R. K. 2006, ApJ, 637, 1025Gaensler, B. M., & Slane, P. O. 2006, ARA&A, 44, 17G´asp´ar, A., Su, K. Y. L., Rieke, G. H., Balog, Z., Kamp, I., Mart´ınex-Galarza, J. R., &Stapelfeldt, K. 2008, ApJ, 672, 974 16 –Geballe, T.R., Rigaut, F., Roy, J.-R., & Draine, B.T. 2004, ApJ, 602, 77Geballe, T. R., Najarro, F., Rigaut, F., & Roy, J.-R. 2006, ApJ, 652, 370Hanson, M. M., Howarth, I. D., & Conti, P. S. 1997, ApJ, 489, 698Huthoff, F., & Kaper, L. 2002, A&A, 383, 999Kaper, L., van Loon, J. T., Augusteijn, T., Goudfrooij, P., Patat, F., Waters, L. B. F. M.,& Zijlstra, A. A. 1997, ApJ, 475, L37Kurucz, R. 1993, ATLAS9 Stellar Atmosphere Programs and 2 km/s grid. Kurucz CD-ROMNo. 13. Cambridge, MA: Smithsonian Astrophysical ObservatoryMartin, D, C. et al , 2007, Nature, 448, 780Nielbock, M., Chini, R., J¨utte, M., & Manthey, E. 2001, A&A, 273, 284Noriega-Crespo, A., van Buren, D., & Dgani, R. 1997, AJ, 113, 780Pellegrini, E. W. et al. 2007, ApJ, 658, 1119Povich, M. S., Stone, J. M., Churchwell, E., Zweibel, E. G., Wolfire, M. G., Babler, B. L.,Indebetouw, R., Meade, M. R., & Whitney, B. A. 2007, ApJ, 660, 346Povich, M. S. et al. 2008, ApJ, submittedPrice, S. D., Egan, M. P., Carey, S. J., Mizuno, D. R., and Kuchar, T. A. 2001, AJ, 121,2842Rauw, G., Manfroid, J., Gosset, E., Naz´e, Y., Sana, H., De Becker, M., Foellmi, C., &Moffat, A. F. J. 2007, A&A, 463, 981Robitaille, T. P., Whitney, B. A., Indebetouw, R., & Wood, K. 2007, ApJS, 169, 328Townsley, L. K., Feigelson, E. D., Montmerle, T., Broos, P. S., Chu, Y.-H., & Garmire, G.P. 2003, ApJ, 593, 874Townsley, L., Feigelson, E., Montmerle, T., Broos, P., Chu, Y.-H. Garmire, G., & Getman,K. 2005, X-Ray and Radio Connections, eds. L.O. Sjouwerman & K.K Dyer (NRAO:Santa Fe, NM)Uzpen, B. et al. 2005, ApJ, 629, 512van Buren, D. & McCray, R. 1988, ApJ, 329, 93 17 –van Buren, D., Mac Low, M. M., Wood, D. O. S., & Churchwell, E. 1990, ApJ, 353, 570van Buren, D., Noriega-Crespo, A., & Dgani, R. 1995, AJ, 110, 2914Vink, J. S., de Koter, A., & Lamers, H. J. G. L. M. 2001, A&A, 369, 574Watson, C., Povich, M. S., Churchwell, E. B., Babler, B. L., Chunev, G. Hoare, M., Indebe-touw, R., Meade, M. R., & Whitney, B. A. 2008, ApJ, 681, 1341 (astro-ph 0806.0609)Whitney, B. A., Wood, K., Bjorkman, J. E., & Wolff, M. J. 2003, ApJ, 591, 1049Wood, K., Whitney, B. A., Robitaille, T. P., & Draine, B. T. 2008, ApJ, in press (astro-ph0807.2398)