The Structure of a Low-Metallicity Giant Molecular Cloud Complex
A. K. Leroy, A. Bolatto, C. Bot, C. W. Engelbracht, K. Gordon, F. P. Israel, M. Rubio, K. Sandstrom, S. Stanimirović
aa r X i v : . [ a s t r o - ph . C O ] J u l Accepted for publication in
The Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 08/22/09
THE STRUCTURE OF A LOW-METALLICITY GIANT MOLECULAR CLOUD COMPLEX
Adam K. Leroy , Alberto Bolatto , Caroline Bot , Charles W. Engelbracht , Karl Gordon , Frank P.Israel , M´onica Rubio , Karin Sandstrom , and Sneˇzana Stanimirovi´c Accepted for publication in
The Astrophysical Journal
ABSTRACTTo understand the impact of low metallicities on giant molecular cloud (GMC) structure, we comparefar infrared dust emission, CO emission, and dynamics in the star-forming complex N83 in the Wing ofthe Small Magellanic Cloud. Dust emission (measured by
Spitzer as part of the S MC and SAGE-SMCsurveys) probes the total gas column independent of molecular line emission and traces shielding fromphotodissociating radiation. We calibrate a method to estimate the dust column using only the high-resolution
Spitzer data and verify that dust traces the ISM in the H I -dominated region around N83.This allows us to resolve the relative structures of H , dust, and CO within a giant molecular cloudcomplex, one of the first times such a measurement has been made in a low-metallicity galaxy. Ourresults support the hypothesis that CO is photodissociated while H self-shields in the outer parts oflow-metallicity GMCs, so that dust/self shielding is the primary factor determining the distributionof CO emission. Four pieces of evidence support this view. First, the CO-to-H conversion factoraveraged over the whole cloud is very high 4–11 × cm − (K km s − ) − , or 20–55 times theGalactic value. Second, the CO-to-H conversion factor varies across the complex, with its lowest(most nearly Galactic) values near the CO peaks. Third, bright CO emission is largely confined toregions of relatively high line-of-sight extinction, A V & masses measured from CO kinematics and dust. Subject headings:
Galaxies: ISM — (galaxies:) Magellanic Clouds — infrared: galaxies — (ISM:)dust, extinction — ISM: clouds — stars: formation INTRODUCTION
Most star formation takes place in giant molecularclouds (GMCs). A quantitative understanding of how lo-cal conditions affect the structure and evolution of theseclouds is key to link conditions in the interstellar medium(ISM) to stellar output. Achieving such an understand-ing is unfortunately complicated by the fact that H doesnot readily emit under the conditions inside a typicalGMC. Astronomers therefore rely on indirect tracers ofH , most commonly CO line emission and dust absorp-tion or emission. These tracers are also affected by envi-ronment, so that assessing the impact of local conditionson GMC structure requires disentangling the effect ofthese conditions on the adopted tracer from their effecton the underlying distribution of H .One way around this problem is to use several in-dependent methods to measure the structure of GMCsin extreme environments, inferring the state of H by Max-Planck-Institut f¨ur Astronomie, D-69117 Heidelberg, Ger-many Department of Astronomy, University of Maryland, CollegePark, MD 20742 UMR 7550, Observatoire Astronomiques de Strasbourg, Uni-versite Louis Pasteur, F-67000 Strasbourg, France Steward Observatory, University of Arizona, Tucson, AZ 85721 Space Telescope Science Institute, 3700 San Martin Drive, Bal-timore, MD 21218, USA Sterrewacht Leiden, Leiden University, PO Box 9513, 2300 RALeiden Departamento de Astronom´ıa, Universidad de Chile, Casilla36-D Department of Astronomy and Radio Astronomy Laboratory,University of California, Berkeley, CA 94720 Astronomy Department, University of Wisconsin, Madison,475 N. Charter St., WI 53711, USA comparing the results. Here we apply this approachto an active star-forming region in the Small Magel-lanic Cloud (SMC). Using far infrared (FIR) emissionmeasured by the
Spitzer
Survey of the SMC (S MCBolatto et al. 2007) and SAGE-SMC (“Surveying theAgents of a Galaxy’s Evolution in the SMC”, Gordon etal. in prep.), we derive the distribution of dust in the re-gion. We compare this to maps of CO and H I line emis-sion (Bolatto et al. 2003; Stanimirovic et al. 1999). Dusttraces the total gas distribution — of which the atomiccomponent is already known — and offers a probe ofshielding from dissociating UV radiation. CO is the mostcommon molecule after H (and the most commonly usedtracer of molecular gas); understanding its relation to H in extreme environments is a long-standing goal. TheCO line also carries kinematic information that allowsdynamical estimates of cloud masses.The SMC is of particular interest because the ISM indwarf irregular galaxies like the SMC contrast sharplywith that of the Milky Way. They have low metallic-ities (e.g., Lee et al. 2006), correspondingly low dust-to-gas ratios (e.g., Issa et al. 1990; Walter et al. 2007),and intense radiation fields (e.g., Madden et al. 2006).These factors should affect the formation and structureof GMCs (e.g., Maloney & Black 1988; Elmegreen 1989;McKee 1989; Papadopoulos et al. 2002; Pelupessy et al.2006). Unfortunately, it has proved extremely challeng-ing to unambiguously observe such effects because theinferred structure of GMCs depends sensitively on themethod used to trace H .Virial mass calculations reveal few differences betweenGMCs in dwarf galaxies and those in the Milky Way. Leroy et al.In this approach, one uses molecular line emission tomeasure the size and line width of a GMC. By as-suming a density profile and virial equilibrium, onecan estimate the dynamical mass of the cloud indepen-dent of its luminosity. Recent studies find the ratioof virial mass to luminosity for GMCs in other galax-ies to be very similar to that observed in the MilkyWay (Walter et al. 2001, 2002; Rosolowsky et al. 2003;Bolatto et al. 2003; Israel et al. 2003; Leroy et al. 2006;Blitz et al. 2007; Bolatto et al. 2008). Further, the scal-ing relations among GMC size, line width, and luminos-ity found in the Milky Way (Larson 1981; Solomon et al.1987; Heyer et al. 2008) seem to approximately apply toresolved CO emission in other galaxies, even dwarf galax-ies (Bolatto et al. 2008).By contrast, observations of low metallicity galaxiesthat do not depend on molecular line emission consis-tently suggest large reservoirs of H untraced by CO(e.g., Israel 1997b; Madden et al. 1997; Pak et al. 1998;Boselli et al. 2002; Galliano et al. 2003; Rubio et al.2004; Leroy et al. 2007; Bot et al. 2007). The most com-mon manifestation of this is an “excess” at FIR or sub-millimeter wavelengths with the following sense: towardsmolecular peaks, there is more dust emission than onewould expect given the gas column estimated from H I + CO. Israel (1997b) treated the abundance of H as anunknown and used this excess to solve for the CO-to-H conversion factor. He found it to depend strongly onboth metallicity and radiation field.These two sets of observations may be reconciled ifCO is selectively photodissociated in the outer parts oflow-metallicity GMCs (e.g. Maloney & Black 1988; Israel1988; Bolatto et al. 1999), a scenario discussed specifi-cally for the SMC by Israel et al. (1986) and Rubio et al.(1991, 1993a). This might be expected if H readily self-shields while CO is shielded from photodissociating ra-diation mostly by dust, which is less abundant at lowmetallicities. In this case, CO emission would trace onlythe inner parts of low-metallicity GMCs.Observations of the Magellanic Clouds as part of theSwedish-ESO Submillimeter Telescope (SEST) Key Pro-gramme (Israel et al. 1993) support this idea: the surfacebrightness of CO is very low in the SMC (Rubio et al.1991); SMC clouds tend to be smaller than their MilkyWay counterparts, with little associated diffuse emission(Rubio et al. 1993a; Israel et al. 2003); and the ratio of CO to CO emission is lower in the Magellanic Cloudsthan in the Galaxy, suggesting that clouds are morenearly optically thin (Israel et al. 2003).The SEST results are mainly indirect evidence. Whatis still needed is a direct, resolved comparison betweenCO, dust, and H . Because dust emission offers a tracerof the total gas distribution that is independent of molec-ular line emission (Thronson et al. 1987, 1988; Thronson1988; Israel 1997b), it allows such a test. If GMCs at lowmetallicity include envelopes of CO-free H , then the dis-tribution of dust (after subtracting the dust associatedwith H I ) should be extended relative to CO emission.Leroy et al. (2007) attempted this measure-ment. They combined S MC with IRIS data(Miville-Deschˆenes & Lagache 2005) to derive thedistribution of dust and compared this to the NANTENCO survey by Mizuno et al. (2001). They derived a distribution of H ∼ . in the SMC may liein envelopes surrounding the CO peaks. The resolutionof the CO and IRIS data limited this comparison toscales of &
45 pc. SMC GMCs are often much smallerthan this (e.g. Rubio et al. 1993a; Mizuno et al. 2001;Israel et al. 2003). Therefore while this measurementindicated that SMC GMC complexes may be immersedin a sea of CO-free cold gas, it was not yet a truecomparison of dust and CO on the scales of individualGMCs.Here, we focus on a single region, N83/N84 (hereaftersimply N83). This isolated star-forming complex lies inthe eastern Wing of the SMC and harbors ∼
10% ofthat galaxy’s total CO luminosity (Mizuno et al. 2001).Combining FIR, CO, and H I data we attempt to answerfollowing questions:1. What is the CO-to-H conversion factor, X CO (i.e.,the ratio of H column density to CO intensityalong a line of sight) in this region?2. Is there evidence that CO is less abundant relativeto H (i.e., that X CO is higher or that there is H without associated CO) in the outer parts of thecloud?3. Is the distribution of CO consistent with dustshielding playing a key role in its survival?4. Can dynamical masses measured from CO kine-matics be brought into agreement with H massesestimated from dust? What is the implied distri-bution of H ?To meet these goals, we first estimate the dust opti-cal depth at 160 µ m, τ ( § τ traces H I column density in the (assumed) H I -dominated ISM near N83, make a self-consistent deter-mination of the dust-to-gas ratio, and then combine τ with the measured H I column density to estimate the H column density in the star forming region ( § τ and H with COand H I data to answer the questions posed above ( § DATA
We use FIR imaging from two
Spitzer surveys. S MCmapped 70 and 160 µ m emission from most active starforming regions in the SMC, including N 83. More re-cently, SAGE-SMC observed a much larger area, includ-ing the Magellanic Bridge and nearby emission-free re-gions. We use a combination of these data sets carriedout by Gordon et al. (in prep.) that dramatically im-proves the quality of the 70 µ m image compared to S MCalone, thus enabling this analysis. At 36 ′′ resolution, thenoise (1 σ ) in the Spitzer maps is σ = 0 .
13 MJy ster − (70 µ m) and σ = 0 . − (160 µ m) in the neigh-borhood of N83.We compare the Spitzer data to the IRIS 100 µ m image.IRIS is a re-processing of the IRAS data carried out byMiville-Deschˆenes & Lagache (2005). These data have ∼ . ′ resolution.Bolatto et al. (2003) used SEST to map CO J = 2 → J = 1 → ′′ ( J = 2 →
1) and 45 ′′ tructure Of A Low Metallicity GMC 3 Fig. 1.— H I (left) and FIR emission at 70 (middle) and 160 µ m (right) in a two degree wide field centered on N 83. A thin black contouroutlines the region where we can clearly distinguish FIR emission from the background (see § ( J = 1 → ′′ ( J = 2 →
1) and 55 ′′ ( J = 1 → σ is typically 0 .
16 K km s − (CO J = 2 →
1) and0 .
22 K km s − (CO J = 1 → I ′′ and sensitivity sufficient to detect H I emis-sion along every line of sight within ∼ ◦ of N83. Wecorrect for H I optical depth and self-absorption follow-ing Stanimirovic et al. (1999, their Equation 6) based onthe H I absorption study by Dickey et al. (2000). Themaximum correction factor near N83 is ∼ . § I from Br¨uns et al. (2005). Galactic H I isdistinguished from SMC gas by its radial velocity. Thesedata have a resolution of 14 ′ .We move all data to three astrometric grids: one cov-ering the entire SMC, a two degree wide field surround-ing N 83 (Figure 1), and the SEST field. In the SESTfield, we use the kernels of Gordon et al. (2008b) to placethe 70 µ m image at the 160 µ m resolution ( ∼ ′′ ), whichmatches that of the SEST CO J = 2 → ′′ )well. We also convolve the 70 and 160 µ m maps to the55 ′′ resolution of the SEST CO J = 1 → µ m data at the 98 ′′ resolution of theH I . Over the whole SMC, we degrade the 70 and 160 µ mimages to the 4 . ′ IRIS resolution.
Additional Processing of the FIR Maps
For consistency among the 70, 100, and 160 µ m data,we move flux densities at 70 and 160 µ m from the MIPSscale (which assumes F ν ∝ ν across the bandpass) tothe IRAS scale (which assumes F ν ∝ ν − ). We do soby dividing the 70 map by 0 .
918 and the 160 µ m map by0 . µ m maps. We estimate this from GalacticH I assuming the average cirrus dust properties measuredby (Boulanger et al. 1996). At 100 µ m we use their fit di- rectly; at 160 µ m we interpolate their fits assuming a typ-ical cirrus dust temperature ( T = 17 . β = 2).To refine the foreground subtraction, we assume thatH I and infrared intensity from the SMC are correlatedat a basic level. As the column density of SMC H I ap-proaches 0, we expect the IR intensity of the SMC toalso approach 0. Therefore, we adjust the zero point ofthe IR maps using a fit of IR intensity to N (H I ) SMC where N (H I ) SMC < × cm − (we subtract the fit-ted y -intercept). This leads us to add 0 . − at 70 µ m, subtract 4 . − at 160 µ m, and sub-tract 0 . − from the IRIS 100 µ m map. Theseoffsets are a natural consequence of the uncertainty inthe reduction and foreground subtraction (which mustremove zodiacal light, Milky Way cirrus, and any cosmicinfrared background). Deviations from the average cir-rus properties are particularly common, being observednear a number of galaxies by Bot et al. (2009).Based on carrying out this exercise in several differentways, we estimate the zero level of our maps to be un-certain by 0 .
25 MJy ster − at 70 µ m and 1 MJy ster − at 160 µ m. We take these uncertainties into account inour calculations ( § I > . − and I > − after the foreground subtraction(i.e., twice the uncertainty in the background). A Word on Resolution
In the rest of this paper we will combine the data de-scribed above in several ways. Two of these combinationslead to maps combining data with different resolutions.We comment on these here and the reader may wish torefer back to this section while reading the paper.First, we subtract a foreground component measuredat 14 ′ resolution from IR maps with 4 . ′ and ∼ ′′ (160 µ m) resolution. Any small scale variation in theMilky Way cirrus will therefore be left in our maps. Thisis only a concern in the diffuse region of the Wing (andso only in § &
10 times higher than the foreground,so variations in the foreground are not a concern.Second, when estimating the distribution of H in N83, Leroy et al.we derive the total amount of hydrogen (H I + H ) alonga line of sight and then subtract the measured H I col-umn density. The total amount of hydrogen is based onFIR dust emission, measured at 36 ′′ resolution (or 55 ′′ resolution when we compare to the SEST CO J = 1 → I column density is measured at 98 ′′ reso-lution. We assume it to be smooth on smaller scales, anassumption born out to some degree by the reasonablecorrelation that we find between H and CO. Nonethe-less, the detailed distribution of H on scales less than98 ′′ ( ∼
29 pc) is somewhat uncertain. DUST TREATMENT
We use the optical depth at 160 µ m, τ , as a proxyfor the amount of dust along a line of sight. For anoptically thin population of grains with an equilibriumtemperature T dust , τ is related to the measured 160 µ mintensity, I , by τ = I B ν ( T dust , µ m) . (1)Here B ν ( T dust , λ ) is the intensity of a blackbody of tem-perature T dust at wavelength λ .Calculating τ thus requires estimating T dust . Be-cause only the 70 and 160 µ m maps have angular res-olution appropriate to compare with CO, we must doso using this combination. Unfortunately, I /I doesnot trivially map to T dust because the 70 µ m band in-cludes non-equilibrium emission from small grains (e.g.,Desert et al. 1990; Draine & Li 2007; Bernard et al.2008). We therefore take an indirect approach: we as-sume that most of the dust mass resides in large grainswith equilibrium temperature T dust that contribute all ofthe emission at 100 µ m and 160 µ m. We use I /I toestimate I /I and then solve for T dust from I I = (cid:18) (cid:19) − . B ν ( T dust , µ m) B ν ( T dust , µ m) , (2)which assumes that dust has a wavelength-dependentemissivity such that τ λ ∝ λ − β with β = 1 . I /I and I /I at the 4 . ′ resolution of IRIS, where both colorsare known and exhibit a roughly 1-to-1 relation. We thenassume this relationship to apply to the smaller ( ∼ ′′ )angular scales measured only by the Spitzer data. NearN83, the two colors are related by: I I = 0 . x + 0 . x + 0 . , where x = I I . (3)Note that this is not a general relation. It does not gothrough the origin and is only 1-to-1 over a limited rangeof I /I ; we fit and apply over the range I /I ∼ .
15 – 1 .
2, where it is a good description of the SMC.
Motivation
In assuming that I /I traces T dust or itsmore sophisticated analogs (e.g., Dale & Helou 2002;Draine & Li 2007), we follow several recent studies of theMagellanic Clouds (Bot et al. 2004; Leroy et al. 2007;Bernard et al. 2008; Gordon et al. 2008a). Schnee et al.(2005, 2006, 2008) have demonstrated that a similar Fig. 2.—
FIR color-color plot for the SMC. The x -axis shows I /I , which is measured by Spitzer at high resolution butincludes contamination by non-equilibrium emission. The y -axisshows I /I , which is only available at 4 . ′ resolution but ismore likely to trace exclusively equilibrium emission. Horizontaldashed lines show the temperatures associated with a few values ofthis color. Shaded contours show the distribution of data for thewhole SMC; the lowest contour includes all data and the contourincrement is a fact of 4 in data density. Black circles show meanand 1 σ scatter for data in a 2 ◦ field centered on N83 (binned by I /I ). The histogram above the plot shows the distribution of I /I over the SEST field (i.e., N83 itself). The dashed curveshows the color-color relation expected for a modified blackbody,which is not a good description of the SMC. On the other hand, thesolid and dash-dotted lines, which show the relations that we usein our analysis, can reasonably predict I /I from I /I . approach reproduces optical and near-IR extinction inGalactic molecular clouds, though with some systematicuncertainties.Figure 2 motivates our use of I /I ( x -axis) to pre-dict I /I ( y -axis). Gray contours show the distri-bution of data for the whole SMC. Bins (filled circles)show data from a 2 ◦ square field centered on N83 (i.e.,Figure 1). Both near N83 and over the whole SMC, thetwo colors show a reasonable correlation (rank correla-tion coefficient 0 . ad hoc treatment of theconversion between I /I and I /I . A single mod-ified blackbody (the dashed line shows one with β = 1 . µ m. The simplest explanation is that I /I traces T dust , while the 70 µ m band includes substantial non-equilibrium emission. We tested the possibility of usingthe models of Draine & Li (2007), which include the ef-fects of stochastic heating, to directly derive dust massesfrom I /I . However, the currently available “SMC”models cannot reproduce the data in Figure 2. Bot et al.(2004) and Bernard et al. (2008) showed that a simi-lar case holds for the Desert et al. (1990) models. Themain stumbling block is reproducing the observed 60 µ m(Desert et al. 1990) or 70 µ m (Draine & Li 2007) emis-sion.Equation 3 is not a unique description. A simple alter-native is a modified blackbody with twice the expectedemission at 70 µ m. In this case: I I = 2 . × (cid:18) (cid:19) − . B ν ( T dust , µ m) B ν ( T dust , µ m) , (4)This is shown by the dash-dotted line in Figure 2. Itreproduces the data near N83 with about the sametructure Of A Low Metallicity GMC 5accuracy as Equation 3. If equilibrium emission sets I /I , then Equation 4 implies that other processes(e.g., single-photon heating of small grains) contribute ≈
50% of the emission at 70 µ m near N83 (and across thewhole SMC). This is in reasonable agreement with theresults for the Solar Neighborhood and several nearbyGMCs (Desert et al. 1990; Schnee et al. 2005, 2008).The aim of this paper is not to investigate the details ofsmall grain heating in the SMC, so we move forward us-ing our empirical fit (Equation 3). This appears as a solidline in Figure 2. It is a good match to the data near N83,where the RMS scatter in the color of individual pixelsabout the fit is ≈ .
04. In deriving uncertainties we useEquation 4 as an equally valid alternative to Equation 3.To convert from I /I to T dust we assume that theSED along each line of sight is described by a modi-fied blackbody with τ λ ∝ λ − β . At long wavelengths( λ & µ m), a blackbody spectrum with a wavelength–dependent emissivity is indeed a good description ofthe integrated SED of the SMC (Aguirre et al. 2003;Wilke et al. 2004; Leroy et al. 2007). We take β = 1 . λ ∼ µ m. Thisis not strongly preferred, and so we allow β from 1 . . Uncertainties in τ We assess the uncertainty in τ by repeatedly addingrealistic noise to our 70 and 160 µ m data and then deriv-ing τ under varying assumptions. For each realization,we offset the observed 70 and 160 µ m maps by a randomamount to reflect uncertainty in the background subtrac-tion; these offsets are drawn from normal distributionswith 1 σ = 0 .
25 MJy ster − at 70 µ m and 1 MJy ster − at 160 µ m. We add normally distributed noise to eachmap. This noise has amplitude equal to the measurednoise ( §
2) and is correlated on scales of 36 ′′ .We derive I /I for each realization using either thepolynomial fit (Equation 3) or scaling the 70 µ m intensity(Equation 4), with equal probability of each. We addnormally distributed noise to I /I with 1 σ = 0 . T dust assuming β anywhere from 1.0 to 2.0 withequal probability.This entire process is repeated 1,000 times. We usethe distribution of Monte Carlo τ s for each pixel toestimate a realistic uncertainty, finding individual mea-surements to be uncertain by ≈
40% (1 σ ). We extendthe same approach through our derivation of N (cid:0) H FIR2 (cid:1) in § τ . τ and Extinction It will be useful to make an approximate assessmentof the dust column in terms of V -band line-of-sight ex-tinction, A V , and reddening, E ( B − V ). In the SolarNeighborhood, E ( B − V ) = N (H) / . × cm − (Bohlin et al. 1978) and τ = 2 . × − cm N (H I )(Boulanger et al. 1996, studying the Galactic cirruswhere we may safely assume that N (H) ≈ N (H I )). Then E ( B − V ) [mag] ≈ τ . (5)The reddening law in the SMC yields R V ≈ . A V [mag] = 1910 τ (6)These equations assume the emissivity, τ /E ( B − V ),of Galactic H I but do not depend on the specific dust-to-gas ratio.Estimates of A V and E ( B − V ) based on τ andEquations 5 and 6 agree well with optical- and UV-based measurements. Caplan et al. (1996) compiled A V for a number of SMC H II regions, including N83 andN84A/B (both of which lie within the SEST field). To-wards N83 they find A V in the range 0 . .
79 mag(mean 0 .
63 mag); towards N84A/B they found A V from0 . .
60 mag (mean 0 .
37 mag). Using their positionsand aperture sizes, we derive A V = 1 . ± .
36 magand 0 . ± .
26 mag for the same regions. The opticaland UV measurements are based on absorption towardsources inside the SMC. Therefore they will sample halfthe total line-of-sight extinction on average. Account-ing for this, our FIR-based extinction estimates are inexcellent agreement with optical values. We find thesame good agreement for Sk 159, a B star near N83towards which Fitzpatrick (1984) and Tumlinson et al.(2002) measured E ( B − V ) ≈ .
05 mag, while we esti-mate E ( B − V ) = 0 . ± .
03 mag (see § DUST AND GAS NEAR N83
Following the method described in §
3, we calculate τ over every line of sight in a 2 ◦ field centered on N83(Figure 1) and in the SEST field. In the process, wederive a median T dust = 20 . ± .
5. This agrees withthe T = 22 ± T dust = 22 . ± . ∼ ± II regions.Our goal in this section is to combine τ with themeasured N (H I ) to estimate N (H ) via N (H FIR2 ) = 12 (cid:16) τ DGR − N (H I ) (cid:17) . (7)Here DGR is the dust-to-gas ratio defined by τ = DGR N (H) (cid:2) cm − (cid:3) , (8) N (H) = N (H I ) + N (H ), and H FIR2 refers to the dis-tribution of H derived using this approach. To calcu-late H FIR2 , we first compare τ and N (H I ) in the areaaround N83 where the ISM is likely to be mostly H I ( § τ effectively traces theISM and allows us to directly measure DGR in the dif-fuse ISM. We show that residuals about this τ - N (H I )relation come exclusively from regions of active star for-mation ( § DGR in N83 itself and estimate N (H ) across the com-plex. H I and Dust Near N83 Leroy et al.
Fig. 3.—
Dust column, traced by τ , ( y -axis) as a function ofH I column density ( x -axis) in a 2 ◦ field centered on N83. Blackcircles show average τ and 1 σ variation in bins 5 × cm − wide. The dashed line shows the median ratio τ /N (H I ) = 1 . × − cm . Dotted lines show the Hi column density for which 50and 75% of the pixels are well above the background (see § τ mostly coincide with N83 and other sites of active star formation. In Figure 3, we plot τ as a function of N (H I ) overthe 2 ◦ field centered on N83. Most of the data are well-described by τ = 1 . +0 . − . × − cm N (H I ) (cid:2) cm − (cid:3) , (9)which is shown by the dashed line in Figure 3. We expectthat N (H) ≈ N (H I ) over most of this area. Thus, theclear, linear correlation in Figure 3 demonstrates that τ traces the ISM well here and the slope is an estimateof the DGR in the diffuse ISM of the SMC Wing.Equation 9 is consistent within the uncertaintieswith results of Bot et al. (2004), who found τ ∼ (1 . ± . × − cm N (H I ) (cid:2) cm − (cid:3) for the wholeWing (after adjusting for slight differences in T dust , β , and λ ). In the Solar Neighborhood, τ ≈ . × − cm N (H I ) (cid:2) cm − (cid:3) (Boulanger et al. 1996).Comparing this to Equation 9 implies that the DGR nearN83 is 17 +10 − times smaller than the Galactic value. Thisagrees within the uncertainties with the DGR found forthe SMC Wing by Leroy et al. (2007), which is ≈ +10 − lower than Galactic .From Equations 9 and 5, we estimate N (H) /E ( B − V ) ≈ +6 − × cm − mag − . This matches the SMC–average N (H) /E ( B − V ) ≈ . × cm − mag − mea-sured by Fitzpatrick (1985) using IUE and confirmed byTumlinson et al. (2002) with FUSE. Residuals About the τ - H I Relation Leroy et al. (2007) made no correction for H I opacity. Doingso would improve the agreement with the present measurement. Equation 9 and Figure 3 demonstrate that a single
DGR describes the region near N83 well. The notableexceptions are a small number of points with high τ relative to their H I column density. In Figure 4 we showthe distribution of residuals about Equation 9. Contoursindicate where our Monte Carlo uncertainty estimatesyield 85, 98, and 99.9% confidence that the residuals arereally greater than zero.The neighboring panel shows the same confidence con-tours superimposed on an H α image of the region nearN83 (Winkler et al., private communication). The high-est residuals are associated with N83 itself. Other regionswith higher-than-expected τ are also associated withconcentrations of H α emission. H α emission indicatesongoing massive star formation, which in turn suggeststhe presence of H . N83 also has significant CO emission,another signpost of H (Mizuno et al. 2001). If a largeamount of the ISM is H , we expect high residuals aboutEquation 9 even for a fixed DGR . The Dust-to-Gas Ratio in N83
To derive H
FIR2 from Equation 7 over the SEST field,we must know the
DGR in N83 itself. We cannot mea-sure this directly because we do not have an indepen-dent measure of the H column. We might expect DGR in N83 to differ somewhat from that in the surroundingdiffuse gas of the Wing: stars are more likely to formin regions with high
DGR and the denser environmentmay shelter grains from destruction by shocks or lead tograin growth (e.g., Dwek 1998). In addition to our mea-surement of the diffuse ISM, we consider two pieces ofevidence when adopting a
DGR to use in N83: observa-tions of a nearby B star and the metallicity of the N84CH II region. FUSE and IUE Measurements of Sk 159:
From FUSEand IUE absorption measurements, E ( B − V ), N (H ),and N (H I ) are known towards Sk 159, a B0.5 star nearN83 (marked by a star in Figure 4). H is detectedbut the column density is small ( ≈ × cm − ,Andr´e et al. 2004). The reddening associated with theSMC is ≈ .
05 mag (Fitzpatrick 1984; Tumlinson et al.2002), though somewhat uncertain. The H I columnmeasured from absorption along the same line of sightis 2 ± × cm − (Bouchet et al. 1985), roughly halfof the column inferred from 21 cm emission along theline of sight (two kinematically distinct H I componentsare visible in emission towards Sk 159; only one of themis seen in absorption, implying that Sk 159 sits betweenthe two, behind the smaller one). These values imply N (H) /E ( B − V ) ≈ × cm − mag − , or DGR ≈ × − cm . Metallicity of N84C:
Russell & Dopita (1990) mea-sured the nebular metallicity of the N84C H II region,which lies within the SEST field, finding 12 + log O / H =8 .
27, 2–3 times lower than the Solar Neighborhood valueand among the highest for any region the SMC. Trans-lating metallicity into a
DGR is not totally straightfor-ward, because the fraction of heavy elements tied up indust may vary with environment. For a fixed fraction ofheavy elements in dust, one would expect
DGR ∝ Z − .Fits to samples of galaxies yield power law relation-ships ( DGR ∝ Z α ) with indices in the range α = 1–2(e.g., Lisenfeld & Ferrara 1998; Draine et al. 2007). Thistructure Of A Low Metallicity GMC 7 Fig. 4.— ( left ) Residuals about Equation 9, the average relationship between τ and H I in a 2 ◦ field centered on N83. The contoursshow where our Monte Carlo uncertainty estimate yields 85, 98, and 99.9% confidence that the residual is above zero. A thin gray lineshows where we clearly distinguish FIR emission from the background (see § τ than expected from H I and Equation 9.( right ) The same contours plotted on top of H α emission near N83. Regions with high τ residuals are associated with sites of recenthigh-mass star formation. would imply N (H) /E ( B − V ) ∼ × cm − mag − or DGR ∼ × − cm .H I and τ : Equation 9 offers a lower bound onthe
DGR — N83 is extremely unlikely to have a lower
DGR than the surrounding medium ( N (H) /E ( B − V ) ≈ × cm − mag − ) and from absorption work weknow that there is not a pervasive massive molecularcomponent in the SMC. The magnitude of the resid-uals about this equation towards N83 itself also offera weak upper bound on the quantity. If we assume DGR much above 3 times the value in Equation 9then some lines of sight inside the SEST field will havesignificantly negative residuals. If the star-forming re-gion itself is described by a single
DGR , then it mustbe roughly bounded by this value, which translates to N (H) /E ( B − V ) ∼ × cm − mag − . Assumed
DGR in N83:
The relatively high metallic-ity and the measurement towards Sk 159 are balancedagainst our observations of a very low
DGR in thenearby ISM and the requirement that Σ
FIRH2 not be signif-icantly and systematically negative. The former suggest N (H) /E ( B − V ) ∼ × cm − mag − , while thelatter yields N (H) /E ( B − V ) ∼ × cm − mag − .In the remainder of this paper we adopt assume that inN83 itself N (H) /E ( B − V ) ∼ × cm − mag − , whichis intermediate in this range. Then τ = 2 . × − cm N (H) (cid:2) cm − (cid:3) . (10)This is twice the value found in the diffuse gas of the SMCWing (Equation 9) and more similar to that found in theactively star-forming SMC Bar (e.g., Wilke et al. 2004;Leroy et al. 2007). It is roughly consistent with observa-tions of Sk 159 and the metallicity of N84C. This DGR also leads to reasonable agreement between dynamicaland dust masses in the star-forming region ( § H FIR2 in N83
Combining Equations 7 and 10 we estimate N (H FIR2 )from τ and N (H I ). From N (H FIR2 ), we calculate themolecular gas surface density,Σ
FIRH2 (cid:2) M ⊙ pc − (cid:3) = N (H FIR2 )4 . × [cm − ] , (11)which includes a factor of 1.36 to account for helium (after Wilson et al. 1988). At the same time we estimatethe extinction along each line of sight using Equation 6.Carrying out these calculations, we work with N (H I )only in average, because the resolution of the 160 µ mand CO J = 2 → ∼ ′′ , while that of the H I map is 98 ′′ ( § FIRH2 in N83 at the resolution of the SEST CO 2 → → FIR2 imagesruns linearly from Σ
FIRH2 = 100 M ⊙ pc − to 500 M ⊙ pc − .Several systematic uncertainties may affect N (cid:0) H FIR2 (cid:1) but are hard to quantify and so not reflected in our MonteCarlo estimate of the uncertainties. We consider thesein Appendix A, finding no strong reason to doubt thatEquation 7 yields an approximate estimate of N (H ). H FIR2 , CO, DUST, AND DYNAMICS H FIR2 and H I In the rest of the paper, Σ
FIRH2 includes this correction forhelium, while N (H ) or N (H FIR2 ) refer to column density of H alone Leroy et al.
Fig. 5.—
CO emission (left panels) and Σ
FIRH2 estimated using Equation 7 (right panels). The top panels show CO J = 2 → FIR2 at ∼ ′′ resolution (but see § FIR2 map). The bottom panels show CO J = 1 → FIR2 at 55 ′′ resolution. Dotted contours show the boundaries of the SEST map. In the CO maps, contours show I CO from 1 to8 K km s − spaced by 1 K km s − . In the H FIR2 maps, contours indicate Σ
FIRH2 from 100 to 500 M ⊙ pc − spaced by 50 M ⊙ pc − . Before we consider the relationship between CO, H
FIR2 ,and dust within N83, we briefly examine the transitionfrom atomic (H I ) to molecular (H ) gas in the complex.Krumholz et al. (2009) recently considered the transi-tion from H I to H in galaxies. They argue that insidea complex of mixed atomic and molecular gas, the ratioof H to H I along a line of sight ( R H2 = Σ H2 / Σ HI ) ismainly a function of two factors: total gas surface den-sity (Σ HI +Σ H2 ) and metallicity. Their calculations agree well with a variety of observations, including FIR-basedestimates of Σ H2 in the SMC at lower resolution.Comparing H I and H in the area around N83, we in-deed observe a clear relationship between R H2 and thetotal gas surface density. We show this in Figure 6, plot-ting R H2 against Σ HI + Σ FIRH2 over the whole area whereΣ
FIRH2 >
0. We work at the 98 ′′ (29 pc) resolution of theH I map, with each point in the plot showing an indepen-dent measurement. For this analysis, we are interestedtructure Of A Low Metallicity GMC 9 Fig. 6.—
Ratio of molecular to atomic gas, R H2 = Σ FIRH2 / Σ HI , asa function of total gas column density Σ HI + Σ H2 in the N83 com-plex (after removing H I not associated with the complex). Eachpoint shows an independent measurement at 29 pc resolution. Thedotted line shows the sensitivity of our H FIR2 map. We plot thetheoretical relationships between R H2 and Σ HI + Σ H2 calculatedby Krumholz et al. (2009) for several metallicities. in the gas associated with the star-forming complex it-self (not unassociated gas in front of and behind it alongthe line of sight). To remove H I unassociated with N83itself from Σ HI , we subtract the median Σ HI measuredover the area shown in Figure 1 (53 M ⊙ pc − ) from themeasured Σ HI before plotting. This is only an issue forH I ; H FIR2 does not extend beyond the N83 complex.We overplot the relationship between R H2 and Σ HI +Σ H2 predicted by Krumholz et al. (2009) for three metal-licities: Z = 0 .
5, 0 .
33, and 0 .
125 times solar. Ourdata are consistent with the shape of the Krumholz et al.(2009) calculation. We find R H2 = 1 at Σ H2 + Σ HI =68 ±
12 M ⊙ pc − , which agrees well with their calcula-tions for Z II region (Russell & Dopita 1990). However, it is signifi-cantly higher than the DGR that we adopt ( § R H2 = 1 at a significantly higher value ofΣ HI + Σ H2 than in a solar metallicity cloud. CO and H FIR2
Figure 5 shows that the distributions of H
FIR2 andCO share the same peaks and basic morphology. How-ever, the values of I CO in N83 are low compared to aGalactic molecular cloud, which usually show I CO ∼
10 K km s over a large area, not merely the peaks(e.g., Wilson et al. 2005). By contrast, the values ofΣ FIRH2 (mean 180 M ⊙ pc − ) are similar to the surfacedensity of an average Galactic GMC ∼ ⊙ pc − (Solomon et al. 1987; Heyer et al. 2008).This means that CO is faint compared to H FIR2 in N83. Over the SEST field X CO is (cid:10) X → (cid:11) = 6 . +2 . − . × cm − K km s − (12) (cid:10) X → (cid:11) = 7 . +4 . − . × cm − K km s − These ratios are 34 and 40 times the Galac-tic conversion factor, taken to be X Gal ≈ × cm − (K km s − ) − (e.g., Strong & Mattox 1996;Dame et al. 2001). This value agrees reasonably withprevious FIR-based determinations of X CO in the SMCand N83: comparing IRAS and CO at selected point-ings in the SMC, Israel (1997b) derived X CO ∼ X Gal .Applying the same methodology to N83, Bolatto et al.(2003) found X CO ∼ ± X Gal . Leroy et al. (2007)derived X CO ∼ X Gal comparing NANTEN CO, IRIS100 µ m and Spitzer µ m towards N83 (removing theircorrection for extent).The left panel in Figure 7 compares H FIR2 and I CO forindividual lines of sight. We plot Σ FIRH2 as a functionof I CO over the SEST field. We regrid the data so thateach point corresponds to an approximately independentmeasurement over a ∼
10 pc (CO J = 2 →
1) or ∼
17 pc(CO J = 1 →
0) wide box. Gray curves show fixed CO-to-H conversion factors, starting with Galactic (lowest)and increasing by factors of 3.33.As with Figure 5, Figure 7 shows that despite the verylow ratio of CO to H FIR2 , the two exhibit an overall cor-respondence. High I CO coincides with high Σ FIRH2 andthe reverse, so that a rank correlation coefficient of 0 . I CO and Σ FIRH2 does not gothrough the origin. Instead, I CO = 0 corresponds toroughly Σ FIRH2 = 50–150 M ⊙ pc − . This suggests thepresence of an envelope of H FIR2 with very little or no as-sociated CO. Unfortunately, this result is very sensitiveto the adopted
DGR ( § DGR at the upper end of the plausible range, the dataare consistent with no CO-free envelope although COemission is still faint relative to Σ
FIRH2 in the SEST field.If we take
DGR at the value derived in the nearby diffuseISM, the surface density of the envelope is even higher ∼ ⊙ pc − . Although the observation towardsSk 159 does not actually intersect the envelope in thelatter case, it is very nearby and the low N (H ) derivedfrom absorption towards this star offers some circumstan-tial evidence against a very massive extended envelope.The other notable feature of this plot is that at veryhigh Σ FIRH2
CO intensity increases dramatically (the turnto the right at the top of the plot). We see this in bothCO transitions, but the effect is more pronounced at thehigher resolution of the CO J = 2 → FIR2 to CO is lower forthe regions of brightest CO emission, dropping to ∼ FIR2 and CO emission almostcertainly trace different volumes ( § CO and Extinction
Fig. 7.—
The structure of H , CO, and dust in N83. ( left ) H FIR2 surface density, Σ
FIRH2 , as a function CO intensity, I CO ( x -axis). Dottedgray lines show CO-to-H conversion factors of 1, 3.33, 10 ... 333 times the Galactic value. ( right ) I CO ( y -axis) as a function of line-of-sightextinction, A V ( x -axis), estimated from τ . The vertical line shows the line-of-sight extinction from which most CO emission emerges inmodels of SMC molecular clouds by Lequeux et al. (1994). The gray curve shows the relationship between I CO and extinction observed inthe Pipe Nebula (Milky Way) by Lombardi et al. (2006). In both plots, each data point represents an independent line of sight. We showresults for the J = 2 → ′′ resolution) in black and the J = 1 → ′′ resolution) in gray. In §
1, we highlighted the role of dust in shieldingCO from dissociating radiation. This may provide asimple explanation for the upturn in CO intensity athigh Σ
FIRH2 . Lequeux et al. (1994) modeled CO emis-sion in SMC molecular clouds. For their typical cloud( n H ∼ cm − , illuminated by a radiation field 10times the local interstellar radiation field), they foundthat most CO emission comes from a relatively narrowregion of the cloud centered on A V ∼ A V . We esti-mate A V from τ using Equation 6. For comparison,we mark A V ∼ A V = 1 magfor them corresponds to A V ∼ A K into A V using theiradopted A V = A K / . − about this relation.In agreement with Lequeux et al. (1994), we find thatlines of sight with bright CO emission occur almost ex-clusively above A V ∼ A V to test whether I CO is indeed more orless independent of extinction well above this threshold(as in the Milky Way, Lombardi et al. 2006; Pineda et al.2008). In fact, Figure 3a of Lequeux et al. (1994) seemsa close match to what we observe: a shallow slope that steepens sharply around A V of 2 mag (for us). The radi-ation field that they assume, 10 times the Galactic valueis a rough match to what one would infer comparing T dust in N83 (median ∼
23 K, max ∼
28 K) to that of Galacticcirrus (17.5 K) — median ∼
5, maximum ∼ — es-pecially when one recalls that this is integrated over thewhole line of sight rather than tracing the radiation fieldincident on the cloud surface.N83 shows somewhat less CO at a given extinctionthan the Pipe Nebula. This is also in agreement withthe models by Lequeux et al. (1994), which predict thatCO from Milky Way clouds emerges from a broader rangeof A V and lower values of A V than in the SMC. Theyattribute the difference to lower rates of photodissocia-tion and it certainly seems likely that the radiation fieldincident on the H in N83 is much more intense than inthe relatively quiescent Pipe.Small differences should not overshadow the similar-ities between the CO-extinction relation in the MilkyWay and that in the SMC. Compared to the left panel inFigure 7, the right panel actually shows a striking sim-ilarity between Galactic and SMC clouds. We derive aCO-H conversion that differs with the Milky Way bya factor of ∼
30, while the relationship between extinc-tion and CO is only slightly offset. Figure 7 supportsthe hypothesis that shielding, rather than the distribu-tion of H , determines the location of bright CO emis-sion. Here “shielding” refers to a combination of dustand self-shielding. Both processes are important to set-ting the location at which most C is tied up in CO (e.g.,Wolfire et al. 1993) and the effective shielding from bothsources will be weaker in the SMC than in the Galaxydue to the decreased metallicity.Extinction may also be critical to a cloud’s ability to For our adopted β = 1 .
5, the magnitude of the radiation fieldheating the dust is roughly ∝ T . . tructure Of A Low Metallicity GMC 11form stars. McKee (1989) proposed that ionization byan external radiation field plays an important role in set-ting cloud structure because it determines the degree ofmagnetic support. He predicted that clouds forming low-mass stars in equilibrium will self-regulate to achieve in-tegrated line-of-sight extinctions A V ≈ A V ∼ A V ∼ . A V is fairly robust. It does not depend on our choice of DGR , only on the adopted FIR emissivity ( τ FIR /A V ) andreddening law. The most likely biases in the emissivity(e.g., coagulation of small grains) will lower A V , bringingour results into even closer agreement with those in theMilky Way. H FIR2 and Dynamical Mass Estimates
CO line emission also offers kinematic information.This is the basis of the virial mass method commonlyused to estimate the masses of molecular clouds andderive CO-to-H conversion factors (e.g. Rubio et al.1993a; Wilson 1995; Arimoto et al. 1996), including inN83 (Bolatto et al. 2003; Israel et al. 2003; Bolatto et al.2008). The potential pitfall of this approach may be seenfrom § FIR2 (even over matched areas) because thelatter also traces the outer (CO-free) part of the cloud,which exists in front of and behind the CO-emitting re-gion even over matched lines of sight.In N83, we have the advantage of an independent mea-surement of H
FIR2 and observing the CO emission overa range of scales. Here we test whether these observa-tions can be reconciled using a simple model in whichCO emission comes from only the inner part of a largerH cloud (as appears to be the case in N83). We con-sider a spherical cloud with a radially declining density,such that ρ ∝ r − α , and a radius R beyond which ρ = 0,i.e., the model usually adopted (with α = 1) to calculatecloud virial masses (Solomon et al. 1987) . We assumethat the dynamical mass estimated from CO line data We cap the density at its maximum value over the inner 3%of the cloud to avoid divergence. traces the mass of a fraction of this cloud, out to ra-dius r frac . The ratio of dynamical mass to H FIR2 over amatched area, M vir /M FIRH2 is then a function of α andthe ratio of the true radius of the cloud to the radius ofthe area being considered, r frac /R . The top left panel ofFigure 8 shows this ratio for models with α from 0 to2.0.To compare our observations to this model, we mea-sure the line width and radius of CO emission over aseries of scales in N83. We consider intensity contoursin position-position-velocity space, beginning with thebright northwestern region and including progressivelymore of the cloud (but always including that region, seeFigure 8). We estimate the radius and line width of eachregion from the area (for the radius) and second moment(for the line width). To account for the finite resolutionof SEST, the radius of each cloud is adjusted by R = r A cloud π . − R . (13)Here A cloud is the area of the cloud and R beam =0 . F W HM is the “radius” of the beam (Solomon et al.1987). We combine the RMS line width, σ v , and cloudradius, R , to derive the virial mass via M vir = 1040 Rσ v [M ⊙ ] , (14)with σ v in km s − and R in pc. For details of measuringthe properties of extragalactic GMCs from CO emission,we refer the reader to Rosolowsky & Leroy (2006) andreferences therein.For each contour, we measure M vir /M FIRH2 . We com-pare this ratio as a function of R to a range of densityprofiles and cloud radii. The resulting distribution ofreduced χ is shown in the bottom left panel of Figure8. Our measurements, along with the best-fit model areshown in the bottom right panel of the same figure.The best-fit model has ρ ∝ r − . and R = 70 pc,though these numbers are not strongly constrained. The χ = 1 surface spans R = 50 – 140 pc and α = 0 . . — can relate dynamics measured frommolecular line emission and H FIR2 . The best fit radius, R = 70 pc, is quite similar to that needed to achievethe extinction threshold for CO emission ( A V ≈
1) us-ing our adopted
DGR and n ≈
100 cm − — a typicalaverage volume density for Galactic GMCs and perhapsappropriate for the diffuse gas between dense molecularclumps in the SMC. These three numbers combine toyield a depth of ∼
60 pc. Meanwhile, the density pro-file is similar to the α = 1 commonly used to describeGalactic clouds (Solomon et al. 1987).The strong dependence of M vir /M FIRH2 on the size-scale sampled at least partially motivates the discrep-ancy between CO-to-H conversion factors measured us-2 Leroy et al. Fig. 8.—
Reconciling dynamics and H
FIR2 in N83. ( top left ) The ratio of virial mass, M vir , to total H mass, M FIRH2 , ( y -axis) expected forthe simple case where M vir traces only an inner portion of a cloud (the fraction traced is shown on the x -axis). Each line shows a cloudwith a different density profile. ( top right ) N83 divided into concentric regions defined by CO intensity. We measure M vir and M FIRH2 foreach region. ( bottom left ) Results of fitting the models in the top left panel to M vir /M FIRH2 measured from the regions in the top rightpanel. The x -axis show the power law index of the cloud density profile; the y -axis shows the cloud radius. Contours show reduced χ ,starting at 0.5 and increasing by a factor of 2 each step. The white cross marks the best-fit model ( ρ ∝ r − . , R = 70 pc). ( bottom right ) M vir /M FIRH2 as a function of region radius (black points), along with the best fit model (gray line). ing CO observations and those derived from dust. Atthe high resolutions achieved by millimeter-wave inter-ferometers in Local Group galaxies, CO-emitting cloudsare resolved from their surroundings. By concentrat-ing on these clouds, one samples only dense regionswhere CO is well-shielded by dust. This naturally leadsto relatively modest conversion factors. On the otherhand, dust measurements and dynamical measurementsmade on larger scales sample the whole complex. Inthe SMC this appears to includes a large amount ofpoorly-shielded gas and such methods therefore returnsignificantly larger conversion factors. One manifestationof this phenomenon is that dynamical mass determina-tions from CO measurements with larger physical beamsizes often return systematically and significantly higher conversion factors than those obtained from CO mea-surements in much smaller beams (Rubio et al. 1993b;Wilson 1995; Israel 2000; Bolatto et al. 2003). For in-terferometer measurements to properly sample the fullcloud structure a multi-scale analysis, such as that pre-sented here or the more rigorous “dendogram” approachrecently described by Rosolowsky et al. (2008), is neces-sary.Although our dynamical and dust-based results appearconsistent with this simple picture, other recent resultssuggest a more complex relationship between the twomeasurements. Bot et al. (2009, in prep.) recentlymeasured the relationship between sub-millimeter dustemission and CO-based dynamical masses in the south-west part of the SMC Bar. Even after controlling fortructure Of A Low Metallicity GMC 13contamination by an extended superstructure of CO-freeH , they find that virial masses are systematically lowerthan dust-based H masses on the scale of individualCO-bright regions. This might arise if clouds are short-lived (i.e., presently collapsing) or partially supported bymagnetic fields. Alternatively it may reflect altered dustproperties in dense cloud cores. The virial-dust discrep-ancy measured by Bot et al. and the multiscale virial-dust measurements presented here can both be readilyapplied to simulated clouds and multi-tracer observationsof Galactic GMCs. It will be interesting to see whetherthese measurements can be replicated purely by alteringthe CO-emitting surface inside of a cloud (as it appearsfrom our simple model) or if they constrain SMC cloudstructure to be genuinely different from that in the MilkyWay (as appears to be the case from the Bot et al. re-sults). SUMMARY AND DISCUSSION
We combine far infrared emission, CO line emission,and a 21-cm H I map to study the structure of CO, dust,and H in the SMC star forming complex N83.Two recent surveys of the SMC using Spitzer (S MCBolatto et al. 2007, and SAGE-SMC, Gordon et al., inprep.) allow us to estimate the distributions of dust andH at high spatial resolution. We calibrate a method toderive the equilibrium dust temperature, T dust , and op-tical depth at 160 µ m, τ , along the line of sight usingonly Spitzer data. Applying this method and assum-ing that the diffuse ISM of the SMC Wing is mostlyH I , we determine the dust-to-gas ratio ( DGR ) using the τ and H I maps. We find τ to be a good tracer of N (H I ) with τ = 1 . +0 . − . × − cm N (H), implyinga DGR +10 − (1 σ ) times lower than that in the SolarNeighborhood. High residuals about the τ – N (H I )relation come almost exclusively from regions of activestar formation, with the largest residuals from N83 it-self. The most likely origin for these high residuals isdust associated with H , though several important sys-tematic uncertainties remain unconstrained (AppendixA). Considering several pieces of evidence (the metal-licity of the N84C H II region, UV spectra of a nearbystar, and the DGR in nearby diffuse ISM) we adopta
DGR of N (H) /E ( B − V ) ≈ × cm − mag − ( τ = 2 . × − cm N (H)) for N83 itself, but notethis as a significant uncertainty with the plausible rangespanning N (H) /E ( B − V ) = 3–10 × cm − mag − .Combining this DGR with τ and the measured H I distribution, we derive a map of H FIR2 in N83.Comparing CO intensity, kinematics, dust, and H wefind:1. The CO-to-H conversion factor averaged over thepart of the N83/N84 region mapped by SESTis very high, 4–11 × cm − (K km s − ) − or ≈ X CO , there isreasonable agreement between the distributions ofCO and H traced by dust: a rank correlation co-efficient ≈ . , so that X CO varies across the region, with the lowest (most nearly Galactic) values near the CO peaks. Themagnitude (or existence) of an extended, truly CO-free envelope is a sensitive function of the adopted DGR . Our best estimate is that such an enve-lope does exist, with Σ H2 ≈
100 M ⊙ pc − where I CO ∼ τ . Bright CO emis-sion is largely confined to regions with A V & .4. A simple model can reconcile dynamical masses(measured from CO) with H (measured fromdust). In this model, CO emission comes a surfacewithin the cloud while dust emission traces all H along the line of sight. The best-fit density profileand radius are ρ ∝ r − . and R = 70 pc. These arenot strongly constrained, but the density profile issimilar to that inferred for Galactic clouds and theradius is consistent with that required to achieve A V ≈ DGR and a typicalmolecular cloud density.These results — particularly the confinement of in-tense CO to regions of relatively high line-of-sight ex-tinction — are all consistent with the selective pho-todissociation of CO relative to H at low metallicities(e.g., Maloney & Black 1988; Rubio et al. 1993b,a; Israel1997b; Bolatto et al. 1999). In this scenario, the distri-bution of CO emission is largely driven by need for dustto shield CO from dissociating radiation. The underlyingdistribution of H , while subject to significant systematicuncertainties, appears similar to that in a Galactic GMCcomplex.If the distribution of CO emission is indeed largely de-termined by dust shielding, then we expect that the ra-tio of CO emission to H mass will depend sensitivelyon both the local DGR and the radiation field incidenton the cloud. These effects may largely cancel in moremassive spiral galaxies, yielding a CO-to-H conversionfactor that is fairly robust (e.g., Wolfire et al. 1993). Inlow-mass galaxies, which have high radiation fields andlow DGR , they will tend to compound, producing ex-tended envelopes of H with little or no associated CO.From recent large surveys of the MagellanicClouds at infrared and millimeter wavelengths (e.g.,Fukui et al. 1999; Mizuno et al. 2001; Meixner et al.2006; Bolatto et al. 2007; Ott et al. 2008, Gordon etal., in prep.), it will be possible in the next few yearsto fill the right panel in Figure 7 with points fromacross the Clouds. This will allow the quantification ofthe radiation field (and perhaps density) as a “secondparameter” in the I CO - A V relation. It may also allowan improved calibration of X CO as a function of both DGR and local radiation field, extending the pioneeringwork by Israel (1997b) to the scale of individual clouds.Even with such data, it is unclear if CO emission canremain an effective tracer of H on the scale of individualclouds. Tracing local variations in DGR and radiation4 Leroy et al.field to apply a spatially variable X CO may not be pos-sible or practical. Of course, CO is already well-knownto be a flawed tracer of H within Galactic clouds (e.g.,Pineda et al. 2008) but retains significant utility for trac-ing H on large scales. Over a sizable portion of a galaxy,variations in the radiation field and DGR may averageout and allow a calibration to work at a basic level. Giventhat the options to trace H in low-metallicity galaxiesremain limited, a combination of dust and molecular lineemission is likely to be the only widely available optionin the near future. Herschel spectroscopy of the [CII] line and
Fermi observations of γ ray emission from the Mag-ellanic Clouds, while both likely to illuminate the issuesignificantly, will only target a small sample of galaxies.We thank the anonymous referee for a detailed andhelpful critique. We thank Henrik Beuther for helpfulcomments on a draft of the paper and Mark Krumholz fora helpful discussion. We acknowledge the use of NASA’sAstrophysics Data System Bibliographic Services. REFERENCESAbergel, A., Boulanger, F., Mizuno, A., & Fukui, Y. 1994, ApJ,423, L59Aguirre, J. E., Bezaire, J. J., Cheng, E. S., Cottingham, D. A.,Cordone, S. S., Crawford, T. M., Fixsen, D. J., Knox, L.,Meyer, S. S., Norgaard-Nielsen, H. U., Silverberg, R. F.,Timbie, P., & Wilson, G. W. 2003, ApJ, 596, 273Andr´e, M. K., Le Petit, F., Sonnentrucker, P., Ferlet, R., Roueff,E., Civeit, T., D´esert, J.-. M., Lacour, S., & Vidal-Madjar, A.2004, A&A, 422, 483Arce, H. G., & Goodman, A. A. 1999, ApJ, 512, L135Arimoto, N., Sofue, Y., & Tsujimoto, T. 1996, PASJ, 48, 275Bell, T. A., Roueff, E., Viti, S., & Williams, D. A. 2006, MNRAS,371, 1865Bernard, J. P., Abergel, A., Ristorcelli, I., Pajot, F., Torre, J. P.,Boulanger, F., Giard, M., Lagache, G., Serra, G., Lamarre,J. M., Puget, J. L., Lepeintre, F., & Cambr´esy, L. 1999, A&A,347, 640Bernard, J.-P., Reach, W. T., Paradis, D., Meixner, M., Paladini,R., Kawamura, A., Onishi, T., Vijh, U., Gordon, K.,Indebetouw, R., Hora, J. L., Whitney, B., Blum, R., Meade, M.,Babler, B., Churchwell, E. B., Engelbracht, C. W., For, B.-Q.,Misselt, K., Leitherer, C., Cohen, M., Boulanger, F., Frogel,J. A., Fukui, Y., Gallagher, J., Gorjian, V., Harris, J., Kelly,D., Latter, W. B., Madden, S., Markwick-Kemper, C., Mizuno,A., Mizuno, N., Mould, J., Nota, A., Oey, M. S., Olsen, K.,Panagia, N., Perez-Gonzalez, P., Shibai, H., Sato, S., Smith, L.,Staveley-Smith, L., Tielens, A. G. G. M., Ueta, T., Van Dyk,S., Volk, K., Werner, M., & Zaritsky, D. 2008, AJ, 136, 919Blitz, L., Fukui, Y., Kawamura, A., Leroy, A., Mizuno, N., &Rosolowsky, E. 2007, in Protostars and Planets V, ed.B. Reipurth, D. Jewitt, & K. Keil, 81–96Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978, ApJ, 224, 132Bolatto, A. D., Jackson, J. M., & Ingalls, J. G. 1999, ApJ, 513,275Bolatto, A. D., Leroy, A., Israel, F. P., & Jackson, J. M. 2003,ApJ, 595, 167Bolatto, A. D., Leroy, A. K., Rosolowsky, E., Walter, F., & Blitz,L. 2008, ApJ, 686, 948Bolatto, A. D., Simon, J. D., Stanimirovi´c, S., van Loon, J. T.,Shah, R. Y., Venn, K., Leroy, A. K., Sandstrom, K., Jackson,J. M., Israel, F. P., Li, A., Staveley-Smith, L., Bot, C.,Boulanger, F., & Rubio, M. 2007, ApJ, 655, 212Boselli, A., Lequeux, J., & Gavazzi, G. 2002, A&A, 384, 33Bot, C., Boulanger, F., Lagache, G., Cambr´esy, L., & Egret, D.2004, A&A, 423, 567Bot, C., Boulanger, F., Rubio, M., & Rantakyro, F. 2007, A&A,471, 103Bot, C., Helou, G., Boulanger, F., Lagache, G., Miville-Deschenes,M.-A., Draine, B., & Martin, P. 2009, ArXiv e-printsBouchet, P., Lequeux, J., Maurice, E., Prevot, L., &Prevot-Burnichon, M. L. 1985, A&A, 149, 330Boulanger, F., Abergel, A., Bernard, J.-P., Burton, W. B.,Desert, F.-X., Hartmann, D., Lagache, G., & Puget, J.-L. 1996,A&A, 312, 256Boulanger, F., Bronfman, L., Dame, T. M., & Thaddeus, P. 1998,A&A, 332, 273Br¨uns, C., Kerp, J., Staveley-Smith, L., Mebold, U., Putman,M. E., Haynes, R. F., Kalberla, P. M. W., Muller, E., &Filipovic, M. D. 2005, A&A, 432, 45Cambr´esy, L., Boulanger, F., Lagache, G., & Stepnik, B. 2001,A&A, 375, 999 Cambr´esy, L., Jarrett, T. H., & Beichman, C. A. 2005, A&A, 435,131Caplan, J., Ye, T., Deharveng, L., Turtle, A. J., & Kennicutt,R. C. 1996, A&A, 307, 403Dale, D. A., & Helou, G. 2002, ApJ, 576, 159Dame, T. M., Hartmann, D., & Thaddeus, P. 2001, ApJ, 547, 792Desert, F.-X., Boulanger, F., & Puget, J. L. 1990, A&A, 237, 215Dickey, J. M., Mebold, U., Marx, M., Amy, S., Haynes, R. F., &Wilson, W. 1994, A&A, 289, 357Dickey, J. M., Mebold, U., Stanimirovi´c, S., & Staveley-Smith, L.2000, ApJ, 536, 756Draine, B. T., Dale, D. A., Bendo, G., Gordon, K. D., Smith,J. D. T., Armus, L., Engelbracht, C. W., Helou, G., Kennicutt,Jr., R. C., Li, A., Roussel, H., Walter, F., Calzetti, D.,Moustakas, J., Murphy, E. J., Rieke, G. H., Bot, C.,Hollenbach, D. J., Sheth, K., & Teplitz, H. I. 2007, ApJ, 663,866Draine, B. T., & Lee, H. M. 1984, ApJ, 285, 89Draine, B. T., & Li, A. 2007, ApJ, 657, 810Dutra, C. M., Ahumada, A. V., Clari´a, J. J., Bica, E., & Barbuy,B. 2003, A&A, 408, 287Dwek, E. 1997, ApJ, 484, 779—. 1998, ApJ, 501, 643Elmegreen, B. G. 1989, ApJ, 338, 178Fitzpatrick, E. L. 1984, ApJ, 282, 436—. 1985, ApJS, 59, 77Fukui, Y., Mizuno, N., Yamaguchi, R., Mizuno, A., Onishi, T.,Ogawa, H., Yonekura, Y., Kawamura, A., Tachihara, K., Xiao,K., Yamaguchi, N., Hara, A., Hayakawa, T., Kato, S., Abe, R.,Saito, H., Mano, S., Matsunaga, K., Mine, Y., Moriguchi, Y.,Aoyama, H., Asayama, S.-i., Yoshikawa, N., & Rubio, M. 1999,PASJ, 51, 745Galliano, F., Madden, S. C., Jones, A. P., Wilson, C. D., Bernard,J.-P., & Le Peintre, F. 2003, A&A, 407, 159Gordon, K. D., Bot, C., Muller, E., Misselt, K. A., Bolatto, A.,Bernard, J. ., Reach, W., Engelbracht, C. W., Babler, B.,Bracker, S., Block, M., Clayton, G. C., Hora, J., Indebetouw,R., Israel, F. P., Li, A., Madden, S., Meade, M., Meixner, M.,Sewilo, M., Shiao, B., Smith, L. J., van Loon, J. T., &Whitney, B. A. 2008a, ArXiv e-printsGordon, K. D., Clayton, G. C., Misselt, K. A., Landolt, A. U., &Wolff, M. J. 2003, ApJ, 594, 279Gordon, K. D., Engelbracht, C. W., Rieke, G. H., Misselt, K. A.,Smith, J.-D. T., & Kennicutt, Jr., R. C. 2008b, ApJ, 682, 336Heyer, M., Krawczyk, C., Duval, J., & Jackson, J. M. 2008,ArXiv e-printsIsrael, F. 2000, in Molecular Hydrogen in Space, ed. F. Combes &G. Pineau Des Forets, 293–+Israel, F. P. 1988, in Astrophysics and Space Science Library, Vol.147, Millimetre and Submillimetre Astronomy, ed. R. D.Wolstencroft & W. B. Burton, 281–305Israel, F. P. 1997a, A&A, 317, 65—. 1997b, A&A, 328, 471Israel, F. P., de Graauw, T., van de Stadt, H., & de Vries, C. P.1986, ApJ, 303, 186Israel, F. P., Johansson, L. E. B., Lequeux, J., Booth, R. S.,Nyman, L. A., Crane, P., Rubio, M., de Graauw, T., Kutner,M. L., Gredel, R., Boulanger, F., Garay, G., & Westerlund, B.1993, A&A, 276, 25 tructure Of A Low Metallicity GMC 15
Israel, F. P., Johansson, L. E. B., Rubio, M., Garay, G., deGraauw, T., Booth, R. S., Boulanger, F., Kutner, M. L.,Lequeux, J., & Nyman, L.-A. 2003, A&A, 406, 817Issa, M. R., MacLaren, I., & Wolfendale, A. W. 1990, A&A, 236,237Krumholz, M. R., McKee, C. F., & Tumlinson, J. 2009, ApJ, 693,216Larson, R. B. 1981, MNRAS, 194, 809Laureijs, R. J., Clark, F. O., & Prusti, T. 1991, ApJ, 372, 185Lee, H., Skillman, E. D., Cannon, J. M., Jackson, D. C., Gehrz,R. D., Polomski, E. F., & Woodward, C. E. 2006, ApJ, 647, 970Lequeux, J., Le Bourlot, J., Des Forets, G. P., Roueff, E.,Boulanger, F., & Rubio, M. 1994, A&A, 292, 371Leroy, A., Bolatto, A., Stanimirovi´c, S., Mizuno, N., Israel, F., &Bot, C. 2007, ApJ, 658, 1027Leroy, A., Bolatto, A., Walter, F., & Blitz, L. 2006, ApJ, 643, 825Lisenfeld, U., & Ferrara, A. 1998, ApJ, 496, 145Lombardi, M., Alves, J., & Lada, C. J. 2006, A&A, 454, 781Madden, S. C., Galliano, F., Jones, A. P., & Sauvage, M. 2006,A&A, 446, 877Madden, S. C., Poglitsch, A., Geis, N., Stacey, G. J., & Townes,C. H. 1997, ApJ, 483, 200Maloney, P., & Black, J. H. 1988, ApJ, 325, 389McKee, C. F. 1989, ApJ, 345, 782Meixner, M., Gordon, K. D., Indebetouw, R., Hora, J. L.,Whitney, B., Blum, R., Reach, W., Bernard, J.-P., Meade, M.,Babler, B., Engelbracht, C. W., For, B.-Q., Misselt, K., Vijh,U., Leitherer, C., Cohen, M., Churchwell, E. B., Boulanger, F.,Frogel, J. A., Fukui, Y., Gallagher, J., Gorjian, V., Harris, J.,Kelly, D., Kawamura, A., Kim, S., Latter, W. B., Madden, S.,Markwick-Kemper, C., Mizuno, A., Mizuno, N., Mould, J.,Nota, A., Oey, M. S., Olsen, K., Onishi, T., Paladini, R.,Panagia, N., Perez-Gonzalez, P., Shibai, H., Sato, S., Smith, L.,Staveley-Smith, L., Tielens, A. G. G. M., Ueta, T., Dyk, S. V.,Volk, K., Werner, M., & Zaritsky, D. 2006, AJ, 132, 2268Miville-Deschˆenes, M.-A., & Lagache, G. 2005, ApJS, 157, 302Mizuno, N., Rubio, M., Mizuno, A., Yamaguchi, R., Onishi, T., &Fukui, Y. 2001, PASJ, 53, L45Ossenkopf, V., & Henning, T. 1994, A&A, 291, 943Ott, J., Wong, T., Pineda, J. L., Hughes, A., Muller, E., Li, Z.-Y.,Wang, M., Staveley-Smith, L., Fukui, Y., Weiß, A., Henkel, C.,& Klein, U. 2008, Publications of the Astronomical Society ofAustralia, 25, 129Pak, S., Jaffe, D. T., van Dishoeck, E. F., Johansson, L. E. B., &Booth, R. S. 1998, ApJ, 498, 735Papadopoulos, P. P., Thi, W.-F., & Viti, S. 2002, ApJ, 579, 270Pelupessy, F. I., Papadopoulos, P. P., & van der Werf, P. 2006,ApJ, 645, 1024Pineda, J. E., Caselli, P., & Goodman, A. A. 2008, ApJ, 679, 481Rosolowsky, E., Engargiola, G., Plambeck, R., & Blitz, L. 2003,ApJ, 599, 258Rosolowsky, E., & Leroy, A. 2006, PASP, 118, 590Rosolowsky, E. W., Pineda, J. E., Kauffmann, J., & Goodman,A. A. 2008, ApJ, 679, 1338Rubio, M., Boulanger, F., Rantakyro, F., & Contursi, A. 2004,A&A, 425, L1 Rubio, M., Garay, G., Montani, J., & Thaddeus, P. 1991, ApJ,368, 173Rubio, M., Lequeux, J., & Boulanger, F. 1993a, A&A, 271, 9Rubio, M., Lequeux, J., Boulanger, F., Booth, R. S., Garay, G.,de Graauw, T., Israel, F. P., Johansson, L. E. B., Kutner,M. L., & Nyman, L. A. 1993b, A&A, 271, 1Russell, S. C., & Dopita, M. A. 1990, ApJS, 74, 93Schnee, S., Bethell, T., & Goodman, A. 2006, ApJ, 640, L47Schnee, S., Li, J., Goodman, A. A., & Sargent, A. I. 2008, ApJ,684, 1228Schnee, S. L., Ridge, N. A., Goodman, A. A., & Li, J. G. 2005,ApJ, 634, 442Solomon, P. M., Rivolo, A. R., Barrett, J., & Yahil, A. 1987, ApJ,319, 730Stanimirovic, S., Staveley-Smith, L., Dickey, J. M., Sault, R. J.,& Snowden, S. L. 1999, MNRAS, 302, 417Stepnik, B., Abergel, A., Bernard, J.-P., Boulanger, F.,Cambr´esy, L., Giard, M., Jones, A. P., Lagache, G., Lamarre,J.-M., Meny, C., Pajot, F., Le Peintre, F., Ristorcelli, I., Serra,G., & Torre, J.-P. 2003, A&A, 398, 551Strong, A. W., & Mattox, J. R. 1996, A&A, 308, L21Thronson, Jr., H. A. 1988, in NATO ASIC Proc. 232: Galacticand Extragalactic Star Formation, ed. R. E. Pudritz &M. Fich, 621–+Thronson, Jr., H. A., Greenhouse, M., Hunter, D. A., Telesco,C. M., & Harper, D. A. 1988, ApJ, 334, 605Thronson, Jr., H. A., Walker, C. K., Walker, C. E., & Maloney,P. 1987, ApJ, 318, 645Tumlinson, J., Shull, J. M., Rachford, B. L., Browning, M. K.,Snow, T. P., Fullerton, A. W., Jenkins, E. B., Savage, B. D.,Crowther, P. A., Moos, H. W., Sembach, K. R., Sonneborn, G.,& York, D. G. 2002, ApJ, 566, 857Walter, F., Cannon, J. M., Roussel, H., Bendo, G. J., Calzetti,D., Dale, D. A., Draine, B. T., Helou, G., Kennicutt, Jr., R. C.,Moustakas, J., Rieke, G. H., Armus, L., Engelbracht, C. W.,Gordon, K., Hollenbach, D. J., Lee, J., Li, A., Meyer, M. J.,Murphy, E. J., Regan, M. W., Smith, J.-D. T., Brinks, E., deBlok, W. J. G., Bigiel, F., & Thornley, M. D. 2007, ApJ, 661,102Walter, F., Taylor, C. L., H¨uttemeister, S., Scoville, N., &McIntyre, V. 2001, AJ, 121, 727Walter, F., Weiss, A., Martin, C., & Scoville, N. 2002, AJ, 123,225Wilke, K., Klaas, U., Lemke, D., Mattila, K., Stickel, M., & Haas,M. 2004, A&A, 414, 69Wilson, B. A., Dame, T. M., Masheder, M. R. W., & Thaddeus,P. 2005, A&A, 430, 523Wilson, C. D. 1995, ApJ, 448, L97+Wilson, C. D., Scoville, N., Madore, B. F., Sanders, D. B., &Freedman, W. L. 1988, ApJ, 333, 611Wolfire, M. G., Hollenbach, D., & Tielens, A. G. G. M. 1993,ApJ, 402, 195APPENDIX
SYSTEMATIC UNCERTAINTIES IN N ` H FIR2 ´ Several systematic uncertainties may affect N (cid:0) H FIR2 (cid:1) but are hard to quantify and so not reflected in our MonteCarlo estimate of the uncertainties. Here we discuss these for the specific case of N83 (for a more general discussionsee Israel 1997a). We find no strong reason to doubt that Equation 7 yields an approximate estimate of N (H ). N83appears unlikely to harbor a significant population of cold dust and we do not observe compelling evidence that dusttraces mostly warm ionized gas or high optical depth H I . There is likely some blending of populations along theline of sight, but the magnitude of the effect is unclear. Grain processing is largely unconstrained, but we note thedissimilarity between N83 and the dense, cold cores where these effects are usually discussed. Blending of Populations Along the Line of Sight:
N83 is a dense, active region and the line-of-sight distance throughthe SMC may be very long. As a result, the observed dust emission may represent a blend of several dust populationswith different temperatures. The likely effect is that we overestimate the average T dust along the line of sight andtherefore underestimate τ and H FIR2 (e.g., see tests on simulated clouds by Schnee et al. 2006).
Cold Dust:
A related concern is that our longest wavelength data are at 160 µ m. As a result, we would miss anypopulation of cold dust. In the Milky Way, when cold, molecular filaments can be isolated from embedded star6 Leroy et al. Fig. 9.— ( left ) The average H I spectrum over the region of high residuals (black) and spectra from individual lines of sight in this area(gray). The spectrum of CO emission (with an arbitrary normalization) is shown below the H I . ( right ) The distribution of opacities in the(integrated) 21cm line required to explain the residuals in highest contour in Figure 4. Although individual spectra show some evidenceof optical thickness, we see no clear signature of self absorption. The line-integrated values of τ required to explain τ in N83 aremostly higher than the peak values of τ measured anywhere in the SMC by Dickey et al. (2000). Fig. 10.— ( left ) H α emission (gray scale) with Σ FIRH2 shown in contour (both at 36 ′′ resolution). Although H α and H FIR2 roughly coincideon large scales (Figure 4), the detailed distributions are not a good match. ( right ) The effect of changing
DGR to the most extremeplausible values on the relationship between Σ
FIRH2 ( y -axis) and I CO ( x -axis, only J = 2 → DGR in N83, that found in the nearby diffuse gas. The gray points show the highest plausible
DGR , ∼ formation, they are often observed to have low dust temperatures ( T .
15 K) and little out-of-equilibrium emission(e.g., Laureijs et al. 1991; Bernard et al. 1999; Stepnik et al. 2003). As with blending of several populations, cold dustis most likely to be associated with the dense, molecular environment of N83. Missing cold dust would lead us tounderestimate τ and N (H FIR2 ).Given the high T dust in N83 and the presence of ongoing, vigorous star formation we consider it unlikely that thereis a significant amount of cold dust present. We attempt a simple test that reveals the presence of cold filaments inGalactic GMCs (Abergel et al. 1994; Boulanger et al. 1998): we take the median I /I over the region, scale the70 µ m map by this value, and subtract it from the 160 µ m map. This should reveal the location of any local 160 µ mexcess, a likely signature of cold dust. We find no such excess associated with N83 as a whole or the CO peaks inparticular. Other Gas Phases:
We refer to the results of Equation 7 as ”H
FIR2 ” but this is actually an estimate of all gas nottraced by the 21-cm transition. Some of this might be high optical depth H I or warm ionized gas. Neither appears totructure Of A Low Metallicity GMC 17be a plausible explanation for the majority of such gas in N83. This agrees with results from the Milky Way, whereexcess dust emission identified in a similar way also appears to correspond mostly to H (Dame et al. 2001).The right panel in Figure 9 shows the H I opacities required to account for τ in N83 given our adopted DGR .These values, τ = 2–4, are higher than those implied by the fit of Dickey et al. (2000), which yields a maximumline-integrated τ ∼ .
55 (correction factor ∼ .
3) near N83. Indeed, most of the line-integrated values of τ inFigure 9 are higher than any of the peak τ values (i.e., τ in the most opaque velocity channel) measured byDickey et al. (2000) in the SMC (maximum ∼ . τ cm values of ∼
2, which is still toosmall to achieve the line-integrated value of τ cm required account for τ in N83. The 21cm spectra do show someevidence of optical thickness at a brightness temperature of ∼
120 K, but no clear signs of self-absorption at thevelocity of the CO peak (left panel in Figure 9). We cannot rule out unlucky geometry, but achieving line-integratedoptical depths of 2–4 without invoking a contrived scenario appears difficult.Warm ionized gas also seems unlikely to account for most of H
FIR2 . The left panel in Figure 10 shows contours ofΣ
FIRH2 over an H α image (at matched resolution) in the SEST field. Although high τ residuals correspond to H α emission on large scales, the detailed distribution is not a particularly good match. The rank correlation coefficientbetween H α and H FIR2 over the area observed by SEST is ∼ .
1, much lower than the 0 . FIR2 and CO. H α emission is proportional to R n dl and so obviously a flawed tracer of the true warm gas column ( R ndl ), but the poorcorrespondence on small scales still argues that most H FIR2 is not actually warm ionized gas.
Dust Processing in Molecular Clouds:
A significant but hard-to-constrain uncertainty in Equation 7 is if and how dustproperties vary between N83 and the surrounding ISM. The most likely variations are increases in the FIR emissivityor the
DGR . In the Milky Way, the FIR emissivity of dust ( τ FIR /A V ) does appear to increase towards dense regions,increasing by ∼ A V ∼ DGR ratio near star-forming regions to be higher than in the surrounding ISM.The magnitude of grain growth in GMCs remains very poorly constrained and in an active environment like N83it will be balanced against grain destruction (e.g., in shocks). Further, the high dust temperatures, low integratedextinctions ( A V . I /I compared to its surroundings.Because it is unclear what, if any, grain processing is at work in N83, we make no correction to the emissivity. Ifdust in N83 indeed has a high emissivity compared to the diffuse ISM, we will derive values of both N (H FIR2 ) and A V that are too high. Note that our adopted DGR is already twice that in the surrounding diffuse gas. Increasing ordecreasing the adopted
DGR will not affect τ or A V , but will lower or raise N (H FIR2 ). Effect of Changing
DGR on the CO-H relation: The exact value of the
DGR in N83 is the largest systematicuncertainty in our analysis. We discuss the constraints on this quantity in § and DGR in the limiting cases:
DGR equal to that in the diffuse ISM of theSMC Wing (black) and