An X-ray View of Star Formation in the Central 3 kpc of NGC 2403
Mihoko Yukita, Douglas A. Swartz, Allyn F. Tennant, Roberto Soria
aa r X i v : . [ a s t r o - ph . C O ] D ec Submitted to Astronomical Journal
Preprint typeset using L A TEX style emulateapj v. 08/22/09
AN X-RAY VIEW OF STAR FORMATION IN THE CENTRAL 3 KPC OF NGC 2403
Mihoko Yukita , Douglas A. Swartz , Allyn F. Tennant , and Roberto Soria Submitted to Astronomical Journal
ABSTRACTArchival
Chandra observations are used to study the X-ray emission associated with star formationin the central region of the nearby SAB(s)cd galaxy NGC 2403. The distribution of X-ray emissionis compared to the morphology visible at other wavelengths using complementary
Spitzer , GALEX ,and ground-based H α imagery. In general, the brightest extended X-ray emission is associated withH ii regions and to other star-forming structures but is more pervasive; existing also in regions devoidof strong H α and UV emission. This X-ray emission has the spectral properties of diffuse hot gas( kT ∼ . . Subject headings: galaxies: individual (NGC 2403) — galaxies: nuclei — galaxies: evolution — X-rays:galaxies INTRODUCTION
Whether merger-induced collapse(Barnes & Hernquist 1992; Cole et al. 2000), bar-driveninflow (Kormendy & Kennicutt 2004), or dynamicalfriction (Noguchi 2000) builds structure at the center ofa particular galaxy, it is likely that star formation and(if present) central black hole growth will be stronglyregulated by feedback from massive stars (and AGNactivity).Since much of the current central massive objectgrowth and star formation is occurring in small, low-density, disk-dominated spirals rather than the massivebut fuel-starved ellipticals (Heckman et al. 2004), nearbylate-type galaxies are the ideal laboratories to view thegrowth of galactic structure in the current epoch. Here,we investigate star formation, feedback, and the growthof the central region of NGC 2403. We focus on theunique perspective enabled by X-rays in viewing the vi-olent and inherently high-energy phenomena associatedwith these dynamical processes and compare X-ray be-havior to that exhibited at other wavelengths.NGC 2403 is a SAB(s)cd galaxy in the M81 group ofgalaxies ( D = 3 . ′ =1 kpc, Madore & Freedman1991). It is the most massive, ∼ M ⊙ , galaxy inthe second largest of three M81 subgroups. There are7 dSph and dIrr satellites of NGC 2403 known withinthis subgroup (Karachentsev et al. 2002) ranging in sizefrom ∼ × to ∼ × M ⊙ (Karachentsev et al. 2004).As with many disk-dominated late-type spirals,NGC 2403 lacks a central bulge (Kent 1987) but doeshost a luminous compact nuclear star cluster first identi-fied in IR images (Davidge & Courteau 2002). The mass University of Alabama in Huntsville, Dept. of Physics,Huntsville, AL, USA Universities Space Research Association, NASA MarshallSpace Flight Center, VP62, Huntsville, AL, USA Space Science Office, NASA Marshall Space Flight Center,VP62, Huntsville, AL, USA Mullard Space Science Laboratory, University College London,Holmbury St. Mary, Surrey RH5 6NT, UK and luminosity of this cluster are comparable to thosefound in many late-type spirals (B¨oker et al. 2002) butit is older and less compact than typical for late-typegalaxies (Yukita et al. 2007). The dominant age of starswithin the nuclear star cluster is ∼ α surface brightness near thecenter is lower than in surrounding regions while theyoungest ( ∼ ii regions(Drissen et al. 1999) are 0.7 to 1.6 kpc from the cen-ter. Davidge & Courteau (2002) speculate that this agegradient may be due to the growth of a superbubble inthe central region that has quenched star-formation inthe central region while triggering activity further outthrough compression of surrounding gas during bubbleexpansion. In this picture, nuclear star formation maybe an episodic phenomenon with roughly a few 100 Myrinterval between major star-forming events.Such a star formation cycle requires a replenishingsource of cold gas. Sheth et al. (2005), using molecu-lar gas maps, showed that bulges can be built by bar-driven gas inflow but that the process requires of or-der a Hubble time in galaxies like NGC 2403. Otherstudies have shown that the molecular and atomic gas inthe central regions of NGC 2403 amount to only a smallfraction of the dynamical mass (e.g., Thornley & Wilson1995). Importantly, the central total gas surface den-sity (e.g., Thornley & Wilson 1995; Martin & Kennicutt2001), at least on kiloparsec spatial scales, is below thecritical value for star formation (Kennicutt 1989) yet theglobal star-formation rate in NGC 2403 is a moderate1.2 M ⊙ yr − .More pertinent to the current study are several in-dependent investigations of the dynamics of the H i gas in NGC 2403 (Sicking 1997; Schaap et al. 2000;Fraternali et al. 2002b). The H i rotation curve showsa flat gravitational potential typical of late-type spiralsbut there is also slower-rotating neutral hydrogen ex- Fig. 1.—
Left: The
GALEX
FUV image of NGC 2403. The 21 . ′ × . ′ D isophote of NGC 2403. The central 6 ′ × ′ region is shown as a square. The gray scale indicates intensity level in the unit of ct s − pixel − . One pixel corresponds to 1 . ′′
5. Right:The same region in the merged X-ray data. The outer boundaries of the ACIS S2-S3-S4 combination are shown as thin-lined rectanglesfor each of the 4 overlaid observations. The central 6 ′ × ′ region is shown as a thick-lined square. The large circle in the center is the 2 . ′ − . One pixelcorresponds to 0 . ′′ tending up to 15 kpc above the disk. Many propertiesof this anomalous H i gas can be explained as galacticfountains (Shapiro & Field 1976; Bregman 1980; Spitzer1990) but an additional, external, source of cold gassuch as infalling clouds or even small satellite galax-ies might also be required (Fraternali & Binney 2006,2008; Struck & Smith 2009). High-resolution VLA ob-servations of NGC 2403 (Fraternali et al. 2002b) showvery anomalous kinematic features on small scales thatmay be evidence of such clouds. Direct evidence offountain activity is less conclusive. Measurements ofH α emission line widths at several locations across thegalaxy (Fraternali et al. 2004) found only a few small-scale fountain-like features. To match the observed H i in-flow with hot gas outflow would require many thousandsof these small-scale fountains. Analysis of the spatial dis-tribution of diffuse hot X-ray-emitting gas in NGC 2403(Fraternali et al. 2002a) concluded that as little as a fewpercent or as much as the majority of the inflowing H i canbe matched by outflowing diffuse hot gas, depending onassumptions made. Thus, it is not yet clear whether thebuildup of the central regions of NGC 2403 is being fu-eled cyclically through galactic fountains, or by accretinggas from the intergalactic medium, or through mergersof small-scale gaseous satellites.If galactic fountains are at work and are the source ofthe majority of the observed infalling H i gas, then thereshould be an imprint of this process in the X-ray emissionfrom the disk correlated with the source of the fountains;namely localized star-forming regions that heat the gasthrough massive stars to the point of breakout from thedisk.We confine our study of NGC 2403 to the inner 6 ′ × ′ nuclear region (corresponding to 5.6 × ii regions identified by Drissen et al. (1999).We begin ( §
2) with an independent analysis of the X-ray data; examining X-ray-detected discrete sources andthe underlying residual X-ray emission. This analysisextends the previous work of Schlegel & Pannuti (2003)and of Fraternali et al. (2002a) by including 3 subsequent
Chandra observations which allow us to better quantifythe spectral and temporal behavior of the X-ray emis-sion. Our results are consistent with both these previ-ous investigations. We expand our analysis to includeoptically-identified supernova remnants (SNRs) and H ii regions. Although individually X-ray-faint, we are ableto characterize their bulk (average) X-ray temperaturesand other X-ray properties using stacking analysis.We then turn to analysis of individual massive star-forming regions ( §
3) in an attempt to determine theX-ray emission properties of these regions as a functionof their age, mass, and extinction properties. We usea combination of (ground-based) H α , mid-IR ( Spitzer ),and UV (
GALEX ) images to define these regions and todetermine their basic physical characteristics. We thenuse our knowledge of the differences in X-ray propertiesbetween H ii regions and SNRs assembled in § Chandra
X-RAY OBSERVATIONS AND PRELIMINARYANALYSIS
In this section, archival
Chandra images are exam-ined to derive the X-ray properties of four differentsource populations in NGC 2403; namely, bright X-ray-detected sources (mainly luminous X-ray binaries andbackground AGNs), optically-identified SNRs, optically-identified H ii regions, and the underlying unresolved X-ray emission. This allows us to parameterize these dif-ferent types of sources in terms of their X-ray temper-atures, luminosity distributions, and emission measures.This information will be used in § TABLE 1
Chandra
Observation log
Date ObsID Instruments & Mode GTI2001/04/17 2014 ACIS-S TE 36.1 ks2004/08/13 4628 ACIS-S TE 47.1 ks2004/10/03 4629 ACIS-S TE 45.1 ks2004/12/22 4630 ACIS-S TE 50.6 ks dividual massive star-forming regions defined at otherwavelengths.NGC 2403 was observed in full-frame mode with
Chan-dra
ACIS-S on four occasions for a total of ∼
180 ks(Table 1). We obtained level 1 event lists for all fourobservations from the
Chandra data archive and re-processed them using the CIAO (version 3.3.0.1) tool acis process events and calibration database CALDB3.2.1. Reprocessing removed pixel randomization, ap-plied CTI- and time-dependent gain corrections, and re-moved events with bad grades or bad status bits as wellas bad and hot pixels. We created lightcurves for eachobservation using a 1 ks binning to check for periods ofhigh background. We excluded intervals with total countrates > σ above the mean rate for each observation. Thefinal Good Time Intervals for the observations are listedin Table 1.SN 2004dj was used to define a common registrationamong the data sets. For this purpose, a circular Gaus-sian was fit to the image of the supernova to obtainan accurate centroid then the coordinates were adjustedto agree with the known position of SN 2004dj (J2000R.A.=7 h m s , Decl.=+65 ◦ ′ ′′ , Argo et al.2004). To register pre-supernova data, we bootstrappedusing a bright point source near the aimpoint that is com-mon to both pre- and post-supernova data. The fourco-aligned Chandra data sets were combined using theFTOOL utility fmerge to form a fifth merged dataset.The right panel of Figure 1 displays the merged X-rayimage. The longest cumulative exposure is near the cen-ter of the galaxy where we have selected a 6 ′ × ′ regionfor analysis as indicated. Also shown are the galaxy’s D ellipse, corresponding to the 25 mag-sec − contourin B . The 6 ′ × ′ region contains many point-like sources,extended star forming regions, and an extended regionof X-ray emission near the galactic center. With theexception of the bright point-like X-ray sources, we areprimarily concerned with the soft X-ray emission compo-nents of NGC 2403. For the soft emission, we fit spectrain the 0.4 − − X-ray Source and Region Definitions
The source-finding tool described by Tennant (2006)was applied to all five datasets to search for discreteX-ray sources. The search was limited to the central6 ′ × ′ region and to events within the full Chandra en-ergy range 0.3 − σ above background uncertainty (corre-sponding to a detection limit of about 8 −
10 counts for http://cda.harvard.edu/chaser/ TABLE 2X-ray Point sources
R.A. Decl Count Rate a L X b VariabilityJ2000 J2000 10 − cts s 10 erg s −
07 36 24.5 +65 37 13.1 1.2 ± ± − ± − ± ± ± ±
12 456.6 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± − ± − ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± − ± − ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note . — Units of right ascension are hours, minutes, and seconds,and units of declination are degrees, arcmintes, and arcseconds. a Count rates for ASIC-S3, taking the 0.5 − b L X for the 0.5 − L X is computed using thebremsstrahlung model obtained from Table 3 (see text). a typical on-axis source). This tally included all sourcesdetected in the individual observations (provided theywere within the FOV of the individual images) and allthose previously reported by Schlegel & Pannuti (2003)within our 6 ′ × ′ field. The X-ray point sources and theirproperties are listed in Table 2.Since we are primarily interested in star-formation andthe X-ray properties of recent and current star-formingevents in this study, we also examined the X-ray emissionfrom known SNRs and H ii regions. For this purpose, wefirst masked out the X-ray-detected point-like sources inthe field of the merged image using a mask radius equalto 3 times the half-width of a Gaussian approximationto the instrumental point spread function (PSF). Largerradii were used, as necessary, for a few bright sourceswith strong emission in their PSF wings. Spectral infor-mation was then extracted from events falling within thearea defined by optically identified SNRs from [SII]/H α ratio by Matonick et al. (1997) using their listed loca-tions and source radii. One arcsecond ( ∼
16 pc ) radiiare used for the SNRs whose sizes are unlisted. Localbackgrounds were also extracted from surrounding an-nuli. There are 29 cataloged SNRs within our 6 ′ × ′ field.There are also three candidate radio SNRs in our field(Pannuti et al. 2007). One (denoted µ by Pannuti et al.2007) is already included in the list of optically-identifiedSNRs. Another (denoted TH2 by Pannuti et al. 2007) isa strong X-ray source, L X = 2 × erg s − , as previ-ously noted by Pannuti et al. (2007). However, no radiospectral indices are available for any of these candidateradio SNRs. Therefore, we opt to omit TH2 from ourlist of candidate SNRs as it is anomolously bright if it is,in fact, a SNR. It remains classified as a discrete X-raysource in our tabulation.A similar procedure was attempted using the H ii re-gions defined by Sivan et al. (1990) based on photo-graphic images but many of these cataloged objectsdid not correspond to H ii regions visible in a recentcontinuum-subtracted H α CCD image. This is duemostly to crude modeling of the H ii structure which re-sulted in poor estimates of source positions and sizes.Therefore, we constructed our own catalog of H ii regionsdefined as circular approximations to the areas enclosedby surface brightness contours corresponding to a level of20 × the background in our continuum-subtracted H α im-age. There are 58 H ii regions defined in this way withinour 6 ′ × ′ field including 5 of the 6 giant H ii regions iden-tified previously by Drissen et al. (1999).There are four point-like sources detected by oursource-finding trial that spatially coincide with SNRs.Two more detected sources lie within a bright extendedH ii region. Upon further examination, all these sourceswere determined to be steady thermal sources and weretherefore assigned to the SNR (H ii ) category rather thanto the general category of point sources for purposes ofour analysis. Conversely, the PSF wings of bright pointsources overlap with the positions of two of the catalogedSNRs. These two SNRs were therefore excluded from ouranalysis because of this contamination. The three his-torical supernovae, SN 2004dj (a strong X-ray source),SN 2002kg and SN 1954J (both fainter than our detec-tion limit) and the X-ray-bright candidate radio SNRwere also not considered further in this study.Finally, we also examined the properties of the X-rayemission remaining after the X-ray point sources and theoptically-identified SNRs and H ii regions are excluded.This residual component is likely comprised of unresolved(faint) XRBs and other low-luminosity objects related tothe stellar content of NGC 2403, extended diffuse hot gas TABLE 3Point sources fitting results
Fit Parameters Bremsstrahlung Powerlaw N H (10 cm − ) 1.9 +0 . − . +0 . − . kT (keV) 4.7 +0 . − . Γ 1.87 +0 . − . Normalization (10 − ) 5.9 +0 . − . +0 . − . L X (0.5-2.0 keV)/10 erg s − +0 . − . +0 . − . L X (2.0-10 keV)/10 erg s − +0 . − . +0 . − . χ /dof 185/197 211/197 within the disk and halo of NGC 2403, and unresolvedbackground (primarily AGN) and foreground (local dif-fuse Galactic) emission. Inspection of the residual emis-sion image shows a clear excess above the background inthe central regions of NGC 2403.We examine the X-ray properties of these four sourcetypes in the following subsections. Discrete X-ray-Detected Sources
Schlegel & Pannuti (2003) have already presented theproperties of the discrete source population detected inthe first
Chandra observation of NGC 2403. Here, ourinterest is only in the average spectral shape of the dis-crete sources and their luminosity distribution. The aver-age spectrum can be represented by co-adding the sourcespectra from only the first observation. This providessufficient source counts for characterizing the spectrumyet avoids complications that arise when combining ob-servations made at different times during the
Chandra mission. The spectra of the individual point sources (andcorresponding backgrounds) were co-added and weightedancillary response (ARF) and response matrix (RMF)files were generated using CIAO tool acisspec . Theresulting spectrum was grouped to ensure at least 20counts in each spectral bin. The source spectrum wasmodelled using the XSPEC spectral-fitting utility (ver-sion 11.3.2t).We applied both absorbed powerlaw andbremsstrahlung models to the co-added discretesource spectrum. The fitting results are shown in Ta-ble 3, and the spectrum is shown in Figure 2. Althoughwe are mainly interested in the energy range 0.5 − − Chandra observa-tions. This check helps to differentiate between diffuseemission associated with star-formation that has a rela-tively soft X-ray spectrum and no variability and emis-sion from XRBs that is spectrally hard and often variableover time. . . r m a li z ed c oun t s / s e c / k e V − . . r e s i dua l s channel energy (keV) Fig. 2.—
Top: The co-added (stacked) spectrum of all 58discrete sources detected in the
Chandra
X-ray data. The co-added spectrum was constructed using only the first observation.Also shown are the best-fit model (an absorbed bremsstahlung)and the fit residuals. Bottom: The X-ray luminosity func-tion on the energy range of 0.5 − × − ct s − corresponds to 10 − ergs cm − s − ). Thesingle powerlaw model fitted to the luminosity function for sourcesbrighter than 2 × − ergs cm − s − ∼ × erg s − is shownas a solid curve and gives a slope of − If the flux from a source during one observation devi-ated from its average over the other available observa-tions by > σ , then it was flagged as long-term variableand designated by the letter “L” in column 5 of Table 2.(Because of differences in spacecraft roll and target aim-point, not all sources were within the field of view duringall four observations.) Short-term variability during in-dividual observations was checked by binning the sourcelight curve (event arrival time) for each observation us-ing 2 ks temporal bins. The shape of the light curve wascompared to a constant using the χ statistic to check forvariability. Short-term variability is also given in column5 of Table 2 by the letter “S”.The cumulative X-ray luminosity function (XLF) forthe 58 sources detected within the 6 ′ × ′ central re-gion of NGC 2403 is shown in Figure 2. Luminosi-ties for each source were computed from their aver-age flux over up to four observations, depending onavailability, on the energy range 0.5 − × − ergs cm − s − in the 0.5 − × erg s − ), a single powerlaw gives a best-fitting slope of − . ± .
02, roughlyconsistent with the value of − . ± .
02 reported bySchlegel & Pannuti (2003). For reference, a flux of10 − ergs cm − s − in the 0.5 − × − ergs cm − s − in the 2.0–10.0 keV band assuming our best-fit co-added discretesource spectral shape.An estimate of the contribution from unrelated back-ground AGN to the detected source population in thefield can be made by comparing the observed XLF to theanalytical fits to the deep field cosmic X-ray backgroundlog N –log S (Moretti et al. 2003). This fit is shown in Fig-ure 2. An estimated 21 ± ± Supernova Remnants
Most of the individual SNRs (and H ii regions discussedin the next subsection) are too faint to be detected as dis-crete X-ray sources. The possibility exists that many ofthe X-ray events detected within the spatial regions de-fined by the individual optical SNRs are unrelated “back-ground” events. Figure 3 shows the XLF for the SNRs.Also shown is a pseudo-random sampling of the back-ground defined using the same locations and sizes of thecataloged SNRs but with their RA coordinates reflectedabout the center of NGC 2403. (For both populations,only those regions with > > − × − ergs cm − s − ) gives a best-fitting slope of − ± α line emission. However,there is no overall correlation between X-ray and H α lu-minosity in the SNR sample. This is consistent with theanalysis of Pannuti et al. (2007). Using the two SNRswhose diameters are listed in Matonick et al. (1997), wealso find the mean diameter of X-ray-detected SNRs issmaller (40 pc) than the mean diameter of X-ray non-detections (70 pc) again comfirming the Pannuti et al.(2007) result.A co-added spectrum of the cataloged SNRs was cre-ated by merging the first three Chandra images (thefourth observation does not adequately cover the cen-tral 6 ′ × ′ region of interest). Twenty-four SNRs (of 27total) are imaged on S3 in the first three observations.The same averaging method used for the discrete sources( § − − − no r m a li z ed c oun t s / s e c / k e V − − − r e s i dua l s channel energy (keV) Fig. 3.—
Top: The co-added (stacked) spectrum of the 24optically-identified SNRs imaged on S3. The best-fitting model(an absorbed apec model with hydrogen column density, N H , al-lowed to be a free parameter) is shown along with the fit resid-ual. The spectrum has been re-binned to have a width of 73eV for display purposes. Bottom: The luminosity function ofthe background-subtracted SNRs (solid line) and of the pseudo-random positions (dashed line, see text). The single powerlawmodel fitted to the luminosity function of sources brighter than1.1 × − ergs cm − s − ∼ × erg s − is shown as a curveand gives a slope of − TABLE 4SNRs X-ray fitting results
Fit Parameters apec apec apec N H (10 cm − ) 0.4 1.4 2.8 +3 . − . T e (MK) 4.8 +2 . − . +1 . − . +2 . − . Normalization a +0 . − . +0 . − . +2 . − . L X /10 erg s − +0 . − . +0 . − . +0 . − . L int X /10 erg s − b V [ f ] (10 cm ) 0.13 0.13 0.13 n e [ f − / ] (cm − ) 0.04 0.05 0.07 P/k [ f − / ] (10 K cm − ) 3.5 4.1 5.1 M x [ f +1 / ] (10 M ⊙ ) 0.05 0.07 0.11 E th [ f +1 / ] (10 erg) 0.09 0.1 0.1 t c [ f +1 / ] (Myr) 48 30 16 a K = (10 − / πD ) R n e n p dV b volume filling factor scaling in [ ] in Column 1. ray events in the co-added spectrum between 0.4 and2.0 keV is 153. This reduces to 94 . ± . apec , represent-ing thermal emission from an optically-thin collisionally-ionized plasma, combined with a model phabs for anintervening absorption column were applied to the spec-trum. The parameters of the apec model are the plasmatemperature, elemental abundances, and emission inte-gral. Elemental abundances were fixed to the solar val-ues as given in Anders & Grevesse (1989). Three modelfits were attempted as detailed in Table 4. These dif-fered in the values assigned to the absorbing columns.In two of these models, the absorbing column was heldfixed at either the Galactic value along the line of sight, N g H = 4 × cm − , implying no local absorption withinNGC 2403, or N H = 14 × cm − which is equivalentto the mean value of the optical extinction derived for ac-tive star-forming regions ( § ii regions and of the residualX-ray emission to be discussed below, allowing the ab-sorbing column, plasma temperature, and normalization(emission integral) to vary in the fitting process oftenleads to large uncertainties in some of these parametervalues. The reason for this is simply that the plasmatemperature is quite low, of order a few 10 K (a few0.1 keV), so that most of the emission is at low energieswhere the sensitivity to absorption is most acute. Thisintroduces a degeneracy in which an equally acceptablefit is possible by increasing the column density, N H , whilesimultaneously increasing the model normalization, usu-ally with a moderate decrease in the fitted temperature.The X-ray luminosities of the individual SNRs werecomputed by adopting the spectral model that best fitsthe stacked spectrum (Table 4, column 4). This modelwas applied to the distribution of events of each indi-vidual SNR using the best-fit values of T e and N H , andallowing only the model normalization to vary in the fi-nal fitting. This is equivalent to scaling by the numberof counts in the individual spectra. The X-ray countrates, corresponding luminosities, and H α luminositiesare listed in Table 5.A number of other physical parameters can be derivedfrom the fitted values of the plasma temperature andemission integral provided some estimate of the volumeoccupied by the hot gas can be made. The total volumeof the 24 SNRs is 1 . × cm assuming a spheri-cal geometry for each remnant and adopting the radiireported in Matonick et al. (1997; the same radii usedhere to define the X-ray spectral source regions). Theactual volume occupied by the hot gas is some fraction, f , of this total volume. Assuming the ion and electronnumber densities are equal, which is roughly true for ahydrogen-dominated hot plasma, then the emission in-tegral R n e n H dV ∼ n e f V so that the electron density n e ∝ ( K/f V ) / where K is the spectral model nor-malization parameter which is itself proportional to theemission integral.From n e and the flux-weighted mean temperature T e ,we can estimate the hot gas pressure P/k = 2 n e T e , mass M X = n e µm p V where m p is the proton mass and µ = 1 . E th = 3 n e V kT e and the cooling time t c = E th /L bol of the X-ray emitting TABLE 5SNRs
RA Dec Count Rate a L X b L Hα (J2000) (J2000) 10 − cts s − erg s − erg s −
07 36 42.8 +65 34 52.7 − ± − ± ± ± ± ± ± ± − ± − ± − ± − ± ± ± ± ± − ± − ± ± ± − ± − ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± − ± − ± − ± − ± − ± − ± ± ± ± ± ± ± a Count rates for ACIS-S3, taking the 0.5 − b Luminosity in the 0.5 − L X is computedusing the one temperature model with variable N H model as listedin Table 4 (see text) plasma as listed in Table 4. Here, L bol is estimated byintegrating the spectral model from 0.02 keV to 20 keVwithin XSPEC. For plasma temperature typical of SNRs(and of other thermal sources analyzed in this paper), L bol is ∼ L X , where L X is the observed X-rayluminosity quoted on the standard 0.5 − f as indicated in the table.The total X-ray luminosity from the 24 SNRs is only3 × erg s − . Inspection of Figure 3 shows that mostof this emission comes from the few brightest sourceswith the most luminous being ∼ × erg s − . Thethermal energy content of the SNRs, ∼ f / erg, cor-responds to up to ∼
40% (in the limit f = 1) of the initialkinetic energy released by the supernova explosions, as-suming 10 erg per event. The total mass of hot gas issubstantial. It corresponds to an average mass per SNRof M X ∼ ± f / M ⊙ (uncertainties given by therange of values deduced from the three X-ray spectralmodels). This can be compared to the amount of ma-terial swept up by the SNRs, ∼ n H M ⊙ per SNR,assuming a uniform density n H cm − , over the SNR vol-ume. H ii Regions
We performed an analysis of the 48 H ii regions, im-aged on S3 in the first three observations, analogous tothat of the SNRs. Figure 4 shows the XLF for the H ii regions and the pseudo-random sampling defined usingRA coordinates reflected about the center of NGC 2403.Again, only the most X-ray-luminous H ii regions aresignificant detections. Similar to the SNRs, we alsofitted a single power law model to the XLF of the 7 brightest H ii regions, ( > × − ergs cm − s − ). The best-fitting slope is − . ± .
13, somewhat shallower than for the SNRs.The co-added X-ray spectrum is shown in Figure 4.Similar to the SNR spectrum, we also applied thermalmodels to the H ii region spectrum. There are sufficientsource counts (858 cts, 422 ±
32 after the backgroundsubtraction) in the H ii region spectrum to allow 2-temperature model fits in addition to the 1-temperaturemodels that were applied to the SNR spectrum. The fit-ting results and derived values are listed in Table 6. Weassumed pressure equilibrium between the two temper-ature components and identical filling factors to derivethe values of n e , E th , and M x .Note that the best-fitting 1-temperature model withvariable column density results in a very high model nor-malization which leads to unreasonable values of the de-rived parameters. However, this does not seem to be theissue with the 2-temperature model. This model (col-umn 6 in Table 6) results in N H ∼ N g H (though withlarge uncertainty) and hence with values of the remain-ing parameters similar to those of the fixed N H model.The X-ray luminosities of the individual H ii regionswere computed in the same manner as for the SNRs; us-ing the best-fitting two temperature spectral model withvariable N H in this case. The X-ray count rates, corre-sponding luminosities, and H α luminosities are listed inTable 7.There are twice as many H ii regions in our sampleas there are SNRs in our SNR sample. The X-raytemperatures of the H ii regions are about 40% lowercompared to the SNRs. The total X-ray luminosity is ∼ × erg s − for the H ii regions which is only a Fig. 4.—
Top: The stacked spectrum of the identified H ii regions.The best-fitting model (an absorbed two-temperature apec modelwith hydrogen column density, N H , as a free parameter) is shownalong with the fit residuals. The spectrum has been re-binned tohave a width of 73 eV for display purposes. Bottom: The luminos-ity function of the background-subtracted H ii regions (solid line)and of the pseudo-random positions (dashed line, see text). Thesingle powerlaw model fitted to the luminosity function of sourcesbrighter than 1.1 × − ergs cm − s − ∼ × erg s − is shownas a curve and gives a slope of − factor of 3 higher, on average, for individual H ii regionscompared to the SNRs. Similarly, the average hot gasmass is ∼ ⊙ , per H ii region, based on the best-fitting 2-temperature spectral model. This is about afactor of 3 larger than the hot gas mass in individualSNRs. Residual Emission
The left panel of Figure 5 displays a smoothed image ofthe soft X-ray emission from the underlying unresolvedcomponent. This residual emission is defined as the netemission left after masking out all detected point sources(including SN 2004dj and TH2), H ii regions and SNRs.For spectral fitting, an X-ray map of this component wascreated using only the first Chandra observation. This isa compromise in that a deeper image could be made usingan image merged from two or more individual observa-tions but the area of the region of overlap sampled bythese multiple observations is too small to define botha sizable source region and a surrounding backgroundregion that are both wholly contained within the over-lap region (see Figure 1). The first
Chandra observationplaced the center of NGC 2403 near the S3 aimpoint and thus is the most useful for our purposes. The sourceregion is defined as a disk of radius 2 . ′ dmfilth , then divided by an exposuremap evaluated at an energy of 0.5 keV. This exposure-corrected image was then smoothed using the CIAO tool aconvolve using a Gaussian function with a width of 10pixels ( ∼ ′′ ).There are several nearly point-like regions that remainin the residual emission image. These may be faint pointsources that were hidden in the PSF wings of brighternearby sources or extended non-spherical features in H ii regions not covered by our (circular) masks. We exam-ined the five brightest of these regions and find they con-tribute only 4% of the counts in the residual emissionspectrum. Therefore, these regions were left in as partof the residual emission.In general, the residual emission fills a broad regionof the central part of NGC 2403 extending from north-west of the nucleus (located at the center of the image)to southeast of the nucleus. This region is not a par-ticularly strong source of emission at UV, mid-IR, norH α wavelengths. For example, a grayscale UV imageis shown in the right panel of Figure 5 with the X-raycontours overlaid. This image suggests that much of theresidual X-ray emission is centrally located relative tothe UV-bright zones that contain stars < ∼
100 Myrs old.Several models were fitted to the spectrum of this resid-ual X-ray emission. Area-weighted ARFs and RMFswere created for this purpose using the CIAO script specextract . The spectral fitting results are listed inTable 8.The three thermal models with the absorption columndensity frozen during the fitting all give low values forthe model normalization, electron density, pressure, hotgas mass, and thermal energy. The best-fitting of thesemodels is the 2-temperature model with temperaturesof 2.3 and 8.6 MK with most of the flux in the coolercomponent (Table 8, column 5), shown in Figure 6. Fix-ing the absorption column density to the Galactic valueimplies that the residual emission lies above the disk ofNGC 2403 where there is little or no overlying cold gas.This is a reasonable assumption if the X-ray-emitting gaswas ejected from the galaxy through galactic-scale windsor smaller fountains or chimneys. Fixing the absorptioncolumn density to the equivalent mean value of the op-tical extinction toward the active star-forming regionsin NGC 2403, N H = 1 . × cm − , corresponds to amodest layer of overlying cold gas.The models with the absorption column density freeto vary during the fit result in statistically improvedfits with absorbing columns higher than found for anyof the other types of X-ray sources considered in theprevious subsections. Such high columns imply thatthe hot gas is located behind a layer of neutral gassuch as would be the case if the hot gas were con-fined to the disk. However, the observed H i col-umn densities through the central disk of NGC 2403 See http://cxc.harvard.edu/ciao/threads/diffuse emission/
TABLE 6Hii regions X-ray fitting results
Fit Parameters apec apec apec apec+apec apec+apec N H (10 cm − ) 0.4 1.4 7.8 +1 . − . +2 . − . T e (MK) 3.5 +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . Normalization K a +0 . − . +0 . − . +33590 − +0 . − . +4 . − . T e (MK) 8.8 +2 . − . +1 . − . Normalization K +0 . − . +0 . − . L X /10 erg s − +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . L int X /10 erg s − b V [ f ] (10 cm ) 2.2 2.2 2.2 2.2 2.2 n e [ f − / ] (cm − ) 0.02 0.03 1.1 0.07,0.01 0.07,0.01 P/k [ f − / ] (10 K cm − ) 1.4 1.9 40 2.5 2.5 M x [ f +1 / ] (10 M ⊙ ) 0.5 0.8 28 0.5 0.5 E th [ f +1 / ] (10 erg) 0.6 0.9 18 1.1 1.1 t c [ f +1 / ] (Myr) 55 32 0.1 32 32 a K = (10 − / πD ) R n e n p dV b volume filling factor scaling in [ ] in Column 1. Fig. 5.—
The residual X-ray emission from NGC 2403. Left: The soft X-ray image (0.4 − ′ × ′ region. Allfour observations have been co-added and exposure-corrected and all discrete sources, optically-identified SNRs and H ii regions have beenremoved. The image has been aggressively smoothed with the CIAO tool aconvolve . The contours indicate 10, 20, 30, and 40 σ above thebackground level. The gray scale indicates intensity levels in units of photon cm − s − pixel − . One pixel corresponds to 0 . ′′ GALEX
FUV image. The gray scale of the image indicates intensity levels in unitsof ct s − pixel − . One pixel corresponds to 1 . ′′ range from only ∼ cm − (Thornley & Wilson 1995)to ∼ × cm − (Schaap et al. 2000; Fraternali et al.2002b) which is considerably less than implied by theseX-ray model fits, suggesting the models are unphysi-cal. Furthermore, inspection of the images of this hotgas component and comparison to images at other wave-lengths suggest that the hot gas is not emitted from re-gions with high levels of overlying cold gas.Therefore, the models with the absorption columnfixed are the most realistic models. They imply (4 − × M ⊙ of hot gas is present in the central regions ofNGC 2403, assuming a 2.3 kpc radius spherical emissionvolume. This hot gas is rather tenuous, n e ∼ .
003 to0.004 cm − , with, formally, a long cooling time of 200 −
340 Myr. We note that the actual X-ray emitting re- gion is slightly smaller than the 2.3 kpc radius spectrumextracting region, so that the assumed volume may belarger than the volume containing the hot gas. If we as-sume instead that the hot gas is confined to a disk withthickness 200 pc and 2.3 kpc radius, then the inferreddensity increases by about a factor of 4 in compensationfor the smaller volume for a given emission integral. Thisreduces the cooling time to 50 −
90 Myrs. The true emis-sion volume is probably somewhere between our sphereand disk estimates. Table 8 contains derived propertiesof the residual gas for both these geometries.An estimate of the contribution to this residual X-rayemission from unresolved sources fainter than, but oth-erwise similar to, the discrete source population ( § TABLE 7Hii regions
R.A. Decl Count Rate a L x b L Hα J2000 J2000 10 − cts s 10 erg s − erg s −
07 36 35.7 +65 35 09.0 3.8 ± ± ± ± ± ± ± ± − ± − ± ± ± ± ± ± ± − ± − ± − ± − ± ± ± − ± − ± − ± − ± − ± − ± − ± − ± ± ± ± ± ± ± − ± − ± − ± − ± − ± − ± − ± − ± − ± − ± ± ± − ± − ± − ± − ± − ± − ± − ± − ± − ± − ± − ± − ± ± ± ± ± − ± − ± ± ± ± ± − ± − ± ± ± ± ± − ± − ± − ± − ± ± ± ± ± − ± − ± − ± − ± ± ± ± ± ± ± Note . — Units of right ascension are hours, minutes, and seconds,and units of declination are degrees, arcmintes, and arcseconds. a Count rates for ASIC-S3, taking the 0.5 − b Luminosity in the 0.5 − L X is computed usingthe two temperature model with variable N H model listed in Table 6. minosity function slope (Figure 2) to lower luminosities.For this, we scale the XLF to include only those sourceswithin the 2 . ′ § × erg s − in the 0.5 to 2.0 keV energy range orabout a 10–15% contribution to the diffuse emission.As an alternative estimate, we added a bremsstrahlungmodel component to the hot gas spectral model to esti-mate the contribution from unresolved sources. Addinga bremsstrahlung component does not improve statisti-cal significance of the fit but can contribute up to ∼ − ii regions to the residual X-ray emis- sion can also be estimated from their XLFs (Figures 3,and 4). These estimates are 9 × erg s − for the SNRsand 1 × for the H ii regions. The estimated contribu-tion from SNRs is relatively high because the XLF slopeis steep for this source population. In any event, faintSNRs and H ii regions can contribute only a small frac-tion to the total X-ray luminosity of the residual emissionwhich is 1.6 × erg s − or higher. Discussion of Preliminary Analysis of X-rayObservations
The bulk X-ray properties of the SNRs, H ii regions,and diffuse (residual) emission are summarized in Ta-bles 4, 6, and 8, respectively. They show that the hot1 TABLE 8Residual emission X-ray fitting results
Fit Parameters apec apec apec apec+apec apec+apec N H (10 cm − ) 0.4 1.4 5.6 +1 . − . +3 . − . T e (MK) 3.1 +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . Normalization K a +0 . − . +1 . − . +1716 − +1 . − . +2639 − T e (MK) 8.6 +1 . − . +697 − . Normalization K +0 . − . +13 . − . L X /10 erg s − +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . L int X /10 erg s − b V csphere [ f ] (10 cm ) 1535 1535 1535 1535 1535 n e [ f − / ] (cm − ) 0.003 0.004 0.026 0.006,0.002 0.08,0.003 P/k [ f − / ] (10 K cm − ) 0.2 0.2 0.8 0.3 2.5 M x [ f +1 / ] (10 M ⊙ ) 47 66 460 43 167 E th [ f +1 / ] (10 erg) 51 67 270 90 830 t c [ f +1 / ] (Myr) 311 203 7.3 342 30 V ddisk [ f ] (10 cm ) 97 97 97 97 97 n e [ f − / ] (cm − ) 0.01 0.01 0.10 0.023,0.006 0.31,0.01 P/k [ f − / ] (10 K cm − ) 0.7 0.8 3.3 1.1 10 M x [ f +1 / ] (10 M ⊙ ) 11 17 114 11 42 E th [ f +1 / ] (10 erg) 12 17 66 23 211 t c [ f +1 / ] (Myr) 78 51 1.8 86 7.5For two tempareture models, the average n e is estimated as M x / ( µm p V ) and used in text. a K = (10 − / πD ) R n e n p dV b Volume filling factor scaling in [ ] in Column 1. c A 2.3 kpc radius spherical emission volume. d A disk with the thickness 200pc and 2.3 kpc radius. − . . r m a li z ed c oun t s / s e c / k e V − . . r e s i dua l s channel energy (keV) Fig. 6.—
The X-ray spectrum of the residual emission from thecentral 6 ′ × ′ region of NGC 2403. Data from only the first observa-tion are included. The best-fitting model, a two-temperature apec model with hydrogen column density fixed to the galactic value,and the fit residuals are shown. The spectrum has been re-binnedto have a width of 73 eV for display purposes. gas densities in SNRs and H ii regions are comparablebut that the SNR temperature is somewhat higher re-sulting in higher pressures, P ∝ nT X , by about a factorof 2. This is not suprising since the SNR, which musthave ages of order 10 yr or less (Matonick et al. 1997),are expected to be further from pressure equilibrium withtheir surroundings than the fully-developed H ii regionswhich have ages of 10 to > yr.Another difference between the SNRs and H ii regionsis the amount of hot gas estimated from the X-ray data.There is, on average, about 300 ±
100 M ⊙ of hot gas per SNR compared to about 1000 M ⊙ per H ii region. Ofcourse, H ii regions are composed of many stars and SNswhereas the SNRs identified by Matonick et al. (1997)are morphologically consistent with single events. Evi-dently, SNe occurring within the relatively tenuous andhot interior of H ii regions are much less efficient at pro-ducing hot gas than are isolated SNe. The ratio of thehot gas mass to the swept up mass for SNRs is there-fore ∼ − ii regions it is only ∼ − . This indi-cates, again, that shock heating and thermal evaporationfrom the swept-up shells of young isolated SNRs is moreefficient than in H ii regions.The diffuse (residual) emission has a density and pres-sure lower by about a factor of 10 (assuming the sphericalvolume) compared to SNRs and H ii regions but a tem-perature that is comparable to both these sources. Ifwe assume a disk geometry, then the density is closerto the density in SNRs and H ii regions. There is an-other parameter to consider. The volume filling factor, f , may be lower for this residual emission compared tothe SNRs and H ii regions because, by definition, thisresidual shares the galaxy disk with the much colderneutral and molecular gas and dust. Assuming f ∼ ii regions and pressure equilibrium suggests afilling factor f ∼ (0 . − .
3) for the residual emissionwith correspondingly higher density and shorter coolingtime. Alternatively, we may assume pressure equilib-rium between the hot residual gas component and thecolder gas and dust. This cold gas resides primarily inclouds of typical temperatures of T ∼
100 K and den-sities of 60 −
100 cm − . This corresponds to a typicalpressure P/k ∼ cm − K which would imply, at theobserved temperature of (2 − × K, a hot gas density2of ∼ − − consistent with the derived value(Table 8) if f ∼
1. We note (Fraternali et al. 2002b) thatthe mean H i density in the central regions of NGC 2403is about 0.4 cm − . Thus its filling factor must be onlyaround 5% for a cloud density of 80 cm − . A similarargument can be made for the H content (Sheth et al.2005). Thus the filling factors for the cold gas may bequite small so that the bulk of the volume in the cen-tral regions of NGC 2403 may be filled with hot gas athigh filling factor ( f ∼ ii regions which is notwholly unreasonable.In any case, the mass of hot gas in this residual com-ponent, M X . × M ⊙ , is a small fraction of themass contained in H i (5 × M ⊙ ) and comparable tothe molecular (8 × M ⊙ ) gas mass.We cannot tell with certainty where this X-ray emit-ting residual gas is located. There is not a strong spatialcorrelation between this gas and star-formation indica-tors including H α and UV emission. Combining this factwith an estimated cooling time of up to 340 Myr (de-pending on the geometry assumed and the filling factor)allows for several possibilities. The residual gas may havebeen created in situ during star-formation activity thatended up to 100 Myr ago (in regions currently lackingstrong UV emission) or it may have been produced morerecently but has escaped from its place of origin eitherinto the halo or into low-density regions of the disk.At this point, we only have crude estimates of the agesof the various star-forming regions that are present in thecentral regions of NGC 2403. The ages of H ii regions are .
10 Myr, equivalent to the lifetimes of the least-massiveLyC-producing stars. The ages of UV-emitting regionscan be as old as 100 Myr. What is needed is a betterestimate of the age of individual star-forming regions.From this, we can better estimate how X-ray propertiesevolve through time and so better constrain the originof the residual X-ray emission – whether it traces recentstar-formation activity or is a relic of past activity. Thisis the subject of the next section. INDIVIDUAL STAR-FORMING REGIONS AND THEIRANALYSIS
We now turn to the analysis of individual young star-forming regions that have well-defined ages and masses.We choose young regions because they are more likelysources of X-ray emission from hot gas than are olderclusters; and have more likely retained their identity asa coherent (though not necessarily bound) and coevalsystem under the prevailing tidal forces. Theoretically,X-ray emission from stellar winds and hot gas associ-ated with supernovae (SNe) occurs only within the first ∼
40 Myr in the lifetime of a star cluster correspondingto the longest-lived stars that become core collapse SNe,roughly the 8 −
10 M ⊙ stars. Star clusters in this agerange emit strongly in UV light and (during the first ∼
10 Myr) in H α line emission. However, these star-formation indicators are sensitive to extinction by overly-ing dust. Dust heated by UV and optical radiation emitsthis energy in the mid-IR and longer wavelength bands.Therefore, we define our regions as those bright in bothUV radiation from massive young stars and in mid-IRradiation from reprocessing by dust. In this way, we canbetter estimate the extinction along the line of sight us- ing the mid-IR flux and thus correct the UV emission toobtain a better estimate of the age and mass of the starclusters.Star-formation tends to propagate spatially by com-pressing surrounding colder regions of the interstellarmedium through the action of massive star winds andsupernovae. Thus, small isolated regions are more likelyto have a single characteristic age than are large ex-tended regions. Therefore, we apply our source-findingalgorithm to identify isolated star-forming regions quan-titatively rather than to rely on a pre-determined ob-ject size (such as a fixed aperture or spatial grid, for in-stance, which is known to sample a range of stellar ages;cf., Calzetti et al. 2005). The drawback to our selectionmethod is that the flux from individual regions may beweak at wavelengths other than the UV and mid-IR weuse to select them. Complementary Observations
We use
GALEX
UV and
Spitzer mid-IR measurementsto identify young star-forming regions. We also use H α , atraditional star-formation tracer, in our analysis to con-firm a young age for the regions. GALEX observed NGC 2403 on December 5, 2003(Tilenum 5087) as part of the Nearby Galaxy Survey(Gil de Paz et al. 2007). Corrected intensity maps atFUV (1529 ˚A central wavelength) and NUV (2321 ˚A)were obtained from the
GALEX archive . The imageshave a pixel scale of 1.5 ′′ and a spatial resolution ofabout 5 ′′ . We applied the standard flux calibrations of1.40 × − ergs cm − s − cts − ˚A − to the FUV and2.06 × − ergs cm − s − cts − ˚A − to the NUV im-ages.NGC 2403 was observed with Spitzer
IRAC and MIPSas a part of the Spitzer Infrared Nearby Galaxies Survey(SINGS) legacy program (Kennicutt et al. 2003). MIPSimages at 24 µ m were taken on 2004 October 13 and16 (key 5549568 and 5549824, respectively). The IRAC3.6 µ m, and 4.5 µ m images were taken on 2004 October 8(key 5505792) and 12 (key 5505536). Final mosaiced im-ages provided by the SINGS program were used here. De-tails of their data reduction and calibration are reportedin the SINGS data delivery paper . The 24 µ m imagehas an ∼ ′′ resolution which is comparable to GALEX .The final mosaiced image has a pixel scale of 1.5 ′′ . Boththe IRAC images have a spatial resolution of ∼ ′′ and apixel scale of 0.75 ′′ .As part of the SINGS project, NGC 2403 was ob-served with the Kitt Peak National Observatory 2.1mtelescope in H α bands on 2001 November 8. We obtaineda continuum-subtracted H α image, made by subtractingan R -band image from a narrow-filter image at H α (D.Calzetti, private communication). The pixel scale of theH α image is 0.3 ′′ and the resolution is about 1 ′′ . TheH α flux is estimated from this image by assuming the[NII]/H α ratio and final H α correction factor as given inPrescott et al. (2007).As with the X-ray data, SN 2004dj was used to de-fine a common registration among the data sets. Un-saturated, foreground stars were used to register pre- http://galex.stsci.edu/GR2/ http://data.spitzer.caltech.edu/popular/sings/20070410 enhanced v1/Documents/sings fifth delivery v2.pdf Fig. 7.—
Identification of star-forming regions in NGC 2403.Top: The central 6 ′ × ′ region of NGC 2403 at FUV ( GALEX λ µ m are encircled in black. The sizes of the circles in-dicate photometric apertures defined as the 3 σ width of a circularGaussian model fitted to the respective surface brightness distri-bution. The gray scale is the same as the right panel of Figure 5.Middle: The same region at 24 µ m ( Spitzer ). Symbols same as forthe FUV image. The gray scale indicate intensity levels in unitsMJy str − for 24 µ m image. One pixel corresponds to 1 . ′′ µ m imagesoverlaid on the continuum-subtracted H α image. Sixteen of the19 sources were also selected as H ii regions as described in § − pixels − ,which corresponds to 2.1 × − Jy pixels − . The image has apixel scale of 0 . ′′ supernova ground-based, and GALEX images.
Star-forming Region Definitions
The same source finding tool applied to the X-ray im-ages was used to identify candidate star-forming regionsin the FUV and 24 µ m images. This tool is optimized fordetecting point sources but can be applied to moderately-extended sources by increasing the characteristic size ofthe model PSF parameter. We used a circular GaussianPSF in the application because most of the star-formingregions appeared centrally-peaked at the moderate res-olution of the 24 µ m and FUV images. We examinedtheir radial profiles post facto using the higher resolutionH α images to confirm that they are basically extendedand either centrally-peaked or centrally-cratered so thatthe Gaussian model is adequate for estimating both theemission centroids and the spatial extent.We detected 58 bright sources in the central 6 ′ × ′ FUVimage (Figure 7, top panel) defined as those with a
S/N above 10 and with a minimum of 15 source counts perunit uncertainty in the background. Similarly, 47 brightsources are detected in the 24 µ m image (Figure 7, middlepanel) with a S/N above 50 and with 100 or more elec-trons per unit uncertainty in the background. There are19 sources common to both the FUV and 24 µ m images.These 19 regions are shown in Figure 7 (bottom panel)overlaid on the H α image. (Sixteen of these regions wereselected as H ii regions based on their H α brightness in § S/N . Note (Figure 7) that thereare sources brighter than these 19 in one or the other ofthe two selection wavebands, but not in both. Examin-ing all available images, none of the sample regions haveproperties of foreground stars or of background AGN.The observed luminosities and angular sizes of the 19regions are listed in Table 9. We performed aperturephotometry on the individual sources with correspond-ing aperture sizes for each region and waveband. Al-though these 19 sources are extended, in order to de-fine sizes of aperture, we apply the circular Gaussianmodel. The radii in each region and each band are de-termined as 3 σ circular Gaussian widths, which contains99% of flux of the sources. For most regions, a sur-rounding annulus that extends to twice the source’s aper-ture radius was used for background. In more crowdedregions, a non-contiguous nearby region was used in-stead. Luminosities are computed using definitions simi-lar to those used by the SINGS team (e.g., Calzetti et al.2005); specifically, L band = λL λ,band for broad bands and L band = δλL λ,band for narrow bands. Luminosities werecorrected for the Galactic extinction along the line ofsight, E B − V = 0 .
040 mag (Schlegel et al. 1998), withthe Cardelli et al. (1989) dust model. The aperture cor-rection values listed on the
Spitzer web page were appliedto the 24 µ m data. Thumbnail images in the FUV and24 µ m bands are shown in Figure 8. The circles shown inthese images define the adopted radii in each band. Notethat they are rarely concentric. Also shown in Figure 8are the H α radial profiles for each region.The sizes of the regions in the FUV band are system-atically smaller than in the H α and 24 µ m bands. The4 Fig. 8.— ′′ × ′′ close-up views of the 19 star-forming regions. FUV images are on the left and 24 µ m images are in the middle.Circle locations, sizes, colors, and gray scales are the same as in Figure 7. On the right are higher-resolution H α radial profiles shown withbest-fitting Gaussian model curves. Each radial profile extends to 5 Gaussian widths. One pixel corresponds to 0. ′′ natural interpretation is that the FUV originates fromyoung star clusters whereas the source of H α and 24 µ mis re-radiation from surrounding warm ionized gas andcold dust regions, respectively, that lie at the outskirts ofwind-blown bubbles created by the stars. This is also thereason why the regions are not concentric in the differ-ent wavebands. The H α radial profiles, Figure 8, whichoften show a crater-like morphology is also indicative ofthis shell-like structure. Age, Mass, & Extinction Determinations
The five wavelength bands FUV, NUV (both from
GALEX ), 3.6 µ m, 4.5 µ m, and 24 µ m (all from Spitzer )were used to define the spectral energy distributions(SEDs) of light from the individual regions. These werecompared to theoretical spectra of instantaneous star-bursts based on the models of Leitherer et al. (1999)as improved by V´azquez & Leitherer (2005) by convolv-5
TABLE 9Luminosities and sizes of 19 star-forming regions
SRC L F UV R F UV L NUV L . µm L . µm L µm R µm L Hα R Hα L obs X erg s − arcsec 10 erg s − erg s − erg s − erg s − arcsec 10 erg s − arcsec 10 erg s − SF J073623.1+653649 4.8 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note . — FUV, NUV, and H α luminosities are Galactic exinction corrected. TABLE 10Derived values of 19 star-forming regions
SRC Mass Age A v N H L w L intF UV L int H α L int X M ⊙ Myr 10 cm − erg s − erg s − erg s − erg s − SF J073623.1+653649 0.4 9 0.67 1.7 5.2 9.8 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note . — Listed N H values include a Galactic component of N H = 0.4 × cm − . ing these spectra with the spectral response functionsof the appropriate filters. For computing the theoreti-cal spectra, we assumed the Kroupa (2001) initial massfunction (IMF), namely a power-law IMF on the range0.1 −
100 M ⊙ with a break at 0.5 M ⊙ and slopes of 1.3below and 2.3 above the break. A solar metallicity wasassumed; consistent with estimates from the literature(e.g., Martin & Belley 1996; Garnett et al. 1997). Weaccounted approximately for the wavelength-dependentattenuation by dust by using the starburst dust modelof Calzetti (2001). We modeled the re-emission of thisradiation in the mid-IR by assuming energy conserva-tion and a blackbody emission profile at a temperatureof 75 K. This temperature was estimated by fitting ablackbody curve to the Spitzer
MIPS 24, 70, and 160 µ m measurements from the brightest of our star-forming re-gions. (This region is fully resolved and isolated evenin the 160 µ m image so that a distinct local backgroundcould be identified.) Varying the blackbody temperatureby ∼
30% has little impact on our results. However, muchhigher or lower temperatures result in unrealistic esti-mates of the age and mass of several of the star-formingregions.By using five-band SEDs we are thus able to samplethe UV region which is most sensitive to age in youngstar clusters yet is most affected by dust attenuation, thenear-IR at 3.6 and 4.5 µ m that is least affected by dustand hence is sensitive primarily to cluster mass, and themid-IR 24 µ m which lies within the dust emission bandthat extends from several microns out to much longer6wavelengths. The 24 µ m waveband is longward of thePAH emission features which greatly complicate the dustemission spectrum (e.g., P´erez-Gonz´alez et al. 2006) butshortward of the lower spatial resolution long wavelength Spitzer bands.We used a least-squares fit of the model SEDs to theobserved luminosities using equal weights for each spec-tral band. Operationally, the model extinction and clus-ter mass parameters were allowed to vary, at a fixed clus-ter age, until a minimum χ was found. We then re-peated this process for discrete cluster ages ranging from1 to 200 Myr (in steps of 1 Myr for ages ≤
20 Myr andsteps of 10 Myr otherwise) to obtain the overall best-fitting age, mass, and extinction combination. That is,we do not interpolate the model spectra to intermediateages.The results are summarized in Table 10. In additionto the best-fitting mass, age, and extinction values basedon the SEDs fitting, Table 10 also includes the corre-sponding hydrogen column density, N H = ( A V /R V )5 . × cm − , assuming an optical parameter R V = 3 .
1, andthe corresponding observed H α , and FUV luminositiesafter correcting for extinction.Among the most massive young clusters in our sampleare the giant H ii regions NGC2403-I and NGC2403-IIstudied by Drissen et al. (1999). Drissen et al. estimatethe ages of these clusters as between 2 and 6 Myr basedon their detection of Wolf-Rayet stars in these regions.A crude estimate of the initial cluster mass can be madefrom the number of WR stars detected. This estimate isvery sensitive to the upper mass cutoff assumed of theinitial mass function and, of course, to the true age ofthe cluster. Interestingly, the age and mass values ob-tained from the studies of Drissen et al. are consistentwith our findings and independently validates our SEDfitting method. Garnett et al. (1999) estimated line-of-sight extinction to several of our sample star-formingregions from Balmer line ratios. Our estimates are inagreement with these results within A V of 0.3 mag. Forexample, we derive A V = 0 . ii region, compared to A V = 0 . A V = 0 . < >
50 Myrwhich we designate young, intermediate, and old, respec-tively. We believe the two regions designated as old arelikely younger than their best-fitting ages. Inspection ofthe distribution of the χ fit statistic shows, for thesetwo old regions, that there is a second local minimum at13 Myr which is probably a more realistic age estimate.Note that they have moderately high FUV luminosities,by selection, which, for their fitted age, forces their best-fitting mass to be very high relative to the other groups.There is not a similar trend in the fitted extinction val-ues.There are distinct differences among the members ofthe remaining two age groups. Members of the younggroup are typically much more luminous, per unit mass,at UV, H α , and 24 µ m compared to the intermediateage group members. Based on the star cluster model,the UV luminosity is expected to rise slightly during the first 4 Myr then drop rapidly as the most massive starsbecome supernovae. Similarly, the H α luminosity, whichcomes from recombination of circumstellar gas photoion-ized by massive stars, also drops quickly as these massivestars disappear. The observed trends are consistent withthis scenario.Dust emission, accounting for most of the 24 µ m lumi-nosity, also represents re-radiation of starlight. It is mostsensitive to UV but also responds to longer wavelengthlight. One would expect, therefore, that the 24 µ m lu-minosity per unit mass would decrease more slowly thanwould the UV and H α luminosities. However, as thecluster evolves and its wind-driven bubble expands, sur-rounding dust clouds will evaporate and the solid anglesubtended by remaining dust clouds will decrease. Thiscould account for the rapid drop with age observed inthe 24 µ m luminosity.These trends do not apply to the near-IR 3.6 and4.5 µ m luminosities. Light at these wavelengths comesdirectly from stars of practically all masses (ages) and isnot strongly affected by dust. Thus, the near-IR lumi-nosity per unit mass is nearly independent of the age ofthe underlying star cluster; as observed.One puzzling result is that the young star forming re-gions in our sample are much more massive than theintermediate-age regions. The average mass of the youngregions is 1.9 × M ⊙ and only 0.5 × M ⊙ for theintermediate-age regions. Notably, there is one regionthat is 3 times as massive as any other. Excluding thisregion, the average mass of the young group is still twicethat of the intermediate group. This may be an indi-cation of a selection bias. If more massive star clus-ters more efficiently destroy or disperse surrounding dustclouds, then our selection criterion that requires regionsbe bright in both UV and 24 µ m would tend to selectagainst more massive clusters. Naturally, the destruc-tion of dust takes time so this bias may only work againstmassive intermediate-age star-forming regions. Compari-son of the two panels of Figure 7 shows that there are sev-eral regions bright in UV that lack a strong IR counter-part. In particular, there is a very UV-bright region justbelow the center of the field at L intFUV ∼ × erg s − .This cluster was not selected because it lacks strong dustemission. It is likely of intermediate age as it also is weakin H α emission. Assuming an age of 10-15 Myr suggestsa mass in the range of about 2 × to 4 × based onthe values of L intFUV for the clusters listed in Table 10.This makes this cluster more massive than all but thetwo most massive young clusters.Crude estimates of the star-formation rate in the cen-tral regions of NGC 2403 can be made from the val-ues listed in Table 10. The total mass in stars in theyoung age group clusters is 1.1 × M ⊙ . These starswere formed in the recent 2 Myr interval for a mean star-formation rate of 0.5 M ⊙ yr − . Adding the contributionfrom the intermediate-age group gives 1.7 × M ⊙ ofstars formed over a 16 Myr period or a 0.1 M ⊙ yr − av-erage rate. These rates can be compared to rates deducedfrom H α luminosities. For the young age group clustersthis rate is only 0.1 M ⊙ yr − and is only slightly higherwhen the intermediate-age clusters are included (see Ta-ble 10). The two methods give quite different estimates.It is unlikely that the escape fraction of ionizing radia-tion is large enough to compensate for the discrepancy.7 − . r m a li z ed c oun t s / s e c / k e V − . . r e s i dua l s channel energy (keV) Fig. 9.—
Co-added spectrum of the 6 young star-forming regions.The best-fitting model, an absorbed two-temperature apec modelwith the hydrogen column density allowed as a free parameter, andthe fit residual are also shown. The spectrum has been re-binnedto have a width of 73 eV for display purposes.
Finally, the values can be compared to the galaxy-widestar-formation rate of 1.3 M ⊙ yr − based on the H α lumi-nosity and assuming a nominal A (H α ) = 1 mag reportedby Kennicutt et al. (2003). This suggests that 10% orup to 38% of the current galaxy-wide star formation isoccurring within these 19 star clusters. This is not anunreasonable estimate since the brightest few clusters inthe sample clearly dominate the relevant emission fromNGC 2403. Comparison to X-ray Properties
In most cases, the X-ray emission from these individ-ual star-forming regions is too faint to be detected as adiscrete source using our standard source-detection al-gorithm. Therefore, we define the X-ray source radiusfor each region to be the largest of the values deter-mined from fitting the surface brightness distribution inthe UV, mid-IR, and H α bands. Furthermore, we com-bine (stack) the spectra of the young and of the inter-mediate age regions to accumulate sufficient counts forfitting of the spectra of these two age groups. For thispurpose, we used the first observation only, in which allregions are imaged on the S3 CCD. We used the samebackground as was used for analysis of the residual emis-sion ( § § A V from theSED fitting ( § N H , namely 1.4 × cm − for the young and1.1 × cm − for the intermediate regions. The fittingresults are shown in Tables 11 and 12 for the young andintermediate age star-forming regions, respectively. Thespectra with best-fitting models are shown in Figure 9for the young regions and Figure 10 for the intermediateregions.Not surprisingly, the star-forming regions have X-raytemperatures and densities similar to the H ii regions se-lected by their H α brightness (Table 6 and § − − − . r m a li z ed c oun t s / s e c / k e V − × − × − r e s i dua l s channel energy (keV) Fig. 10.—
Co-added spectrum of the 11 intermediate-agestar-forming regions. The best-fitting model, an absorbed one-temperature apec model with the hydrogen column density allowedas a free parameter, and the fit residual are also shown. The spec-trum has been re-binned to have a width of 73 eV for displaypurposes. ble 4) and the densities are higher than deduced for theresidual emission (Table 8). The largest differences arefound for the intermediate-age group: the derived elec-tron density and pressure are a factor of two lower thanthat of the H ii regions. For models allowing the interven-ing column density to vary, the resulting N H are compa-rable to the Galactic value or about a factor of ten lowerthan for any of the other source populations consideredincluding the young star-forming region group.These results indicate that the star clusters begin toshow their age even after just 10 −
20 Myrs. Their hotgas densities and hence pressures begin to fall althoughtheir temperatures remain a moderate 2 − § −
20 Myrs so that the traditional current star-formation tracer, H α , will no longer be strong.We can estimate the X-ray luminosities of the individ-ual star-forming regions by scaling by the ratio of countsin the stacked spectra (for each age group) to the num-ber of counts detected in individual regions. The resultsare listed in Table 10. These luminosities have been cor-rected for extinction local to each region using the valuesof N H tabulated in column 7 of Table 10 (derived fromtheir best-fitting A V ) and for Galactic extinction. Theluminosities were computed using the two temperaturemodel with variable N H for the young region group, andthe one temperature model with variable N H for the in-termediate age regions.Figures 11 through 13 compare these intrinsic X-rayluminosities to the (extinction-corrected) FUV, 24 µ m,and H α luminosities, respectively. In all cases, there isonly a weak correlation between X-ray luminosity andthese star-formation indicators. This is due to the largescatter in the X-ray luminosities.A simple calculation shows that in one or two regionsa relatively high X-ray luminosity may be the result of afaint undetected X-ray point source in the region. Fromthe XLF, § TABLE 11Young regions X-ray fitting parameters
Fit Parameters apec apec apec apec+apec apec+apec N H (10 cm − ) 0.4 1.4 4.7 +1 . − . +2 . − . T e (MK) 3.4 +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . Normalization K a +0 . − . +0 . − . +1600 − +0 . − . +111 − T e (MK) 58 +690 − +1 . − . Normalization K +1 . − . +2 . − . L X /10 erg s − +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . L int X /10 erg s − b V [ f ] (10 cm ) 2.0 2.0 2.0 2.0 2.0 n e [ f − / ] (cm − ) 0.03 0.04 0.13 0.54,0.02 0.13,0.03 P/k [ f − / ] (10 K cm − ) 1.8 2.2 5.5 28 4.2 M x [ f +1 / ] (10 M ⊙ ) 0.6 0.8 3.1 1.3 3.5 E th [ f +1 / ] (10 erg) 0.7 0.9 2.3 2.7 3.4 t c [ f +1 / ] (Myr) 40 21 2.5 90 6.2 a K = (10 − / πD ) R n e n p dV b volume filling factor scaling in [ ] in Column 1. TABLE 12Intermediate age regions X-ray fitting parameters
Fit Parameters apec apec apec N H (10 cm − ) 0.4 1.1 0.4 +5 . − . T e (MK) 3.1 +0 . − . +0 . − . +0 . − . Normalization K a +0 . − . +0 . − . +51 − . L X /10 erg s − +0 . − . +0 . − . +0 . − . L int X /10 erg s − b V [ f ] (10 cm ) 4.0 4.0 4.0 n e [ f − / ] (cm − ) 0.01 0.02 0.01 P/k [ f − / ] (10 K cm − ) 0.8 1.0 0.8 M x [ f +1 / ] (10 M ⊙ ) 0.6 0.8 0.6 E th [ f +1 / ] (10 erg) 0.7 0.9 0.7 t c [ f +1 / ] (Myr) 60 50 60 a K = (10 − / πD ) R n e n p dV b volume filling factor scaling in [ ] in Column 1. with 0.5 − . ′ L − . .) Comparison to Theoretical Expectations
Stellar winds and supernovae within a young star clus-ter create a hot low-density cavity in the interstellarmedium (ISM) which persists beyond the ∼
40 Myr life-times of the OB stars in the cluster. The cumula-tive effect of winds and individual SNe is to graduallysweep the ISM into a thin dense shell analogous to theshell around a stellar wind bubble. The simple physi-cal model of a stellar wind bubble (Castor et al. 1975;Weaver et al. 1977) can be taken over directly to de-scribe, qualitatively, these large H ii regions or superbub-bles associated with OB associations (McCray & Kafatos1987; Mac Low & McCray 1988). In this model, massivestars inject kinetic energy into their surroundings creat- ing a freely expanding wind. Hot shocked gas surroundsthis free wind region and occupies most of the volume ofthe bubble. Surrounding this hot gas is the dense coldswept-up shell. Beyond the shell is the ambient interstel-lar medium. Diffuse X-ray emission from star-formingregions arises from the tenuous shocked gas in the bub-ble interior and from higher density gas evaporated fromthe shell. The dense evaporated component may domi-nate the X-ray emission which scales as the square of thedensity.Although analytic versions of the original basic modelhave been applied widely in the literature, we note thatnumerous simplifying assumptions are made which maybe poor approximations in reality. Among these arespherical symmetry, constant energy injection rate, adia-baticity, neglect of magnetic fields and of turbulence, anda homogeneous ambient medium. For our present pur-poses, the most critical of these is the assumption of aconstant energy injection rate or mechanical luminosity, L w . Numerical simulations (e.g., Strickland & Stevens1999) show that the expected X-ray luminosity is roughlyproportional to L w and that L w is far from constant.The starburst synthesis models we have used here(Leitherer & Heckman 1995) show that L w (injected bywinds from stars of initial mass >
15 M ⊙ and by suitablytime-averaged SN explosions) is a function of the age ofthe star-forming region, the mass of the star cluster, andthe assumed IMF. Specifically, only stellar winds frommassive stars contributes to the mechanical luminosityduring the first 3 Myrs. Then supernova explosions be-gin to contribute resulting in an increase in L w by abouta factor of 4 from an age of 3 to 6 Myrs. The lumi-nosity decreases again as the contribution from stellarwinds declines due to the decreasing number of O stars.From ∼
10 Myrs and up to the last supernova, which isaround 40 Myrs, L w is almost constant then finally fallsquickly to very low values. The young and intermedi-ate star-forming regions studied here have ages in therange where L w changes significantly on short timescalesso that our estimates of L w (Table 10) are reliable onlyto about a factor of four.Figure 14 displays the (extinction-corrected) X-ray lu-9 Fig. 11.— L FUV vs. L X for the 19 star-forming regions. Lumi-nosities are corrected for intrinsic and Galactic extinction. Trian-gles represent young (less than 2 Myrs), circles intermediate (2 − ± × − . Fig. 12.— L µ m vs. L X for the 19 star-forming regions. Lumi-nosities are corrected for intrinsic and Galactic extinction. Sym-bols are the same as in Figure 11. The straight line indicates thebest-fitting linear function with a slope of (1.9 ± × − . Fig. 13.— L H α vs. L X for the 19 star-forming regions. Luminosi-ties are corrected for intrinsic and Galactic extinction. Symbols arethe same as in Figure 11. The straight line indicates the best-fittinglinear function with a slope of (3.5 ± × − . minosity estimated in § § L X /L w .The resulting slopes are 0.0033 ± ± Chandra result ofthe second brightest H ii region in the local group, NGC604, whose L X /L w is estimated as 0.002 (T¨ullmann et al.2008).Numerical simulations (Strickland & Stevens 1999)found that L X /L w for a thin galactic disk model is0.5 − L X appropriate for the ROSAT band.This must be adjusted for the different sensitivity of
Chandra compared to
ROSAT for the observed spectralshape. Here, the characteristic electron temperature is2 − Chan-dra band (0.5 − ROSAT band (0.1 − L X /L w to 0.25 to 1% forthe Chandra band which is consistent with our results.Figure 15 displays the (extinction-corrected) X-rayemission against the estimated cluster age. Here, theX-ray luminosities are scaled to a fixed cluster massof 10 M ⊙ for easy comparison to L w ( t ). The solidcurve represents the best-fitting scaling, L X = 0 . L w ,from Figure 14 for the intermediate-age clusters and thedashed line is the factor-of-two higher scaling appropriatefor the younger clusters. Note the roughly factor-of-fourchange in L w at t ∼ L X /L w ratiois not a sensitive independent measure of the age of acluster. DISCUSSION OF INDIVIDUAL STAR-FORMINGREGIONS
We have examined 19 star-forming regions in the cen-tral part of NGC 2403. Their basic properties are listedin Tables 9 and 10. They range in age from 1 to 16 Myrwith two much older regions at 50 and 60 Myr. Theselatter ages are likely poor estimates as these two regionsare also strong H α sources which argues against an oldage. The remaining star-forming regions fall into distinctage groups; young (1-3 Myr) and intermediate (7 to 16Myr). The bulk X-ray properties of the hot gas withinthese two age groups, from model fits to their stackedX-ray spectra, are summarized in Tables 11 and 12, re-spectively.There is very little change in the X-ray temperaturesand very little change in the X-ray luminosity per unitstar cluster mass with age (Figure 15). There is also nosignificant evolution in the ratio of the X-ray luminos-ity to the mechanical luminosity (Figure 14). Thus, thebasic X-ray observables, temperature and luminosity, donot trace evolutionary changes for this sample of youngstar-forming regions.However, the density of the hot gas derived from thespectral fitting and hence the gas pressure in the youngstar clusters is about a factor of 2 higher than in the inter-mediate age clusters. This is consistent with the trendsdeduced previously ( §
2) where the youngest sources, the0
Fig. 14.—
Mechanical luminosity, L w calculated from Star-burst99 scaled by the mass and age of each region is shown againstthe intrinsic X-ray luminosity. Symbols are the same as in Fig-ure 11 (the two regions older than 20 Myr are omitted becausethey have very low mechanical luminosities according to Star-burst99). The lines indicate best-fitting slopes of (3.4 ± × − and (1.7 ± × − fitted to the young and intermediate-age re-gions, respectively. Fig. 15.—
Intrinsic X-ray luminosity is shown against age ofstar forming regions. Luminosites are scaled to an initial mass of10 M ⊙ . Symbols are the same as in Figure 11. Open diamondsdepict the average for young and intermediate regions. The solidcurve indicates the X-ray luminosity scaled from the mechanicalluminosity, L w , assuming L X /L w = 0 . L w assuming L X /L w =0 . SNRs, have the highest densities and pressures followedby the (H α -selected) H ii regions with the residual X-rayemission having the lowest density and pressure.There are three plausible scenarios for the nature of thehot residual gas consistent with these observed trends.(1) If the decrease in density is temporal, then the resid-ual emission comes from gas initially heated at timesconsiderably more than ∼
20 Myr in the past (the age ofthe intermediate group of star-forming regions); i.e., it isa relic of past star-formation activity. This is consistentwith the absence of a spatial correlation between trac-ers of recent star formation and this residual emission.(2) The hot residual gas may be escaping from activestar-forming regions into lower-density voids in the disk ISM. These low-density voids may be, for instance, lo-calized remnants of past star-forming activity. If thiswere the case, then we would expect some of this hotgas to be surrounding the star-forming regions analyzedhere. We checked this possibility by taking successivelylarger X-ray source sizes for the 19 star-forming regionsto estimate the true source extent. We found that forsizes up to twice the values adopted above (which werethe maximum radii estimated from the UV and mid-IRimages) there is a clear net X-ray excess (above the back-ground). Nevertheless, on larger scales, the morphologyof the residual hot gas does not correlate well with thestar-formation tracers. (3) The hot residual gas may havemoved out of the plane of the galaxy down the densitygradient into the halo. However, again, as there is noevidence from the distribution of the residual gas thatstrongly correlates it with star-forming regions, it mustbe a relic of past activity. In fact, there are no regions ofactive star formation that have the signature of blowoutfrom the disk. Blowout requires that the bubble size beroughly a few times the density scale height. The largeststar-forming regions in the central regions of NGC 2403are ∼
200 pc compared to a canonical scale height forlate-type spirals of 250 −
500 pc (see also the discussionin Strickland et al. 2004).We note that the age of the hot residual gasinferred from all of these scenarios is roughly thesame as the timescale for gravitational instabilities tocause shells surrounding OB associations to fragment(McCray & Kafatos 1987) at which point hot gas canvent out into the surroundings. It is also an age at whichstrong ionizing continua from massive stars has endedthough SNe may still be active. The current coolingtime for the residual gas exceeds 20 Myr and may beas high as 340 Myrs. Thus, regardless of the actual lo-cation of the residual gas (disk or halo) and the timeelapsed since it formed ( < −
20 Myr if escaping fromactive star-forming regions or perhaps factors of a fewlonger if remnants of past activity), this gas is likely tocontinue to radiate at low levels for a time longer thanthe characteristic timescales for localized star formation. SUMMARY
We have revisited the study of the current and recentstar formation activity in the central regions of normallate-type spiral galaxy NGC 2403 by re-evaluating the ex-isting X-ray data and including supporting observationsat other key wavebands. These include
Spitzer imagesthat trace dust heated by massive stars,
GALEX im-ages showing the young stars, and H α observations thatreveal nebulae photoionized by massive stars and SNe.Through analysis of this multi-wavelength data, we haveobtained estimates of the mass, age, and line-of-sight ex-tinction towards numerous young star clusters. With fewexceptions, these clusters and their environs are weak X-ray sources. Even the most powerful, a giant H ii regioncomparable to 30 Dor in the LMC and NGC 604 in M33,radiates at only ∼ × erg s − in the Chandra
X-rayband which is about 0.5% of its estimated mechanicalluminosity.We have also shown that, after carefully accounting forthe point-source (X-ray binary) population, SNRs andbright H ii regions, there remains a residual X-ray com-ponent pervading the central few kpc of NGC 2403. This1component cannot be accounted for by faint sources be-low our detection limit. It is likely diffuse hot gas but notstrongly correlated with current star-formation activity.It is likely a relic of star formation activity occurringsome 20 Myrs (the age of the intermediate age regions)or older.The geometry of the residual hot gas is not well de-termined, which leads to additional uncertainties in itsphysical properties. It is not clear whether this gas isconfined to the disk of the galaxy or resides in the halo.Nonetheless, it seems to be more centrally-located thanthe star-forming regions. Since the amount of hot gasrepresented by this residual emission is substantial, it im-plies that star-formation activity was much higher in thepast. Combined with its central location, this conclusionis consistent with the suggestion by Davidge & Courteau(2002) who suggested that an earlier episode of star for-mation occurred in the central region of NGC 2403 thathas now propagated outward in the disk to its presentradius of about a kpc from the galactic center. We gratefully acknowledge the anonymous referee forcareful reading and insightful comments that improvedthe paper. Support for this research was provided inpart by NASA through an Astrophysics Data Anal-ysis Program grant NNX08AJ49G and through theNASA/ Chandra
Award Number GO5-6089A issued bythe
Chandra
X-ray Observatory Center, which is oper-ated by the Smithsonian Astrophysical Observatory forand on behalf of NASA under contract NAS8-03060.This work made use of observations made with the
Chan-dra ; of observations made with the Spitzer Space Tele-scope, which is operated by the Jet Propulsion Labo-ratory, California Institute of Technology, under a con-tract with NASA; of observations made with the GalaxyEvolution Explorer, a NASA mission managed by theJet Propulsion Laboratory; and of ground-based obser-vations obtained as part of the Spitzer Legacy Scienceproject SINGS (Kennicutt et al. 2003) to which we aregreatly indebted.
REFERENCESAnders, E., & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53,197Argo, M. K., Muxlow, T. W. B., Beswick, R. J., Pedlar, A., &Marcaide, J. M. 2004, IAU Circ., 8399, 3Barnes, J. E., & Hernquist, L. 1992, ARA&A, 30, 705B¨oker, T., Laine, S., van der Marel, R. P., Sarzi, M., Rix, H.-W.,Ho, L. C., & Shields, J. C. 2002, AJ, 123, 1389Bregman, J. N. 1980, ApJ, 236, 577Calzetti, D. 2001, PASP, 113, 1449Calzetti, D., et al. 2005, ApJ, 633, 871Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345,245Castor, J., McCray, R., & Weaver, R. 1975, ApJ, 200, L107Cole, S., Lacey, C. G., Baugh, C. M., & Frenk, C. S. 2000,MNRAS, 319, 168Davidge, T. J., & Courteau, S. 2002, AJ, 123, 1438Drissen, L., Roy, J.-R., Moffat, A. F. J., & Shara, M. M. 1999,AJ, 117, 1249Fraternali, F., & Binney, J. J. 2006, MNRAS, 366, 449—. 2008, MNRAS, 386, 935Fraternali, F., Cappi, M., Sancisi, R., & Oosterloo, T. 2002a,ApJ, 578, 109Fraternali, F., Oosterloo, T., & Sancisi, R. 2004, A&A, 424, 485Fraternali, F., van Moorsel, G., Sancisi, R., & Oosterloo, T.2002b, AJ, 123, 3124Garnett, D. R., Shields, G. A., Peimbert, M., Torres-Peimbert, S.,Skillman, E. D., Dufour, R. J., Terlevich, E., & Terlevich, R. J.1999, ApJ, 513, 168Garnett, D. R., Shields, G. A., Skillman, E. D., Sagan, S. P., &Dufour, R. J. 1997, ApJ, 489, 63Gil de Paz, A., et al. 2007, ApJS, 173, 185Heckman, T. M., Kauffmann, G., Brinchmann, J., Charlot, S.,Tremonti, C., & White, S. D. M. 2004, ApJ, 613, 109Karachentsev, I. D., Karachentseva, V. E., Huchtmeier, W. K., &Makarov, D. I. 2004, AJ, 127, 2031Karachentsev, I. D., et al. 2002, A&A, 383, 125Kennicutt, Jr., R. C. 1989, ApJ, 344, 685Kennicutt, Jr., R. C., et al. 2003, PASP, 115, 928Kent, S. M. 1987, AJ, 93, 816Kormendy, J., & Kennicutt, Jr., R. C. 2004, ARA&A, 42, 603Kroupa, P. 2001, MNRAS, 322, 231Leitherer, C., & Heckman, T. M. 1995, ApJS, 96, 9 Leitherer, C., et al. 1999, ApJS, 123, 3Mac Low, M.-M., & McCray, R. 1988, ApJ, 324, 776Madore, B. F., & Freedman, W. L. 1991, PASP, 103, 933Martin, C. L., & Kennicutt, Jr., R. C. 2001, ApJ, 555, 301Martin, P., & Belley, J. 1996, ApJ, 468, 598Matonick, D. M., Fesen, R. A., Blair, W. P., & Long, K. S. 1997,ApJS, 113, 333McCray, R., & Kafatos, M. 1987, ApJ, 317, 190Moretti, A., Campana, S., Lazzati, D., & Tagliaferri, G. 2003,ApJ, 588, 696Noguchi, M. 2000, MNRAS, 312, 194Pannuti, T. G., Schlegel, E. M., & Lacey, C. K. 2007, AJ, 133,1361P´erez-Gonz´alez, P. G., et al. 2006, ApJ, 648, 987Prescott, M. K. M., et al. 2007, ApJ, 668, 182Schaap, W. E., Sancisi, R., & Swaters, R. A. 2000, A&A, 356, L49Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500,525Schlegel, E. M., & Pannuti, T. G. 2003, AJ, 125, 3025Shapiro, P. R., & Field, G. B. 1976, ApJ, 205, 762Sheth, K., Vogel, S. N., Regan, M. W., Thornley, M. D., &Teuben, P. J. 2005, ApJ, 632, 217Sicking, F. J. 1997, PhD thesis, PhD Thesis, University ofGroningen, (1997)Sivan, J.-P., Maucherat, A. J., Petit, H., & Comte, G. 1990,A&A, 237, 23Spitzer, L. J. 1990, ARA&A, 28, 71Strickland, D. K., Heckman, T. M., Colbert, E. J. M., Hoopes,C. G., & Weaver, K. A. 2004, ApJ, 606, 829Strickland, D. K., & Stevens, I. R. 1999, MNRAS, 306, 43Struck, C., & Smith, D. C. 2009, MNRAS, 1189Tennant, A. F. 2006, AJ, 132, 1372Thornley, M. D., & Wilson, C. D. 1995, ApJ, 447, 616T¨ullmann, R., et al. 2008, ApJ, 685, 919V´azquez, G. A., & Leitherer, C. 2005, ApJ, 621, 695Weaver, R., McCray, R., Castor, J., Shapiro, P., & Moore, R.1977, ApJ, 218, 377Yukita, M., Swartz, D. A., Soria, R., & Tennant, A. F. 2007, ApJ,664, 277
Fig. 8 Cont..— ′′ × ′′ close-up views of the rest 15 of 19 star-forming regions. FUV images are on the left and 24 µ m images are inthe middle. Circle locations, sizes, colors, and gray scales are the same as in Figure 7. On the right are higher-resolution H α radial profilesshown with best-fitting Gaussian model curves. Each radial profile extends to 5 Gaussian widths. One pixel corresponds to 0. ′′′′