OBSERVATIONAL CONSTRAINTS ON SUPERBUBBLE X-RAY ENERGY BUDGETS

For the purposes of background subtraction we used the blank-sky data sets provided as part of the Chandra CALDB, as the diffuse soft X-ray emission from both DEM L50 and DEM L152 covers a large fraction of the S3 chip in each observation. These blank-sky data sets comprise hundreds of kiloseconds of data per CCD chip, providing a high S/N estimate of the typical background experienced by the ACIS detectors. Since the net background experienced in the ACIS detector does change with time, the blank sky data was renormalized so that the mean X-ray surface brightness (excluding point sources) for each chip in the E = 3–6 keV energy band matched that in each observation. Typical variations in the hard X-ray background in the ACIS instrument are ∼2% (root mean square) from observation to observation (Hickox & Markevitch 2006).

We used a hard X-ray band to calculate the renormalization factor for the blank-sky data sets as soft diffuse X-ray emission is present in both the S2 and S3 chips in the two observations (see Figure 1). The majority of the superbubble emission falls within the field of view of the S3 chip in both observations. Part of DEM L50 spills over onto the S2 chip, however, and N44 shell 3 (an SNR; see Chu et al. 1993) is a strong soft X-ray source in the S2 chip of the DEM L152 observation. Excess soft X-ray emission is not apparent on the S4 chip in either observation.

Note that apparently diffuse hard X-ray emission in a superbubble has only been detected in 30 Doradus C (Smith & Wang 2004; Yamaguchi et al. 2010). Our method of renormalizing the blank-sky data sets means that there will be no significant diffuse X-ray emission in the E = 3–6 keV energy band, once point sources and the background have been subtracted. This approach is valid, in that a visual inspection of the raw, non-background-subtracted data reveals no evidence for hard X-ray emission within DEM L50 or DEM 152, and the mean source-free hard X-ray surface brightness in each chip of our observations is less than or equal to that in the blank-sky data. We cannot rule out the presence of diffuse hard X-ray emission in our observations of DEM L50 and DEM L152, but if such emission is present, it is very faint.

For DEM L50, the S3 chip needed essentially no renormalization, as the observed source-free E = 3–6 keV surface brightness differed from the blank-sky data sets by only 0.9% ± 1.8%. For the front-illuminated S2 and S4 chips the blank sky backgrounds were normalized downward by 5.9% (uncertainty 2.1%) and 14.9% (uncertainty 2.0%), respectively. It is not surprising that the hard background surface brightness in the blank-sky data might be higher than in our observations of DEM L50 and DEM L152. The blank sky data sets are constructed from a large number of data sets and thus from data experiencing a higher (but more normal) fraction of mild background flares than our observations.

For the DEM L152 observation the blank sky background was normalized down by 4.4% (uncertainty ±2.4%). For the S2 and S4 chips the blank sky backgrounds were normalized downward by 2.0% (uncertainty 3.0%) and 16.6% (uncertainty 2.8%), respectively.

The magnitude of the difference in the mean source-free X-ray surface brightness in the E = 3–6 keV energy band for the S4 chip in both observations is somewhat surprising. This CCD chip is known to suffer from significant flaws in its serial readout that leads to artificial events being registered (visible as streaks in images) that can only be partially screened out in software processing. We speculate that this effect is normally exacerbated by the unwanted particle events common in background flares, thus accounting for the pronounced difference between the blank sky data for the S4 chip and our observations.

We altered the coordinate system of the blank-sky data to match the point of the appropriate observation. Thus our background subtraction for image analysis consisted of subtracting an image created from the blank-sky data from an image from the real observational data, scaled by the ratio of total exposure times in the observation to the blank-sky data for the appropriate chip.

In Figures 2 and 3, we present adaptively smoothed X-ray images of DEM L50 and DEM L152 in a variety of soft X-ray energy bands, along with optical Hα+[N ii] imaging and H i column density maps. Figure 4 presents three-color composite images of the superbubbles that combine narrowband optical Hα+[N ii], [O iii], and soft X-ray emission.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Adaptive smoothing using the CIAO tool csmooth (Ebeling et al. 2006) attempts to differentially smooth an input image to achieve a relatively uniform local S/N, so that bright regions are smoothed with a smaller smoothing kernel than low surface brightness regions. We initially smoothed the diffuse image in the E = 0.3–2.8 keV energy band, after point-source removal but prior to background subtraction, with a target S/N of 3 but with a maximum smoothing kernel equivalent to a Gaussian of FWHM =40''. The smoothing map calculated in this step was then applied to all the background-subtracted soft X-ray diffuse images, so that any differences between images in different energy bands are intrinsic and are not due to the application of different smoothing maps.

Given claims that csmooth can generate spurious structure in adaptively smoothed images (Diehl & Statler 2006), we also investigated adaptively smoothed images using the completely independent adaptive smoothing algorithm implemented in the XMM-Newton data analysis software SAS as asmooth. This algorithm generated images very similar to those produced by csmooth, except that the stated statistical significance of individual features within the images differed significantly between the two methods. We also generated images using uniform smoothing with a two-dimensional Gaussian mask of FWHM =30'' (not shown) in order to conservatively verify the existence of features seen in the adaptively smoothed images.

The diffuse soft X-ray emission in DEM L50 is clearly confined within the optical superbubble and SNR shells (Figure 2(b)). The apparent X-ray surface brightness of the SNR N186 D (Oey et al. 2002) on the northern edge of the roughly elliptical superbubble is twice that of the larger superbubble. If we ignore N186 D and concentrate solely on the superbubble we see that the detected X-ray emission in the E = 0.3–2.8 keV energy band is relatively uniformly distributed, although with a slight tendency toward being brighter on the south and eastern interior edges of the optical shell.

The H i column density along the line of sight to DEM L50 peaks in the projected center of the bubble, however (see Figure 2(c)), raising the possibility that the weak limb brightening is due to central absorption rather than true enhancement of the X-ray emission at the shell walls.

That at least some of the H i lies between us and the X-ray-emitting plasma is clear from the way the X-ray-bright southeast limb of the superbubble fills in the gap in H i column density at that location (compare Figures 2(a) and (c)). Based on the observed velocities of the H i gas, Oey et al. (2002) show that the H i associated with DEM L50 is most likely the outer layer of the expanding shell of the bubble.

DEM L50 is most prominent in the E = 0.6–1.6 keV band. The SNR N186 D is prominent in both this energy band and the softer E = 0.3–0.6 keV energy band, despite being associated with a column density of N_H ∼ 10²¹cm⁻² of H i. This suggests that the SNR may be spectrally distinct from the superbubble. Neither the superbubble nor the SNR is apparent in the E = 1.6–2.8 keV energy band.

3.2.1. DEM L152

In contrast to the relative simplicity of DEM L50, the spatial structure of DEM L152 (N44) in the X-ray, optical and H i is complex (see Figure 4). The field of view of the Chandra observations (Figure 3) cover the following features identified in the ROSAT PSPC observations of Chu et al. (1993): Shell 1, the main superbubble, and parts of Shell 3, an SNR, and the South Bar, in particular the suspected break-out region where Shell 1 may be venting its contents through a break in its southern edge.

The relationship between the new Chandra observation and the features identified in prior X-ray observations is shown in Figure 5, where the Chandra ACIS field of view and contours of diffuse X-ray surface brightness are overlaid on a smoothed ROSAT HRI image of N44. The image was taken from the 108.5 ks HRI observation RH600913, obtained from the HEASARC data archive.

Zoom In Zoom Out Reset image size

Shell 1 is the most prominent X-ray source in the E = 0.3–2.8 keV energy band Chandra image, and the soft diffuse X-ray emission is particularly bright around the inner edge of the optical shell. This limb brightening is not an artifact of adaptive smoothing, as it is also clearly visible in images smoothed with a uniform FWHM =30'' Gaussian mask.

The limb-to-center brightness contrast is a factor of ∼2 in the E = 0.3–2.8 keV and E = 0.6–1.6 keV energy band images and is slightly larger in the softer E = 0.3–0.6 keV energy band image (Figure 3(d)).

This central decrement in X-ray surface brightness cannot be due to absorption by intervening gas as both the X-ray-bright limbs and the fainter center of Shell 1 lie within a region of lower H i column density (Figure 3(c)).

Shell 3 falls partly within the ACIS S3 and S2 chips, although its eastern edge is not covered in the current observations. It has a comparable X-ray surface brightness to Shell 1, even though the H i column density along this line of sight (N_H ∼ 5 × 10²¹cm⁻²) is approximately twice the value toward the center of Shell 1.

Bright, diffuse, soft X-ray emission also extends to both the west and south of Shell 1; the latter feature being known as the South Bar (Chu et al. 1993).

It is also noteworthy that both Shell 1 and Shell 3 are clearly detected in the E = 1.6–2.8 keV energy band, in contrast to the complete lack of emission in this energy band from DEM L50 and N186 D. The South Bar lacks appreciable emission in this energy band, which is consistent with the finding of Magnier et al. (1996) that the South Bar was spectrally softer than Shell 1.

As a first step to selecting regions for more detailed spectral analysis we constructed three-color images of the two bubbles where the resultant color is representative of the mean energy of the diffuse X-ray emission (Figures 4(b) and (e)) within the soft X-ray band. In these panels, red represents the lowest energy photons (E = 0.3–0.6 keV), green intermediate energy photons (E = 0.6–1.6 keV), and blue relatively harder emission in the E = 1.6–2.8 keV band.

As we shall later show, these figures somewhat exaggerate the spectral differences between the two objects and the degree of small-scale spectral variation within each object. Nevertheless they do serve some purpose by more clearly illustrating some of the possible spatial variation in the distribution of diffuse emission at different energies and help inform our choice of regions over which to study the spectral properties of the diffuse emission.

In Figure 4(b), the emission from the superbubble DEM L50 appears slightly spectrally harder than the brighter emission from the SNR N186 D, but there is only a marginal suggestion of spectral hardness variation within DEM L50 itself. Based on this and the optical morphology of the bubble, we choose to separate the X-ray data into two regions for spectral fitting: an elliptical region encompassing SNR N186 D and another elliptical region covering the superbubble but excluding the SNR. These are the regions shown in Figure 2(a). In order to examine spectral variations on smaller scales, we further subdivided DEM L50 into a Center and Limb region when performing spectral fits (see Section 4.2). Note that only the parts of those ellipses within the S3 chip are actually included within the spectral extraction regions.

The soft X-ray emission in the vicinity of DEM L152 shows more evident spectral variation than does DEM L50 (Figure 4(e)). Based on this and the optical morphology of the N44 nebula we chose spectral extraction regions similar to those used earlier by Chu et al. (1993) and Magnier et al. (1996): Shell 1 (the main bubble), Shell 3 (the SNR), and the South Bar, with most of the remaining emission associated with a new region we simply name West (see Figure 3(a)).

Diffuse X-ray count rates within the spectral extraction regions of both sets of observations are given in Table 2. Column 3 gives the area of each region. Columns 4–6 show the background-subtracted ACIS S3 count rates in the E = 0.3–2.8 keV energy band within each specified region, where R_tot gives the count rate for the full X-ray emission, R_diff gives the estimated diffuse emission count rate, and R_PS gives the total point-source count rate. Emission from point-like X-ray sources accounts for ∼14% of the total soft X-ray emission within the boundaries of the DEM L50 bubble, and less than 1% of the emission within the boundary of the DEM L152 Shell 1.

Given Chandra's high spatial resolution and sensitivity we should be able to determine whether the spectral properties of the diffuse emission vary on spatial scales smaller than the entire shell or bubbles. In Figures 4(b) and (e), we show energy-color-coded images of the diffuse emission around DEM L50 and DEM L152. Such images can be prone to smoothing-related artifacts, so we also mapped the hardness ratios in 1' × 1' pixels. These large pixels were required in order to minimize the statistical uncertainty in the hardness ratio in each pixel. The spectral hardness ratio, which will be discussed in more detail in Section 4.2, is defined as Q = (H − S)/(H + S), where H and S are the count rates in the specified hard (E = 0.6–1.6 keV) and soft (E = 0.3–0.6 keV) energy bands, respectively. Thus, higher values of Q correspond to a harder spectrum. Images of the hardness ratio itself tend to be difficult to interpret, as they can be dominated by noise in the lower signal-to-noise regions. For this reason we constructed maps of the statistical significance of the deviation of the local hardness ratio from the average hardness ratio $\overline{Q}$ over each bubble. For a given pixel with hardness ratio Q_pix and uncertainty σ_Q,pix, Figures 4(c) and (f) show $(Q_{\rm pix} - \overline{Q})/\sigma _{\rm Q, pix}$. In other words, the value in each pixel corresponds to the significance of the deviation in Gaussian σ. While Figures 4(c) and (f) hint at some of the intrinsic large-scale variations across the superbubbles, variations in column density and the resultant absorption of soft X-ray photons have a substantial effect on the observed hardness ratios. Figures 4(c) and (f) should only be used to identify regions that appear spectrally distinct from one another.

Although some spatial variation in the spectral properties of the diffuse X-ray emission is present in both objects, it is not particularly strong, and emission in the E = 0.6–1.6 keV energy band dominates both sets of observations. Hardness ratio variations are present on large scales, for example between the bubbles and the nearby SNR, or between Shell 1 and the Southern Bar region of DEM L152. On smaller angular scales (∼1') we find that any spatial variations in spectral hardness are of marginal statistical significance in our current X-ray data.

For the spectral fits, we removed background flares from both the S3 and S2 chips with the lc_clean routine in CIAO. We then used the renormalized ACIS stowed background files to subtract the approximately constant particle background from both chips. To renormalize the background files, we compared the relative count rates of the stowed background and our observations between 8 keV and 9.4 keV and renormalized the particle background to match the count rate of our data. The 8–9.4 keV energy range was chosen to exclude real X-ray photon events and a line feature above 9.4 keV.

We used the specextract script in CIAO to extract spectra from the specified regions and subtract the particle background. To account for the X-ray background, we fit the extracted spectra from the S2 chip, following Kuntz & Snowden (2000) and Henley et al. (2007), and used the resulting fits as a component of our subsequent models to the S3 data. A thermal plasma in collisional ionization equilibrium fits the local X-ray background, while two additional thermal plasmas in equilibrium are used for the Galactic background, and the cumulative emission from extragalactic sources is fit by a power law. Our best fits to the S2 background are shown in Table 3. The power-law index was fixed to the value found by Chen et al. (1997), and the column densities were obtained from HEASARC from Kalberla et al. (2005). Temperatures were left as a free parameter. The Local Hot Bubble temperature is given in Column 3, the Galactic H i column density is shown in Column 4, the temperatures of the two thermal components of the Galactic background are given in Columns 5 and 6, and the extragalactic power law index is shown in Column 7. Normalizations were re-scaled to account for the relative difference in area between the S3 extracted regions and the S2 chip. For the sake of comparison, we include two fits to the S2 background for DEM L50; the "Low S2" fit was a local minimum in χ² before the global best fit was found. The effect of varying the background fit can be seen by comparing the DEM L50 models with the "High S2" and "Low S2" background fits.

All extracted regions are shown in Figure 6. In addition to the regions in Figures 2(a) and 3(a), we extracted spectra for smaller subregions. We divided DEM L50 into Limb and Center regions a separate spectrum for the bright southern portion of the Limb. For DEM L152, we extracted additional spectra for the Bubble (that is, Shell 1) Limb, West Limb, and West Blowout regions as well as for the two knots shown in Figure 6. While some clumps seen in the smoothed image of DEM L152 are artifacts of the smoothing algorithm, these two knots are statistically significant (see Figure 4(f)) and appear when different smoothing algorithms are used.

Zoom In Zoom Out Reset image size

We fit each superbubble region and subregion with one- and two-temperature plasmas in collisional ionization equilibrium, using the XSPEC Astrophysical Plasma Emission Code (APEC) models (Smith et al. 2001) with variable abundances (Tables 4 and 6). We examined an additional, non-equilibrium ionization model for DEM L50 (Table 5). In all models, He, C, N, Al, and Ni abundances were fixed to the solar values. We varied the α elements O, Ne, Mg, Si, S, Ar, and Ca in proportion to one another, and Fe was left as a free parameter. Although absolute abundances are not well constrained, the α/Fe ratio is more reliable. To ensure that the fits were not local minima, we looked at possible values of N_H, temperature, and abundances spanning one to two orders of magnitude with the XSPEC command steppar. As a further check on the validity of our fits and to establish error bars for each variable, we ran the error command on the free parameters.

Table 4. Fit Parameters for DEM L50

Model	Bkgd Fita	NH (1022 cm−2)	kT1 (keV)	Norm1	kT2 (keV)	Norm2	Eff kT (keV)	α/O☉	Fe/Fe☉	$\frac{\alpha /\rm Fe}{(\alpha /\rm Fe_{\odot })}$	Red. χ2	dof
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
Center 1T	High S2	0.38+0.03−0.04	0.15+0.02−0.01	3.5E-03	⋅⋅⋅	⋅⋅⋅	0.15	0.88+0.87−0.35	6.00+12.87−3.42	0.15+0.05−0.04	1.743	211
Center 1T solar	High S2	0.31 ± 0.05	0.19 ± 0.01	1.3E-03	⋅⋅⋅	⋅⋅⋅	0.19	1.00	1.00	1.00	1.943	213
Center 2T	High S2	0.40+0.08−0.06	0.11+0.80−0.11	4.7E-03	0.17 ± 0.03	1.6E-03	0.13	1.10+1.18−0.60	6.03+10.61−3.35	0.18+0.14−0.05	1.757	209
Center 2T solar	High S2	0.32+0.06−0.09	0.11+0.04−0.02	3.1E-03	0.23+0.06−0.02	5.9E-04	0.13	1.00	1.00	1.00	1.835	211
Limb 1T	High S2	0.29 ± 0.05	0.18+0.02−0.04	2.7E-03	⋅⋅⋅	⋅⋅⋅	0.18	0.54+0.29−0.17	1.93+8.04−0.94	0.28+0.10−0.06	1.744	302
Limb 2T	High S2	0.32+0.03−0.02	0.08+0.20−0.08	1.4E-02	0.22+0.07−0.02	7.3E-04	0.09	1.25+0.62−0.36	2.78+1.51−1.37	0.45+0.27−0.17	1.718	300
Limb 2T solar	High S2	0.21 ± 0.04	0.11+0.04−0.02	1.4E-03	0.27+0.03−0.04	2.1E-04	0.13	1.00	1.00	1.00	1.749	302
Limb 1T	Low S2	0.36 ± 0.06	0.18+0.01−0.03	2.6E-03	⋅⋅⋅	⋅⋅⋅	0.18	0.95+2.70−0.39	3.38+13.92−1.95	0.28+0.12−0.08	1.793	302
Limb 2T	Low S2	0.34+0.07−0.11	0.14+0.26−0.14	1.4E-03	0.21+0.16−0.11	7.1E-04	0.16	1.37+2.50−0.71	3.58+9.85−2.05	0.38+0.28−0.16	1.799	300
Bright Limb 1T	High S2	0.37+0.07−0.08	0.19+0.02−0.03	1.9E-03	⋅⋅⋅	⋅⋅⋅	0.19	0.56+0.95−0.25	1.07+7.40−1.23	0.53+0.27−0.18	1.516	137
Bright Limb 2T	High S2	0.37 ± 0.07	0.11+0.15−0.11	1.8E-03	0.22+0.11−0.05	8.3E-04	0.14	0.80+2.35−0.46	1.21+2.72−0.72	0.66+0.44−0.37	1.532	135
SNR 2T	High S2	0.09 ± 0.02	0.19 ± 0.01	4.7E-04	0.72+0.09−0.05	2.6E-05	0.22	0.37+0.23−0.11	1.66+2.21−0.81	0.22+0.13−0.06	2.114	145
SNR 2T solar	High S2	0.20+0.05−0.03	0.08+0.01−0.08	6.2E-03	0.29 ± 0.02	2.8E-04	0.09	1.00	1.00	1.00	2.173	147
SNR 2T	Low S2	0.06+0.05−0.04	0.19+0.02−0.01	4.3E-04	0.73 ± 0.09	3.2E-05	0.23	0.27+0.12−0.06	1.26+1.50−0.54	0.22+0.11−0.06	2.150	145
SNR 2T solar	Low S2	0.25+0.04−0.06	0.08+0.01−0.08	9.9E-03	0.27+0.02−0.01	3.9E-04	0.09	1.00	1.00	1.00	2.255	148

Notes. Errors show the 90% confidence interval when one parameter is varied. ^aThe Low S2 background gave poor fits to the Center region which are not included.

Download table as:  ASCIITypeset image

Table 5. DEM L50 Non-equilibrium Fits

Model	Bkgd Fita	NH (1022 cm−2)	kT1 (keV)	Norm1	τ (s cm−3)	Ion. Time (yr)	α/O☉	Fe/Fe☉	$\frac{\alpha /\rm Fe}{(\alpha /\rm Fe_{\odot })}$	Red. χ2	dof
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Center NEI	High S2	0.31+0.05−0.08	0.28+0.07−0.06	5.1+12.0−3.3E-04	5.57E10 ± 4.68E10	22,000	0.76+0.36−0.24	0.74+0.36−0.26	1.04+0.27−0.20	1.838	210
Limb NEI	High S2	0.26+0.04−0.03	0.34+0.11−0.05	3.2+6.3−2.3E-04	3.31E10 ± 2.84E10	46,000	0.83+0.46−0.24	1.100.62−0.34	0.76+0.16−0.12	1.718	301
Limb NEI	Low S2	0.34+0.09−0.08	0.28+0.11−0.07	5.5+18.7−4.4E-04	6.26E10 ± 4.69E10	65,000	1.28+1.48−0.49	1.51+0.87−0.60	0.85+0.19−0.15	1.718	301
Bright Limb NEI	High S2	0.34+0.12−0.10	0.32+0.14−0.10	3.8+15.8−1.8E-04	8.33E10 ± 2.88E10	53,000b	0.70+0.69−0.36	0.58+0.55−0.27	1.21+0.38−0.27	1.543	136
SNR NEI	High S2	0.11+0.05−0.03	0.58+0.23−0.08	1.2+0.04−0.06E-04	2.60E10 ± 6.39E9	11,000	0.29+0.10−0.09	0.45+0.14−0.08	0.64+0.15−0.12	2.265	146

Notes. Errors show the 90% confidence interval when one parameter is varied. τ errors are estimates only. ^aThe Low S2 background gave poor fits to the Center region which are not included. ^bAssuming n_e ∼ 0.05 cm⁻³.

Download table as:  ASCIITypeset image

Table 6. Fit Parameters for DEM L152

Region and Model	NH (1022 cm−2)	kT1 (keV)	Norm1	kT2 (keV)	Norm2	Eff kT (keV)	α/O☉	Fe/Fe☉	$\frac{\alpha /\rm Fe}{(\alpha /Fe_{\odot })}$	Red. χ2	dof
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Bubble 1T	0.27 ± 0.02	0.37 ± 0.02	2.5E-03	⋅⋅⋅	⋅⋅⋅	0.37	0.68+0.61−0.26	0.21+0.16−0.07	3.27+0.58−0.43	1.360	151
Bubble 2T	0.29+0.04−0.02	0.17+0.14−0.07	5.5E-04	0.37+0.05−0.02	2.2E-03	0.33	0.76+0.78−0.27	0.26+0.12−0.10	2.95+0.65−0.47	1.352	149
Bubble Limb 1T	0.28 ± 0.03	0.33 ± 0.02	3.3E-03	⋅⋅⋅	⋅⋅⋅	0.33	0.48+0.38−0.17	0.18 ± 0.05	2.65+0.59−0.43	1.794	138
Bubble Limb 2T NH fixed	0.28	0.14+0.10−0.04	4.7E-04	0.35+0.02−0.01	1.7E-03	0.30	0.88+0.92−0.35	0.32+0.39−0.15	2.76+0.55−0.38	1.782	137
West 1T	0.27+0.03−0.02	0.30 ± 0.02	3.6E-03	⋅⋅⋅	⋅⋅⋅	0.30	0.36+0.24−0.11	0.10+0.07−0.04	3.42+1.34−0.77	1.088	154
West 2T	0.28+0.04−0.03	0.25+0.04−0.03	2.5E-03	0.79+0.15−0.21	2.9E-04	0.31	0.50+0.43−0.18	0.29+0.29−0.15	1.74+0.97−0.58	1.027	152
West Limb 1Ta	0.24+0.06−0.03	0.32+0.04−0.05	7.3E-04	⋅⋅⋅	⋅⋅⋅	0.32	0.49+16.89−0.29	0.21+2.28−0.09	2.31+1.21−1.89	1.387	78
West Limb 2T NH fixeda	0.28	0.21+0.03−0.02	6.6E-04	0.83+0.09−0.14	5.7E-05	0.26	0.68+6.12−0.36	1.22+2.33−0.75	0.56+0.35−0.22	1.355	77
West Blowout 1T NH fixed	0.28	0.25 ± 0.02	2.0E-04	⋅⋅⋅	${\bf... }$	0.25	2.34b	0.32+2.56−0.32	7.39−4.22	1.143	77
West Blowout 2T NH fixed	0.28	0.14+4.30−0.14	2.5E-05	0.27+0.28−0.04	3.7E-05	0.22	10.10b	1.64+374.11−1.64	6.08−3.23	1.156	75
South Bar 1Ta	0.23 ± 0.02	0.35+0.03−0.02	1.0E-03	⋅⋅⋅	⋅⋅⋅	0.35	0.75+2.71−0.35	0.21+0.700.08	3.56+1.35−0.74	1.054	128
South Bar 2Ta	0.21± 0.02	0.29+0.04−0.05	4.4E-05	0.86+0.06−0.07	1.5E-05	0.43	10.00+21.02−7.23	6.84b	1.46+1.75−0.46	1.034	126
SNR 1T	0.34+0.06−0.04	0.38+0.07−0.06	9.3E-04	⋅⋅⋅	⋅⋅⋅	0.38	0.36+0.96−0.21	0.04+0.05−0.04	10.24+24.59−5.15	1.176	67
SNR 2T	0.35+0.05−0.04	0.35+0.08−0.23	7.4E-04	0.86b	6.3E-05	0.39	0.44+0.64−0.27	0.05+0.06−0.05	8.18+51.18−6.99	1.198	65
Knot 1 1T solar	0.22	0.30 ± 0.03	8.2E-05	⋅⋅⋅	⋅⋅⋅	0.30	1.00	1.00	1.00	1.471	17
Knot 1 1T free abun	0.22	0.30 ± 0.03	7.3E-05	⋅⋅⋅	⋅⋅⋅	0.30	1.40−1.07b	0.53+9.87−0.33	2.63+4.67−1.20	1.253	15
Knot 2 1T solar	0.22	0.30+0.07−0.05	3.5E-05	⋅⋅⋅	⋅⋅⋅	0.30	1.00	1.00	1.00	1.288	8

Notes. Errors show the 90% confidence interval when one parameter is varied. ^aIt is not clear which fit is statistically better. The p-value for a 2T fit is ∼0.1. ^bErrors are unconstrained.

Download table as:  ASCIITypeset image

To determine fluxes, we ran the XSPEC command flux on each model and determined the absorption-corrected fluxes by setting N_H to zero. We adopted the distance of 50 kpc to the LMC when converting absorption-corrected fluxes to luminosities. Observed fluxes and the calculated luminosities are shown in Tables 7–10, which we discuss below.

We also calculated hardness ratios and median energies for the diffuse emission from each region, shown in Table 11; the hardness ratio Q = (H − S)/(H + S), where H and S are the luminosities in the hard (E = 0.6–1.6 keV) energy band and soft (E = 0.3–0.6 keV) energy band. Absorption by intervening H i results in an apparently harder spectrum, and Figures 2(c) and 3(c) show that there is significant variation in the H i column density across these objects. The column density was a free parameter in our fits, and the resulting luminosities and hardness ratios account for N_H variations between regions. Q obtained using the count rate (i.e., not absorption-corrected) is shown for comparison in Column 4 of Table 11. The influence of the column density can be seen by comparing the luminosity-determined and count rate-determined hardness ratios for the SNR 0523−679 (Chu et al. 1993) in DEM L152. While the SNR appears to have the hardest spectrum from the count rates, its higher absorption column density masks a much softer spectrum.

Hong et al. (2004) demonstrate that the median energy is a more reliable indicator of X-ray spectral variations than the hardness ratio is. As a result, we include the median energy of each region in Table 11. The median energies generally agree with the hardness ratios. An exception is the DEM L152 SNR, which has a higher median energy due to the H i absorption discussed above.

Modeled fits to DEM L50 are shown in Tables 4 and 5. The fits in Table 4 model the emission as one- and two-temperature thermal plasmas. The best-fit N_H in Column 3 is consistent with the observations of the H i in DEM L50 by Oey et al. (2002). The first temperature component, kT₁, appears in Column 4, with its normalization in Column 5, and the second component, kT₂, appears in Column 6, with its normalization in Column 7. Column 8 shows the "Effective kT," defined as the average kT of the region weighted by the normalization. The "Effective kT" thus shows the dominant temperature component in the region. Columns 9 and 10 give the abundances relative to solar, with the α/Fe ratio in Column 11. Columns 12 and 13 show the reduced χ² and the number of degrees of freedom of the fit. The statistically best fits, determined by running ftest in XSPEC, are shown in bold. Figure 7 shows a sample fit to the Center region of DEM L50.

Zoom In Zoom Out Reset image size

From the best fits for the Limb and Center regions in Table 4, the Center's temperature distribution appears more homogeneous than the Limb's, and the two-temperature best fit to the Limb shows a higher temperature component that may correspond to the Bright Limb subregion. This correlation would be consistent with shock heating of the southern part of the Limb by an expanding SNR, while the rest of the Limb remains at a cooler temperature.

4.3.1. Low α/Fe or Non-equilibrium Ionization?

In each region of DEM L50, spectral fits consistently gave low α/Fe ratios (see Table 4), with typical values of about 0.3 compared to an average LMC value of 0.6 (Smith 1999). These low ratios are especially surprising, given that we might have expected a high α/Fe ratio due to an enhancement in α elements from core-collapse SNe. The bubble has likely experienced ∼2 SNe (Oey 1996b), so an α enhancement would be natural. Raising the α/Fe ratio by forcing the abundances to solar values resulted in the models labeled "solar" in Table 4. In every case, ftest showed that the free abundance fits were statistically better than the solar values. The ftest command calculates a p-value, which here denotes the probability that solar-abundance material could randomly produce a spectrum fit by the more complicated, free abundance fit. The resulting p-values ranged from 0.049 to as low as 4 × 10⁻⁶, showing that the solar fits have a very low probability of being correct. Although the solar abundances are statistically excluded, there may be systematic problems in forward-fitting low-resolution X-ray spectra, which could give us misleading estimates of the abundances and their uncertainties. Using the Mekal or Raymond–Smith models instead of APEC did not affect the α/Fe ratios.

Another possibility is that collisional ionization equilibrium is not a valid assumption for DEM L50. To test this scenario, we performed the non-equilibrium ionization fits shown in Table 5. Temperature is shown in Column 4, the normalization is in Column 5, abundances relative to solar are in Columns 8 and 9, and the reduced χ² and number of degrees of freedom are in Columns 11 and 12, respectively. The α/Fe ratios found with these new fits are all consistent with solar values. A sample non-equilibrium ionization fit is shown in Figure 8 for the DEM L50 Center region.

Zoom In Zoom Out Reset image size

To see whether or not non-ionization equilibrium is a plausible scenario, we consider the ionization timescale. The ionization timescale in seconds is found by dividing τ, the density-weighted ionization timescale in s cm⁻³ (Column 6), by the electron density, n_e. To find the electron density, we used

$$\begin{equation} \eta = \frac{10^{-14}}{4\pi D_{A}^2}\int n_{e} n_{\rm H} dV, \end{equation} \tag{ 1 }$$

where η is the normalization of the fit, in units of cm⁻⁵, and D_A is the distance to the superbubble, assumed to be 50 kpc. We also assumed a spherical geometry, uniform density distribution, and ∼10% He abundance (n_e ∼ 1.2n_H; see, e.g., Peimbert et al. 2007). The resulting ionization timescale estimates are shown in Column 7 of Table 5 and are all on the order of 10⁵ years, about 10% of the superbubble's age. While our observation of the bubble in this stage would be a chance occurrence to some extent, this timescale is not unrealistically brief and a recent SN is also consistent with many of the observed properties of the bubble as we will discuss later. Furthermore, the hypothesized brightening from SNR impacts should enhance the detectability of bubbles with recent SNe, increasing the likelihood of observing a bubble that is not in equilibrium.

If the low α/Fe abundances are indeed real, they may reflect an Fe enrichment in the ISM near DEM L50. DEM L50 is located in a large, 1.4 kpc diameter void in the ISM (Oey et al. 2002); a single Type Ia SN could not noticeably enrich this entire region, nor could it enrich the estimated 1.7 × 10⁵ M_☉ of material swept up by DEM L50 (Oey et al. 2002). On the other hand, if the density is low enough, the interior of DEM L50 could be enriched by a single Type Ia event, and SNR N186 D near the bubble would be a likely candidate. The α/Fe ratio of the SNR, however, does not seem noticeably different from the rest of the bubble (see Section 4.5) and may in fact be less Fe-enriched than the Center region. There is no clear spectral contrast between the SNR and regions dominated by the shocked ISM like that seen in the Type Ia SNR DEM L71 (Hughes et al. 2003), nor are there any noticeable Fe, S, or Si features such as those seen in Type Ia SNR N103B (Hughes et al. 1995). Furthermore, SNR N186D's proximity to young stars and star-forming regions make a core-collapse SN origin more likely (e.g., Chu & Kennicutt 1988).

Dust destruction from an SN shock wave cannot explain the low α/Fe ratio. Although dust destruction would increase the Fe abundance, the destruction of silicates would concomitantly raise the O abundance (M. Compiègne 2010, private communication; see Whittet 2003 for O and Fe abundances in dust). We note that DEM L50 does have an uncommon infrared morphology (Slater et al. 2011); its 8 μm emission appears to fill the bubble's Hα shell, whereas many other LMC superbubbles only show polycyclic aromatic hydrocarbon (PAH) emission external to the bubble. This morphology is likely explained by PAH emission on the interface between the ionized and the foreground neutral layer, which is seen in projection toward the center of the bubble (see Figure 2 and Section 3.2). An unusual dust environment does not appear to be the cause of the low observed α/Fe ratio.

Similarly odd Fe abundances have been observed in other objects. DEM L316 consists of a pair of interacting SNRs; one SNR has a high Fe abundance while the other does not (Williams & Chu 2005; Nishiuchi et al. 2001). As mentioned previously, the LMC α/Fe ratio is low in general, ∼0.6 (Smith 1999); DEM L50 may simply lie in a region of the ISM with a lower than average α/Fe ratio.

Of course, the α/Fe ratio we observe could also result from inadequate X-ray models or spectra, rather than genuine properties of the emitting material. In X-ray spectral models, parameters such as temperature and column density are degenerate with the metallicity, making the model difficult to constrain (Dahlem et al. 2000). A variety of X-ray models may fit the same spectrum equally well. For instance, other authors have found that the need for unusual abundances in a spectral fit may disappear when a more complicated, multi-component X-ray spectral model is used instead (e.g., Weaver et al. 2000; Strickland et al. 2002). The ∼0.12 keV resolution of Chandra ACIS spectra and model degeneracies make it unclear which precise combination of models should be used, however. Determining the origin of the unusual abundances in DEM L50 requires further investigation of both the LMC ISM and the X-ray models used, and the explanation of the observed abundances may have to wait for higher resolution spectra. Our current data suggest, however, that the non-equilibrium ionization scenario is the most plausible explanation for DEM L50's low observed α/Fe ratio.

4.3.2. Luminosities

Like DEM L50, DEM L152 is an X-ray-bright superbubble, although as seen in Figures 2 and 3, its H i environment and X-ray and optical morphology differ substantially from that of DEM L50. Consequently, DEM L152's derived X-ray properties also generally differ from DEM L50; the characteristics the two objects have in common may help resolve the question of why some superbubbles become X-ray-bright. We modeled the emission from DEM L152 with one- and two-temperature thermal plasmas. Table 6 reports the resulting fits; the column headings are the same as in Table 4, except that the background fit, which is the same for all fits, is not listed. In the case of the West Limb and South Bar, it is not clear which fit is statistically better, so neither fit is shown in bold. Figure 9 shows the best fit to the Bubble region as an example.

Zoom In Zoom Out Reset image size

DEM L152 is a few 10⁶ K hotter than DEM L50 on average, a result of the higher number of O stars and SNe in DEM L152 (Oey & Massey 1995; Oey 1996a). The Bubble Limb appears slightly cooler than the Bubble region as a whole, which is reasonable since it makes up the outer edge of the Bubble and likely consists of denser, cooler material. The West region shows a 0.25 keV component as well as an extremely hot 0.79 keV component. This hot component appears in the West Limb region two-temperature fit, but not the West Blowout, and may also account for a substantial portion (∼25%) of the emission from the South Bar. The localization of this hot component to the adjacent West Limb and South Bar regions suggests recent heating in this area, perhaps from an SN explosion. An expanding SNR would also explain the limb brightening seen in the West region. Interestingly, Chu & Mac Low (1990) also suggest the presence of an off-center SNR in the southwest due to the X-ray brightness in the South Bar, and Magnier et al. (1996) find high-velocity gas in the region.

As expected, the two blowout regions are slightly cooler than the Bubble region. Most of the South Bar's emission is fit by a 0.29 keV component, while the hot component mentioned above accounts for the remainder. Likewise, a 0.25 keV plasma characterizes the West Blowout. The blowouts are cooler than the 0.37 keV Bubble region and cooler than (or at least comparable to) the 0.33 keV Bubble Limb region, presumably having cooled due to adiabatic expansion.

Abundances for each region show enrichment from core-collapse SNe (see Table 6), which raises the α/Fe ratio. We discuss the implications of this enrichment for the expected luminosity in Section 6.1.

Two clumps of X-ray-emitting material, a few parsecs in diameter, appear in the South Bar region (see Figure 6). We extracted and fit spectra for these two knots (see Table 6), fixing the column density to the best-fit value for the South Bar region. Due to low signal to noise, we only used single temperature fits and fixed abundances to solar values for the second knot. The resulting luminosities should be viewed as estimates only. Nevertheless, the luminosities are one to two orders of magnitude lower than the luminosity of the other regions and clearly do not have much effect on the total emission. We plot the median energy versus total count rate for the knots and regions of DEM L152 in Figure 10. The knots do not exhibit any detectable spectral difference relative to the rest of the superbubble.

Zoom In Zoom Out Reset image size

The luminosities for each region and subregion of DEM L152 are shown in Table 10; the N_H error bars correspond to an uncertainty of ∼15% in the calculated luminosities. The Bubble region provides ∼37% of the total X-ray luminosity (33% from the Bubble Limb, 4% from diffuse emission), the West region contributes ∼41% (14% from the West Limb, 13% from the West Blowout, and 14% from diffuse emission), and the South Bar provides ∼21% (19% from diffuse emission, 2% from the knots). The intrinsic surface brightnesses are unclear, as they depend on the model and column density assumed. For example, the best fits for the Bubble region imply some slight limb brightening, while the two-temperature fits indicate the opposite. If the limb brightening observed in Figure 4 is not intrinsic, it must instead result from higher absorption toward the Bubble center. This hypothetical absorption would conflict with the H i contours shown in Figure 3, however, which show a lower column density toward the center of the Bubble. Thus, we expect that the observed limb brightening in the Bubble region is real.

A comparison of the hardness ratios and median energies in Table 11 shows that the South Bar and West Blowout regions are softer than the rest of the superbubble, consistent with a blowout origin and with the previous X-ray observations of Chu et al. (1993) and Magnier et al. (1996). This variation in hardness can also be seen in Figure 4(e). The Bubble and Bubble Limb regions are noticeably harder than the other regions of DEM L152. In general, we find softer hardness ratios for DEM L152 than previous authors (e.g., Wang & Helfand 1991), a result of fitting the column density instead of assuming a fixed conversion between count rate and luminosity.

As with DEM L50, limb regions account for roughly half of the observed X-ray emission. The South Bar and West Limb regions alone provide a third of the total emission and may be associated with a recent off-center SNR as discussed above. In contrast, the two knots make up only 2% of the total X-ray emission, suggesting that emission from clumps cannot explain DEM L152's X-ray luminosity.

The SNR N186 D appears at the northern edge of DEM L50. The fact that N186 D is detected in [O iii] while DEM L50 is not suggests they may be two physically distinct objects (Lasker 1977); similar H i kinematics in both objects, however, may indicate that the SNR is associated with the larger bubble (Oey et al. 2002), and thus the exact relationship between the SNR and the superbubble remains unclear.

SNR N186 D's X-ray spectra can be fit by a non-equilibrium model or a two-temperature equilibrium model. With free abundances, the equilibrium fits show a 0.19 keV component and a weaker 0.72 keV component. These temperatures are typical of SNRs in the LMC (see Williams 1999) and in our own Galaxy (e.g., Temim et al. 2009; Hui & Becker 2009). If abundances are forced to solar values, however, the two temperature components are 0.08 keV and 0.29 keV. The dominant 0.08 keV component is a lower temperature than any of the LMC remnants studied by Williams (1999) and even the 0.29 keV component is on the lower end of her sample. The free abundance fit is statistically better than this solar fit, with a p-value of 0.05, and shows the same low α/Fe ratio seen in the rest of DEM L50. Although we cannot rule out the possibility that this SNR is a Type Ia remnant, the consistency of the α/Fe ratio throughout DEM L50 suggests the abundances are associated with the ISM environment of DEM L50 or, alternatively, that the SNR is not in ionization equilibrium. As discussed in Section 4.3.1, the non-equilibrium scenario is the more likely explanation. A non-equilibrium ionization fit gives an ionizing timescale of 11,000 years, a temperature of 0.58 keV, and an α/Fe ratio of 0.64, close to the typical LMC value of ∼0.6 (Smith 1999).

The non-equilibrium ionization model gives a total X-ray luminosity in the E = 0.3–8.0 keV band of 3.73 × 10³⁵ erg s⁻¹. The luminosity varies by an order of magnitude among the potential models, however, from 2 × 10³⁵ to 1 × 10³⁶ erg s⁻¹, a result of the higher N_H in the models with solar abundances. These possible luminosities are all within the range of ∼10³⁴–10³⁷ erg s⁻¹ found by Williams et al. (1999) for SNRs in the LMC and by Long et al. (2010) for M33 SNRs.

The median energy for SNR N186 D, shown in Table 11, is softer than its hardness ratio would suggest. The H i appears patchy in this region of DEM L50, and the observations by Oey et al. (2002) show a substantial gradient in the H i column density near the SNR. These N_H variations may be affecting our modeled temperatures in this region.

DEM L152's SNR, 0523−679 (Chu et al. 1993), consists of a 0.38 keV plasma, close to the median temperature value of Williams' sample of LMC SNRs (Williams 1999). The SNR shows a high α/Fe ratio characteristic of core-collapse SNe. The extremely high and unconstrained value of the SNR's α/Fe ratio is a result of the lower signal to noise in this small region of the bubble. The total X-ray luminosity is 5.73 × 10³⁵ erg s⁻¹, typical of other SNRs (e.g., Williams et al. 1999; Long et al. 2010; Seward et al. 2010). The SNR does appear to differ spectrally from the rest of the bubble, as seen in Figure 10.

We modeled the central temperature and X-ray luminosity in the [0.3–2.0 keV] band for DEM L50 and DEM L152 using the similarity solutions of Weaver et al. (1977) as implemented by Oey & Massey (1995). Energy and mass input were not assumed to be constant; time-dependent energy input from stellar winds and SNe was calculated using the mass loss rates from Schaerer et al. (1993) and the stellar population shown in Table 12 (see Oey & Massey 1995 for details).

For DEM L50, we assumed an ambient density of 1.4 cm⁻³ as explained in Oey (1996b), a metallicity of 0.4 $\thinspace Z_{\odot }$, and an age of 5 Myr, at which point two SNe are expected to have occurred if a Salpeter (1955) initial mass function (IMF) is assumed (Oey 1996b). We modeled DEM L152 using a density of 2.5 cm⁻³ (Oey & Massey 1995), a metallicity of 0.4 $\thinspace Z_{\odot }$, and an age of 6 Myr, incorporating the effects of four SNe as suggested in Oey & Massey's (1995) analysis of DEM L152's mass function. The resulting X-ray luminosities and central temperatures are shown in Table 13 and compared with the observed values. The observed X-ray luminosities are an order of magnitude higher than the standard model predicts. Varying the input ambient density or metallicity did increase the X-ray luminosity, but even increasing these parameters by an order of magnitude still resulted in an X-ray luminosity lower than the observed value.

The radii calculated with the Weaver model likewise do not match the observed values; both bubbles have predicted radii of more than 100 pc (130.6 pc for DEM L50, 146.6 pc for DEM L152), whereas the observed radii are around 50 pc. This "growth-rate discrepancy" has been observed in bubbles with a range of sizes, from bubbles around individual Wolf-Rayet stars to the superbubbles we consider here (e.g., Saken et al. 1992; Brown et al. 1995; Oey & Massey 1995; Cappa et al. 2003; see review by Oey 2009). One possible explanation for the smaller observed radii is that excess radiative cooling has caused the bubble to lose more energy than predicted. To see if the high X-ray luminosities of our objects provide enough cooling to slow the bubble's growth, we use the formula for the bubble radius given by Weaver et al. (1977):

$$\begin{equation} R=27n_0^{-1/5}L_{36}^{1/5}t_6^{3/5} \rm pc, \end{equation} \tag{ 2 }$$

where n₀ is the ambient density in cm⁻³, L₃₆ is the mechanical luminosity in 10³⁶ erg s⁻¹, and t₆ is time in 10⁶ years. Assuming that SNe dominate the mechanical energy input, we can estimate the mechanical luminosity as the energy provided by SNe divided by the age of the superbubble. Two SNe and an age of 5 Myr (DEM L50) or four SNe and an age of 6 Myr (DEM L152) yield a mechanical luminosity of order 10³⁷ erg s⁻¹. The X-ray radiative losses for both bubbles, on the other hand, are on the order of 10³⁶ erg s⁻¹. Although our superbubbles are X-ray-bright, the observed X-ray luminosity is therefore insufficient to account for the small radius observed in both bubbles.

In contrast to the Weaver model, the one-dimensional hydrodynamic model of DEM L50 from Oey & García-Segura (2004) does correctly fit the radius and velocity of the bubble. The mechanical input is the same as before, but this model assumes a higher ambient pressure of P/k = 1 × 10⁵ cm⁻³ K, neglects thermal conduction at the shell walls, and includes the effect of an SNR impact on the shell. While the higher ambient pressure results in the correct radius, the X-ray luminosity predicted by this model is far too low, an order of magnitude lower than the Weaver model, which was itself lower than the observed luminosity. This low luminosity suggests that thermal conduction does play a role in DEM L50 and is not strongly suppressed by magnetic fields. If thermal conduction is occurring, the bubble temperature should be lowest at the outer edge, where dense, cold evaporated material is present, and should increase toward the bubble center. The equilibrium fits to the DEM L50 Limb and Center regions do show this temperature contrast; the effective temperature of the Limb is only 0.09 keV, while the value increases to 0.15 keV for the Center region. The non-equilibrium fits, on the other hand, suggest that the Limb is slightly hotter than the Center. The possibility of a cooler Limb is within the error bars, however, and the contribution of the heated Bright Limb region may raise the average temperature of the Limb.

The X-ray luminosity at different times is shown in Table 14 and compared with the Weaver model luminosity for DEM L50 and the observed luminosity. An SN goes off at t = 5.26 Myr, and the SNR increases the bubble's luminosity by several orders of magnitude. The emissivity profiles of the bubble before and after the SNR hits the shell walls are shown in Figure 11. The SNR impact enhances the emissivity of the bubble's limb and also increases the shell velocity from 1 kms⁻¹ to 23 kms⁻¹ in agreement with the observed value. Although the luminosities of this model are unrealistic, the response of the luminosity and shell velocity to the SNR impact suggests that SNR impacts may be important in our superbubbles.

Zoom In Zoom Out Reset image size

Neither the Weaver model nor the one-dimensional hydrodynamic model is able to match the observed luminosities of the superbubbles. The hydrodynamic model suggests, however, that mass loading via thermal evaporation from the shell walls must play a role in these bubbles. Mass loading from the shell walls and heating from SNR impacts are consistent with our observations and could account for most of the superbubbles' X-ray emission.

Silich (2001) suggests that since emission lines of α elements dominate the soft X-ray emission, metallicity enhancement from SNe could explain the anomalously high X-ray luminosities observed in many superbubbles. As discussed earlier (Sections 4.3 and 4.4), DEM L152 shows an α enhancement, but DEM L50 may not, indicating that a different explanation would be required for that bubble.

Assuming the initial stellar population proposed by Oey & Massey (1995), approximately four SNe have gone off in DEM L152, from one 85 M_☉ progenitor and three 60 M_☉ progenitors (see Table 12). These SNe should input a total 65 M_☉ of O (Maeder 1992) and 0.2–0.8 M_☉ of Fe (see discussion in Silich et al. 2001). Assuming a number density of 0.01–0.2cm⁻³, spherical geometry, a radius of 50 pc, and 10% He abundance, the total mass enclosed by the bubble should be between 100 and 3000 M_☉. As the inferred number density from the observed spectrum (see Equation (1)) is 0.2cm⁻³, we assume an enclosed mass of 3000 M_☉ in what follows. We use Equations (9) and (10) from Silich et al. (2001) to calculate the resulting metallicity:

$$\begin{equation} Z_{\rm O} = \frac{M_{\rm ej,O}/Z_{\rm sol,O}+Z_{\rm ISM}M_{\rm ev}}{M_{\rm ev}+M_{\rm ej}}, \end{equation} \tag{ 3 }$$

$$\begin{equation} Z_{\rm Fe} = \frac{M_{\rm ej,Fe}/Z_{\rm sol,Fe}+Z_{\rm ISM}M_{\rm ev}}{M_{\rm ev}+M_{\rm ej}}. \end{equation} \tag{ 4 }$$

M_ej is the mass ejected in SNe, M_ev is the mass evaporated from the shell walls, and Z_sol,O and Z_sol,Fe are the solar abundances by mass from Grevesse et al. (1996), Z_sol,O = 0.0083 and Z_sol,Fe = 0.00126. Weaver et al. (1977) predict that the evaporated mass will be the dominant source of mass in the bubble interior. We assume that, apart from the mass ejected by the four SNe, the remainder of the mass was contributed by evaporation. Assuming stellar remnant masses of ∼2 M_☉ (Maeder 1992) and the initial masses from Oey & Massey (1995), the total mass returned to the ISM by the four most massive stars is 257 M_☉, resulting in a M_ev of ∼2700 M_☉. Then, from Equations (3) and (4), the predicted metallicities are $Z_{\rm O}=3\hbox{$\thinspace Z_{\odot }$}$ and $Z_{\rm Fe}=0.4\hbox{--}0.6\hbox{$\thinspace Z_{\odot }$}$. If the bubble contains less evaporated mass, these metallicities will be higher.

As noted by Silich et al. (2001), the X-ray emissivity is proportional to the metallicity. Assuming the α elements dominate the soft X-ray emission, the factor of 10 increase from Z_LMC of 0.4$\thinspace Z_{\odot }$ to Z_O of 3 $\thinspace Z_{\odot }$ would be sufficient to explain the factor of 10 increase in DEM L152's luminosity (see Table 13). This predicted metallicity does not match the observed α abundance in DEM L152, however (see Table 6), which is subsolar in all cases with reasonable error bars. The α/Fe ratio is a more reliable diagnostic of the metallicity, and it is possible that the α abundance from the spectral fits is incorrect. Given the additional uncertainties in the bubble's density, geometry, and resulting mass and in the number of SNe that have occurred, this discrepancy between the predicted and observed metallicities is not unexpected. The predicted α/Fe ratio from the four SNe is 5–8, while Table 6 shows typical α/Fe ratios of around 3. When errors are taken into account, the observed α/Fe ratio could be consistent with the predicted value, and therefore it is possible that the higher α abundance is also present.

If we use the observed α abundances, which have an average value of ∼0.6$\thinspace Z_{\odot }$ and are generally consistent with 1$\thinspace Z_{\odot }$, the enhancement over the Weaver model only accounts for 14%–20% of the observed emission. At most, the observed α abundance can account for the diffuse emission in the Bubble and West regions and the West Blowout, but not for the emission from the Bubble Limb, West Limb, or South Bar. If, on the other hand, we accept that this observed α abundance is inaccurate and use the predicted Z_O, the enhanced metallicity explains the X-ray luminosity but not the limb-brightened features or enhanced expansion velocities.

A similar analysis for DEM L50 yields a predicted Z_O of 5$\thinspace Z_{\odot }$ and predicted α/Fe of 7–13, where we have assumed M_ej,O = 36 M_☉, M_ej,Fe = 0.1–0.4 M_☉, a bubble radius of 50 pc, and n_H = 0.06cm⁻³, derived from the non-equilibrium fit normalization. The predicted metallicities are much higher than the approximately solar values obtained from the non-equilibrium fit, and as with DEM L152, the observed metallicities cannot account for the majority of the X-ray enhancement.

Another possible explanation for the high X-ray luminosities of DEM L50 and DEM L152 is mass loading from the evaporation or ablation of overrun clouds. While the distribution of mass from evaporation at the shell walls should be uniform over relatively large scales, evaporating clouds should appear as small clumps of emission in our data. DEM L50 is located in a void and does not exhibit any noticeable clumpiness in its emission. The two knots we observed in DEM L152 and its more crowded H i environment, on the other hand, show that mass loading by clumps could be occurring in that bubble. Silich et al. (1996) investigate clouds with n_e ∼ 10cm⁻³ and radii of 3–5 pc, while Orlando et al. (2005) find their modeled shocked clouds have number densities of 4cm⁻³, radii of 1 pc, and are surrounded by a corona with density 0.4 cm⁻³. To encompass this range of possible cloud parameters, we calculated the emission from clouds with each of these three densities (n_e = 0.4, 4, and 10 cm⁻³), a temperature of 10⁶ K, a metallicity of 0.4 $\thinspace Z_{\odot }$, and radii of 0.5 pc, 1 pc, and 5 pc, all of which should be resolved at Chandra's ∼1'' resolution. The X-ray emissivity in the 0.2–4.5 keV band is approximately Λ_x = 9 × 10⁻²⁴ × Z erg s⁻¹cm⁻³ (Silich et al. 1996); this emissivity will result in more emission than we expect to see in the 0.3–2.0 keV band, but it should allow us to roughly estimate the type of clouds that could be detectable. Assuming a median photon energy of 0.55 keV (Orlando et al. 2010), we can calculate the predicted count rate per arcsec² (Table 15). A comparison with the X-ray contours in Figures 2 and 3 shows that the knots with radii of 0.5 pc are potentially detectable if they have a density of ∼10cm⁻³, while no knots with a density of ∼0.4cm⁻³ are detectable. Interestingly, the detected "knots" in DEM L152 have a count rate of approximately 6.4 × 10⁻⁶ s⁻¹ arcsec⁻², which is within the range of values in Table 15 and a density of ∼1cm⁻³, found using Equation (1), assuming a radius of 2.5 pc. These values suggest that these knots may indeed be genuine clouds swept up by the bubble.

Another possibility is that some clumps appeared as point sources in the images (Figure 1). We would expect predominantly soft emission from overrun clumps, while other point sources, such as background AGNs or X-ray binaries, should appear in the hard band as well. Only two point sources, one in DEM L50 and one in DEM L152, appear in the soft band alone. The point source in DEM L50 is quite faint, containing only ∼15 counts, and its identity is unclear. The soft point source in DEM L152 is located outside the main bubble. This source is likely associated with the highly ionized N44 C region rather than an independent cloud swept up by DEM L152. In any case, the lack of multiple soft point sources suggests that dense, overrun globules are not present to any significant degree in either bubble.

Although DEM L50 and DEM L152 may have overrun additional clouds which are below our detection limit or already destroyed, their morphology and spectral properties suggest that mass loading by clouds is not dominating the bubbles' evolution. Modeled bubbles with high levels of mass loading tend to show a smoother distribution of density and temperature across the bubble, as the clouds add cooler gas to the system (see, e.g., Arthur & Henney 1996; Pittard 2007). The increased central density and reduced central temperature may result in a flat surface brightness profile (Arthur & Henney 1996) or even a centrally peaked profile (Silich et al. 1996). A limb-brightened morphology, on the other hand, appears correlated with SN activity. No pre-SN superbubbles yet studied show limb brightening, while this morphology is common in post-SN superbubbles (Chu et al. 2003). DEM L50 and DEM L152's limb-brightened morphologies likely reflect their SN history rather than a history of high mass loading.

Mass loading should reduce the central temperature of a bubble (e.g., Silich et al. 1996), whereas DEM L152's central temperature is quite high (∼0.37 keV), higher than predicted by the Weaver model. DEM L50's lower central temperature of 0.15 keV in the equilibrium fit could be consistent with mass loading, although uncertainties in the number of SNe or energy input in DEM L50 could also account for the slightly lower temperature. The temperature derived from the non-equilibrium fit to DEM L50 is completely consistent with the Weaver model predictions, however, requiring no additional mass loading. Mass loading by overrun clouds or evaporation from the shell walls is likely occurring to some extent in both bubbles, but it does not fully explain the bubbles' X-ray emission or large-scale properties.

Dense cloudlets have been detected in other superbubbles, but their presence does not necessarily imply a high X-ray luminosity. For instance, although clouds are detected in the superbubble DEM L192 and the bubble appears to be expanding into an inhomogeneous ISM, its X-ray luminosity in the 0.3–3.0 keV band is only moderate, ∼4 × 10³⁵ erg s⁻¹ (Cooper et al. 2004). Cooper et al. (2004) note a correlation between the Hαand X-ray surface brightnesses in this bubble, however, and suggest that mass loading is occurring through evaporation at the shell walls. This analysis is consistent with our findings in this section and in Section 5.2; thermal conduction and the subsequent evaporation of shell material is important in DEM L50 and DEM L152, while the contribution of swept-up clouds to the X-ray emission is negligible.

Thus, we see evidence that both metallicity enhancement and mass loading by clouds has taken place in these superbubbles. The metallicity enrichment is too low to account for the observed luminosities, however. Likewise, few candidate clouds are observed in the superbubbles and the X-ray emission is inconsistent with a high mass-loading scenario. Metallicity and mass loading by clouds do not completely explain the observed X-ray luminosities.