Kinematics: A Clean Diagnostic for Separating Supernova Remnants from H ii Regions in Nearby Galaxies

The study of extragalactic supernova remnants (SNRs) began modestly with the discovery of three radio candidates in the Large Magellanic Cloud (LMC) by Mathewson & Healey (1964), later confirmed through their radio and optical properties by Westerlund & Mathewson (1966). It was over a decade until SNR studies expanded to more distant galaxies through the optical identification of three remnants in M33 by D'Odorico et al. (1978). Great advances have been made since, so that today there are a total of over 1400 known SNRs (or at least strong candidates) in about a dozen galaxies, up to distances of ∼10 Mpc—many times the number known in our own Galaxy (see Long 2017 for a recent discussion of the state of play in the field).

Unlike in our Galaxy, where most SNRs were first discovered through their radio properties and only about a third have even been detected optically (Green 2017), the vast majority of extragalactic remnants have been discovered optically. Some SNRs have been detected at X-ray and radio wavelengths in galaxies outside the Local Group, but due to limited sensitivity and angular resolution few were originally discovered in these wavelength bands. The classic method to identify SNRs optically is through the [S ii]:Hα ratio, where shock heated nebulae (primarily SNRs) typically have [S ii]:Hα ≳ 0.4, compared with photoionized nebulae (H ii regions and planetary nebulae) where [S ii]:Hα is typically ≲0.2. This notable difference results from the fact that the gas behind shocks like those characterizing SNRs has a long cooling tail containing a range of low-ionization species, including S⁺, N⁺, and O⁰, while the photoionized gas surrounding hot stars is generally more highly ionized.

As larger telescopes and more sensitive instrumentation have enabled exploration of deeper extragalactic samples, the tidy [S ii]:Hα ratio criterion for distinguishing SNRs from H ii regions has become blurred. At low surface brightness especially, photoionized regions such as the warm ionized medium in our Galaxy and the so-called diffuse ionized gas (DIG) in nearby galaxies can display [S ii]:Hα ratios that approach or exceed 0.4, as density decreases and the distance from the ionizing sources increases (e.g., Haffner et al. 1999). Hence, a patch of DIG in a nearby galaxy may be misidentified as a potential low-surface-brightness SNR. Also, extragalactic SNRs may be embedded in complex emission regions or located so close to regions of H ii emission that they cannot be resolved with ground-based facilities. Intrinsic observational uncertainty in the derived line ratios for faint emission-line objects can also cause ambiguity for objects with observed ratios near the 0.4 threshold. Therefore, it would be valuable to have an additional optical diagnostic for positive identification of SNRs; hence the motivation for this study.

A potential diagnostic for SNRs is to take advantage of the fact that the material behind an SNR shock will have bulk motions comparable to the local shock speed, several hundred km s⁻¹, whereas material in H ii regions will for the most part have velocities of order 10 km s⁻¹. Unfortunately, although spectroscopic follow-up of a large number of the nebulae identified as SNRs in extragalactic surveys has taken place (e.g., Blair & Long 1997; Matonick & Fesen 1997; Gordon et al. 1998; Lee & Lee 2014a, 2014b; Lee et al. 2015; Winkler et al. 2017; Long et al. 2018, 2019), most of these spectroscopic surveys have not had sufficient spectral resolution to measure the expected line broadening in SNRs and use it as a meaningful diagnostic.

Here we report a survey of selected SNRs in the LMC, along with a few SNR candidates in M83, to test whether measuring line broadening is a practical way to separate SNRs from H ii regions, and to assess what additional information line broadening may provide in attempting to understand the nature of SNRs in our sample. The LMC, with 59 SNRs and 15 additional candidates, is the obvious place to carry out such a study because it has the most fully characterized sample of SNRs of any galaxy (see, e.g., Maggi et al. 2016; Bozzetto et al. 2017). There is, however, one obvious observational difference between the SNRs in the LMC and those in other more distant galaxies: at 50 kpc (Pietrzyński et al. 2013), the SNRs in the LMC are fully resolved, and simple long-slit spectroscopy samples only a small portion of each SNR. This is in contrast to the situation in a galaxy like M83 (d = 4.6 Mpc, Saha et al. 2006), where a spectroscopic aperture (slit or fiber) typically samples most or all of an SNR. To address this situation, we obtained spectra of several SNRs and a few H ii regions in the LMC by scanning the slit over the entire object, thus obtaining an integrated spectrum analogous to what a slit might encompass for an object in a galaxy like M83.

In the following sections, we detail our new kinematic observations of SNRs, present our results, and then provide a discussion of our findings. A brief summary of our conclusions appears at the end.

We used the Goodman High-Throughput Spectrograph (Clemens et al. 2004) at the SOAR 4.1 m telescope on Cerro Pachón to measure the velocity broadening of emission lines as a potential diagnostic of SNRs. Our observations took place on five nights: 2018 December 14–16 and 2019 January 15–16 (UT), as detailed in Table 1. We used the Goodman Red Camera, equipped with a 4096 × 4112 pixel back-illuminated, deep-depletion, astro-multi-2 coated e2v 231-84 CCD, and chose the red-optimized 1200 lines mm⁻¹ Volume Phased Holographic grism for our observations. At a central wavelength of 6840 Å, the total wavelength coverage was 6230–7460 Å, which included lines of interest such as [O i] λλ6300, 6363, [N ii] λλ6548, 6583, Hα, and [S ii] λλ6716, 6731. We used a 06 slit with 1 × 1 pixel binning as the optimum compromise between resolution and throughput. The instrumental profile, as measured from the HgAr + Ne calibration lamp lines was 3.95 ± 0.17 pixels (≈1.2 Å; ≈53 $\mathrm{km}\,{{\rm{s}}}^{-1}$). The spatial scale along the slit was 015 pixel⁻¹.

We obtained spectra toward a sample of LMC and M83 SNRs and H ii regions. For the primary LMC targets, we used a drift-scan technique, allowing the narrow slit to scan gradually across the entire source. This enabled us to sample the full range of filamentary velocities, somewhat analogous to the situation for long-slit spectra toward M83 and other more distant galaxies, where much or all of a marginally resolved source falls within the slit. We selected the LMC targets based on the criteria of size and surface brightness. We chose targets that were small enough (in either the N–S and/or E–W direction) to be covered in a single scan with the 39 long Goodman slit, with at least some margin at one or both ends to facilitate sky subtraction. For simplicity, the slit was aligned either N–S or E–W so we could scan along a single coordinate perpendicular to the slit.

We used flux-calibrated images from the Magellanic Clouds Emission Line Survey (MCELS; Smith & MCELS Team 1999; Paredes et al. 2015; Winkler et al. 2015) to estimate the total Hα flux and surface brightness of SNRs in the LMC. Ideally, for this experiment we would have chosen examples covering the full range from brightest to faintest. However, because drift-scan spectra are "inefficient," in that much of what the slit sees as it scans over an object is blank sky rather than filaments of emission, we discovered that it was not practical to observe the fainter objects with this technique.

With the Goodman spectrograph, it is quite easy to switch between imaging and spectrographic modes. In order to facilitate both the setups and subsequent analysis, we obtained short direct images in Hα, SDSS r', and in some cases [S ii] as well, before the scanned spectra for each object.

In addition to the scanned spectra, we also obtained long-slit spectra for five of the LMC SNRs in our sample, in order to assess the effectiveness of the scans compared with more traditional long-slit spectra. The positions and position angles for these long-slit observations were chosen to minimize stellar contamination while maximizing coverage of the SNR emission regions. Finally, for comparison with the situation typical for more distant galaxies, we also used the same setup to obtain long-slit spectra at two positions in the galaxy M83, chosen to include multiple SNRs from the Blair et al. (2012) catalog. (One or more H ii regions also lay along each slit position.) Table 1 gives a journal of all our observations of LMC and M83 targets. In total, we obtained drift-scan spectra toward 15 LMC SNRs and eight LMC H ii regions (three explicitly targeted and five others covered in SNR scans) plus long-slit spectra of five LMC SNRs, five M83 SNRs, and three M83 H ii regions.

After obtaining the spectra for each object, we took comparison lamp spectra using HgAr+Ne lamps for wavelength calibration. For flux calibration, we observed several spectrophotometric standard stars from the list of Hamuy et al. (1992, 1994) at the beginning and/or end of each night.

We used common IRAF⁷ tasks to perform the basic data reductions, including bias-subtraction, trimming, and flat-fielding (using dome flats obtained during the day), for both the drift-scan and long-slit 2D spectra.

Because the overall exposure time for most of our targets was more than 1800 s, we investigated the wavelength stability of the data for each object by looking for differences in the centroids of telluric emission lines. The positions of the telluric lines in the object spectrum taken closest in time to the corresponding comparison lamp spectrum were used as the fiducial values. If the telluric emission lines in another spectrum of that object were not well aligned with the fiducial spectrum, we determined the pixel offset between that spectrum and the fiducial, shifted the spectrum to match the fiducial, and combined the individual spectra. After all the spectra for a given object had been aligned and combined, we used the comparison spectra and IRAF tasks in the twodspec.longslit package to rectify each 2D spectrum to a linear wavelength scale.

We also derived a wavelength solution for the all the standard star spectra, combined these for each night, and determined a nightly sensitivity function, which we applied to all the object spectra for that night. Finally, for each object, we then used the IRAF task background to subtract the night-sky emission, selecting as background the portion(s) of the slit that fell outside the target object.

Figures 1 and 2 provide representative examples of the spectroscopic and imaging data we obtained toward the LMC SNR sample. The upper panel of Figure 1 shows the Hα image of the SNR N86 obtained by the Goodman using its direct imaging mode, where the blue dashed rectangle outlines the area scanned in spectroscopic mode. Figure 2 shows the Hα image of the LMC SNR B0520−69.4, before (left) and after (right) continuum subtraction. One has no difficulty identifying N86 because of its relatively high surface brightness filaments and its location in a sparse field on the western edge of the LMC. However, the crowded star field and the relatively low surface brightness of the emitting filaments in SNR B0520−69.4 (see Figure 2 top) make it more difficult to identify its structure until after continuum subtraction.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The background-subtracted 2D spectra of N86 and B0520−69.4 (lower panels of Figures 1 and 2, respectively) reveal complex kinematic structure, despite the crowded star field in the latter case. Both of these show multiple red- and blueshifted components, especially near the slit centers that tracked across the central part of the objects, where we expect the line-of-sight velocities to be highest. Such complex structure is seen in virtually all of our scanned data for LMC SNRs.

While the 2D spectra are necessary to reveal the full complexity of the velocity structure, it is also useful to extract 1D spectra from the scanned data in order to provide a single quantitative measure for the velocity broadening. The 1D spectra from LMC objects also provide a direct comparison with spectra from SNRs and other objects in M83 and other more distant galaxies. Therefore, for each of the scanned spectra, we summed over all the lines containing the object, omitting small regions where bright stars crossed the slit. In four of the spectra of our LMC primary targets (N49B, N59B, N63A, and N132D) we also detected a second component attributable to nearby diffuse Hα emission. In these cases we extracted 1D spectra for both the SNR and H ii region components separately.

Figure 3 shows examples of extracted 1D spectra from B0520−69.4. The lower trace shows the extracted spectrum from the scanned observation (lower panel of Figure 2), while the upper trace is from the position of the slit indicated on the upper panels of Figure 2. The scanned data clearly reveal a more complex velocity structure and broader emission lines with multiple peaks than does the long-slit spectrum. The same is true for the other objects observed in both scanned and long-slit modes, as shown in Table 2.

Zoom In Zoom Out Reset image size

Table 2.  Extracted Fluxes and Line Widths

Object	Mode	Hα	Hα	Hα	[N ii]	[N ii]	[S ii]	[S ii]	${L}_{{\rm{H}}\alpha }$ d
	Scan/LS	Fluxa	(FWHM)b	(20% Width)c	Fluxa	(FWHM)b	Fluxa	(FWHM)b
			($\mathring{\rm A}$)	($\mathring{\rm A}$)		($\mathring{\rm A}$)		($\mathring{\rm A}$)
LMC SNRs:
N4	Scan	23.00	2.35	3.62	4.03	2.39	13.90	2.39	22.8
N11L	Scan	35.60	2.83	4.53	7.09	2.89	26.90	2.78	10.8
	LS	90.50	2.59	3.63	17.10	2.63	66.80	2.50	⋯
N86	Scan	62.00	2.40	3.93	11.20	2.46	45.40	2.30	41.1
	LS	108.00	1.87	3.02	16.70	1.80	68.50	1.87	⋯
N103B	Scan	43.30	4.03	5.44	7.57	4.53	11.99	3.90	4.3
N120	Scan	190.00	1.67	2.72	36.30	1.87	96.80	2.75	70.8
	LS	114.00	1.94	3.02	29.00	1.94	107.50	2.04	⋯
B0520-69.4	Scan	18.70	4.94	6.04	6.24	4.65	17.78	4.54	13.0
	LS	9.58	1.55	2.42	2.25	1.59	6.05	1.58	⋯
N132De	Scan	28.10	1.56	2.42	5.94	1.58	15.50	1.40	14.8
N49B	Scan	27.90	5.58	5.44	9.82	3.81	14.30	4.61	5.1
	LS	31.70	5.83	5.43	16.00	4.86	19.90	4.78	⋯
N49	Scan	468.00	5.28	7.22	106.00	5.52	355.00	5.03	232.0
DEML-205	Scan	189.00	1.66	2.72	27.30	1.64	71.40	1.82	235.0
N206	Scan	55.30	1.63	2.42	12.10	1.85	44.00	2.56	48.1
N63A	Scan	149.00	2.21	3.93	26.70	2.34	128.80	1.88	33.3
N59B	Scan	67.30	1.88	2.72	10.20	1.81	32.40	1.87	41.8
DEM-L316B	Scan	27.70	4.28	5.76	4.54	4.72	27.50	4.50	22.0
DEM-L316A	Scan	32.70	3.73	3.93	5.54	4.07	27.50	3.48	13.0
LMC H ii Regions:
N80	Scan	74.30	1.11	1.81	9.49	1.24	15.18	1.28	16.9
N38	Scan	297.00	1.27	1.81	22.70	1.18	33.50	1.17	103.0
HII(N132D)-E	Scan	3.09	1.34	2.11	0.54	1.19	1.17	1.21	1.6
HII(N132D)-W	Scan	2.56	1.30	2.11	0.62	1.16	1.32	1.16	1.3
HII(N49B)	Scan	26.90	1.18	1.81	6.27	1.18	17.11	1.12	4.9
HII(N63A)	Scan	70.80	1.20	1.81	7.94	1.11	10.84	1.15	15.8
HII(N59B)	Scan	351.00	1.53	2.11	33.50	1.52	78.70	1.58	218.0
N175	Scan	130.00	1.25	1.81	15.30	1.15	26.70	1.16	96.9
M83 Objects:
B12-023	LS	1.79	3.66	5.43	1.89	5.40	1.78	5.54	⋯
B12-039	LS	4.06	1.59	2.41	2.74	1.83	2.45	1.94	⋯
B12-180	LS	0.77	2.87	4.83	1.16	3.99	0.97	4.05	⋯
B12-191	LS	0.48	2.14	4.22	0.70	3.52	0.50	2.89	⋯
B12-208	LS	0.27	1.67	2.71	0.29	3.52	0.26	3.39	⋯
M83-HII-1	LS	25.90	1.53	2.41	7.96	1.39	4.64	1.46	⋯
M83-HII-2	LS	3.67	1.19	1.81	1.42	1.10	0.58	1.11	⋯
M83-HII-3	LS	0.52	1.35	2.11	0.22	1.45	0.12	1.32	⋯

Notes.

^aFluxes are in units of 10⁻¹⁵ erg cm⁻² s⁻¹. [N ii] flux is for λ6583 only while [S ii] flux is total for the doublet. Note that these do not represent the total flux from the entire object, but for scans is an average flux through the slit as it moved across the object. For long-slit spectra, these are simply the total flux through the fixed 06 slit. See the text, Section 4.2. ^bSingle Gaussian FWHM values in Å; [N ii] is for λ6583 and [S ii] is for λ6716; see the text. ^cLine widths at 20% of peak value, in Å; see the text. ^dThe Hα luminosities are in units of 10³⁵ erg s⁻¹. ^eN132D is embedded in a region of strong H ii emission, which gives a strong narrow component to the lines, producing artificially narrow Gaussian fits. Fainter broad emission is evident in both the 2D and 1D spectra (see Figure 4).

Download table as:  ASCIITypeset image

For the two slit positions we observed in M83, each object crossed by the slit (SNRs and serendipitous H ii regions) extended over only a few lines (≲2'') along the slit. Here we simply summed over the relevant lines to obtain the 1D spectra. Several examples of the extracted spectra from LMC and M83 SNRs and H ii regions are shown in Figures 4 and 5, respectively. The extracted 1D spectra from the scanned LMC SNR sample and from the M83 long-slit SNR sample both show the same general pattern of emission lines that are broadened in comparison to those in the spectra of H ii regions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To obtain a quantitative comparison between SNRs and H ii regions, we determined the spectral widths of the emission lines in our sample using two different methods for both the scanned and long-slit data. In the first method we simply measured the widths of the lines at 20% of the peak value, while for the second we fit each of the principal lines in the spectra—Hα, the [N ii] doublet, and the [S ii] doublet—with simple, single-component Gaussians. Simply measuring the width of the profile at a particular percentage of the peak has the advantage that it is relatively insensitive to asymmetries in the profiles, but it is quantized at the scale of an individual pixel. Gaussian fits do not suffer from this limitation; they readily give measures for the total flux in the lines; and they make comparisons with spectra obtained from different instruments more straightforward. In Table 2, we present the 20% results only for Hα, while we tabulate the FWHM results for all three lines, both for SNRs and H ii regions. As expected, the Hα 20% widths are larger than the FWHM values for individual objects, and the 20% widths are significantly higher for SNRs than for H ii regions.

From the data presented in Table 2, we conclude the following:

• 1.  
  In general the H ii regions have narrower line profiles than do the SNRs. This can be seen qualitatively in Figures 4 and 5 where the emission lines in the SNRs are broader than the emission lines in the H ii regions. In Figure 6, we plot the spread of values for the FWHM from the Hα line for the scanned LMC SNRs and H ii regions and the long-slit data of M83 SNRs and H ii regions. As seen in this figure, the H ii region line profiles are close to the instrumental resolution of  1.2 Å. Therefore, with sufficiently high spectral resolution, it is easy to separate the H ii regions from the SNRs (see Section 4.1 for details).
• 2.  
  The FWHM and 20% widths of the Hα line give similar results and are correlated, as seen in Table 2 and in the left-hand side of Figure 7. In general the FWHM of the Hα line is smaller than the 20% width of the Hα line. Hence, even though the Gaussian fits to the line profiles do not capture the full complexity, and in some case extent, of the line profiles, they can be used effectively to discriminate between H ii regions and SNRs.
• 3.  
  Our experiments with long-slit spectra make it clear that the kinematic distinction between SNRs and H ii regions is less effective when observing single positions of spatially resolved SNRs in long-slit mode than it is when observing the entire SNR (see Figure 3). This is probably what one would have expected a priori, since one generally selects slit positions to overlay the brightest portion of an SNR, often along the rim with high surface brightness but relatively low line-of-sight shock velocity.
• 4.  
  For the LMC SNRs, the FWHM of Hα and other lines are comparable (see Figure 7); however, in M83 there is a clear tendency for the [S ii] and [N ii] line widths to be greater than those of Hα, indicative of contamination of the Hα line by adjacent or overlying photoionized emission that could not be accurately subtracted. This indicates that velocity broadening of forbidden lines may be the preferred diagnostic in galaxies beyond the Local Group.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We have obtained drift-scan spectra toward 15 SNRs and eight H ii regions in the LMC in addition to long-slit spectra of five SNRs and three H ii regions in M83. We scanned across the LMC targets in order to measure the integrated line profiles of the SNRs and H ii regions so that we can compare the results to the measured properties of SNRs and H ii regions in more distant galaxies where objects can be centered in a single long-slit spectrum. Below we discuss our results in detail.

4.1. LMC H ii Regions versus SNRs

4.2. Integrated Line Flux Measurements and Luminosity Comparison with Other Galaxies

For our spectra scanned over entire objects, we have the opportunity to obtain integrated line fluxes, and corresponding luminosities, for SNRs in the LMC. After sky subtraction and extraction of 1D spectra for each object, we obtained the integrated flux in Hα (and other lines). But since only a small fraction of the emitting region lies within the slit at any time during the scan, we must then divide by an "inefficiency factor": ($\mathrm{slit}\ \mathrm{area})/(\mathrm{scan}\ \mathrm{area})=(\mathrm{slit}\ \mathrm{width})/(\mathrm{scan}\ \mathrm{distance})$. For example, for N4 the 06 slit scanned at 400'' hr⁻¹ during exposures of 1800 s (Table 1), so the fraction of the scan covered at any instant is 06/200'' = 0.003. The results for the 15 LMC SNRs with scanned spectra are given in Table 2. We measure Hα luminosities from $4\times {10}^{35}\,\mathrm{to}\ 2.4\times {10}^{37}\,\mathrm{erg}\ {{\rm{s}}}^{-1}$ (assuming a distance of 50 kpc and no absorption).

As a check on these values, in Figure 8 we compare them with those obtained by summing over regions containing the SNRs in narrowband Hα mosaic images of the LMC from MCELS (Smith & MCELS Team 1999; Paredes et al. 2015). (As noted above, we used these images in selecting targets for scanned spectra.) We find that fluxes (and luminosities, of course) are fairly well correlated in these two independent data sets, though the MCELS image-based values are typically larger than those from scanned spectra by about a factor of 2. This can be attributed to at least three factors: (1) the MCELS Hα filter was broad enough (FWHM 30 Å) to transmit [N ii] emission at the ∼30% level, contaminating the Hα emission; (2) the regions measured in the MCELS images included somewhat more emission than what was covered in the Goodman spectrograph scans; and (3) the presence of nonuniform diffuse Hα emission toward the SNRs complicated sky subtraction such that the sky-background was never completely removed. Each of these three effects would produce an overestimate of the MCELS-derived Hα luminosities of the SNRs in comparison to the values determined using the scanned Goodman data.

Zoom In Zoom Out Reset image size

It is interesting to compare these values with those for more distant galaxies, where the integrated fluxes have been obtained through narrowband imaging and/or through long-slit spectra.⁹ For SNRs in M33, Long et al. (2010) measured Hα fluxes using narrowband images from the Local Group Galaxies Survey (LGGS, Massey et al. 2007), and found values up to a maximum of $2.7\times {10}^{37}\,\mathrm{erg}\ {{\rm{s}}}^{-1}$. Lee & Lee (2014a) also used Hα images from the LGGS to do similar luminosity measurements for M31, where they found SNRs (and candidates) to have Hα luminosities up to 1.0 × 10³⁷ $\mathrm{erg}\ {{\rm{s}}}^{-1}$.

For M83, SNR line fluxes have been measured both in narrowband images from Magellan (Blair et al. 2012) and in multislit spectra from Gemini-S (Winkler et al. 2017). As Figure 5 of the latter paper shows, there is a tight correlation between the two data sets; the vast majority of objects have Hα fluxes from imaging between 1 and 2 times those measured from spectra. The luminosities for M83 SNRs range up to ∼5 × 10³⁷ $\mathrm{erg}\ {{\rm{s}}}^{-1}$, a few times brighter than the most luminous ones in M31 or M33. This is probably attributable to the fact that active star formation is taking place in M83, with an attendant higher mean ISM density compared with the more quiescent M31 or M33 (Winkler et al. 2017).

Within the context of these three much larger galaxies, we see that the brightest SNRs in the LMC have luminosity higher than those in M31, comparable with those in M33, and a factor of about two lower than those in M83. We note, however, that we do not present a complete luminosity function for LMC SNRs, for we have sampled only about 25% of them. Since we intentionally selected objects with high surface brightness, our sample should be at least close to the bright end of the luminosity function; however, there may be other high-luminosity objects that we did not include because they are too large to be covered by the 39 Goodman slit. Nevertheless, we can conclude that the brightest SNRs in the LMC have luminosities that at least approach those in M83.

4.3. Comparison with M83 SNRs and H ii Regions

4.4. Line Widths and Shock Models

In addition to being a discriminant between H ii regions and SNRs, higher resolution observations also communicate something about the SNR itself. The LMC SNRs we have observed range in diameter from approximately 6.8 pc in the case of N103B to nearly 60 pc in the case of N86. Analysis of the X-ray spectrum of N103B suggests it is a relatively young, 860 yr old, SNR produced by a thermonuclear explosion (Hughes et al. 1995), while N86 is an old SNR, dominated by emission from swept-up interstellar material, with an age of perhaps 53,000 yr (Williams et al. 1999). Given this, it is not surprising that the widths of the lines in N103B are greater than in N86. However, as shown in Figure 9, any trends of line width with diameter or age are fairly weak. All very large diameter, very old SNRs do have fairly narrow line widths, but objects with diameters ≲30 pc and ages ≲10,000 yr show a large range in observed line widths.

Zoom In Zoom Out Reset image size

On the other hand, it is important to remember that at 6500 Å, a velocity of 300 $\mathrm{km}\,{{\rm{s}}}^{-1}$ corresponds to 6.5 Å, so the velocity widths we observe are far less than the primary shock velocity (especially for the young objects), even if one allows for the fact that the FWHM is less than the full widths at zero intensity. Qualitatively, we can understand this as follows. Consider an SNR with a primary shock velocity of 1000 $\mathrm{km}\,{{\rm{s}}}^{-1}$ expanding into an ISM with a mean density of 1 cm⁻³. The primary shock will be nonradiative, i.e., material behind the shock will not have had time to cool to ∼10,000 K, as needed to produce bright Hα and the forbidden lines typically observed in optically identified SNRs. There are SNRs where these fast shocks are observed, the so-called Balmer-dominated SNRs (see, e.g., Heng 2010).¹⁰ In these, the only observed optical radiation arises from H atoms undergoing ionization, because the temperature behind the shock is 10⁷ K or more. A few such SNRs do exist in the LMC, but they are faint and are not among the ones we have observed. Furthermore, objects such as these are too faint to be observed in more distant galaxies.

In a typical SNR, the optical emission we detect arises as the primary shock drives secondary shocks into clouds with densities of 10–100 times that of the average ISM. The pressure behind these secondary shocks will be in approximate equilibrium with that of the primary shock, that is ${v}_{2}\sim \sqrt{({n}_{1}/{n}_{2})}\ {v}_{1}$) or 100–300 $\mathrm{km}\,{{\rm{s}}}^{-1}$. The initial temperature behind these shocks is still rather high, of order 10⁶ K, but at these temperatures and densities the shocked gas will cool relatively quickly to the point where it produces significant amounts of Hα and forbidden-line radiation. So except in cases where we are viewing the ejecta directly, as in Cas A, we do not expect the lines to be much broader than we have observed here.

Large numbers of SNRs have been and are continuing to be identified in nearby galaxies using ratios derived from narrowband filter imaging to distinguish SNRs from the principal source of contamination, i.e., H ii regions. Because all of the SNRs in a nearby galaxy lie at essentially the same distance and because absorption along the line-of-sight is generally low and less variable than in the plane of the Galaxy, such samples are important for understanding the class properties of SNRs. To carry out such studies accurately, however, one must be sure that the sample is not contaminated by objects that are not SNRs. Here we suggest that one very clear way to eliminate imposters from a sample is to take spectra with sufficient spectral resolution (∼50 $\mathrm{km}\,{{\rm{s}}}^{-1}$) to separate SNRs from H ii regions.

To test this suggestion, we have obtained spectra of 15 SNRs and of 8 H ii regions in the LMC with the SOAR telescope with a spectral resolution of 53 $\mathrm{km}\,{{\rm{s}}}^{-1}$. To ensure that our spectra of the LMC objects are similar to those typically obtained from objects in more distant galaxies, we scanned the slit across each entire LMC object, in order to obtain "global" spectra. In a few cases, we also obtained simple long-slit spectra at a single position to compare with our scans. We also obtained a few spectra of SNRs and H ii regions in M83 as a preliminary test of this technique for a more distant galaxy. Our conclusions are as follows:

• 1.  
  Spectroscopy with ∼50 $\mathrm{km}\ {{\rm{s}}}^{-1}$ resolution is indeed an effective way to separate SNRs from H ii regions, even for the largest LMC SNRs in our sample. The typical Gaussian velocity widths of the SNRs are ∼130 $\mathrm{km}\ {{\rm{s}}}^{-1}$ (FWHM)¹¹ whereas the H ii region lines appear near the instrumental resolution of ∼50 $\mathrm{km}\ {{\rm{s}}}^{-1}$.
• 2.  
  The widths of lines in the "global" (scanned) spectra are generally larger than the lines in corresponding simple long-slit spectra, because of bulk velocity shifts in the various portions of the SNR. The velocity widths of the lines in our spectra from Hα, [N ii], and [S ii] are all similar for objects with little contaminating H ii region emission. For the M83 SNRs, such contamination does seem to affect Hα more than the [N ii] and [S ii] lines, making these forbidden lines that are typically strong in SNRs somewhat more effective as the kinematic diagnostic.
• 3.  
  The integrated Hα luminosity for LMC SNRs ranges up to at least 3 × 10³⁷ $\mathrm{erg}\ {{\rm{s}}}^{-1}$, somewhat higher than for M31 or M33 SNRs, and approaching the values found in the starburst galaxy M83.
• 4.  
  While the smaller diameter SNRs often have higher velocity widths in their global spectra, there is considerable scatter at small diameter, consistent with the idea that density (and possibly explosion energy) strongly affect the appearance of SNRs at the same diameter.

In summary, we believe that kinematics provide a clean way to discriminate between SNRs and H ii regions at optical wavelengths. We plan, and encourage others, to pursue higher resolution spectroscopy of the various SNR samples in nearby galaxies as a way of resolving the uncertain identifications of objects, especially those near the often-used [S ii]:Hα = 0.4 boundary where the largest uncertainty lies. This will result in more complete, less contaminated samples of bona fide SNRs, which are needed for understanding the global properties of SNRs and their impact on their host galaxies.

This work was based on observations obtained at the Southern Astrophysical Research (SOAR) telescope, which is a joint project of the Ministério da Ciência, Tecnologia, Inovações e Comunicações (MCTIC) do Brasil, the U.S. National Optical Astronomy Observatory (NOAO), the University of North Carolina at Chapel Hill (UNC), and Michigan State University (MSU). S.D.P. would like to thank P. Ugarte, C. Corco, and S. Pizarro for their help at the telescope without which these observations would not have been possible. P.F.W. acknowledges support from the NSF through grant AST-1714281. W.P.B. thanks the Center for Astrophysical Sciences at Johns Hopkins University for support.

Facility: SOAR (Goodman). -