A New, Larger Sample of Supernova Remnants in NGC 6946

We carried out narrowband imaging observations of NGC 6946 from the 3.5 m WIYN telescope and MiniMosaic imager on Kitt Peak on the nights of 2011 June 26–28 (UT).⁶ The so-called "Minimo" was mounted at the f/6.3 Nasmyth port and consisted of a pair of 2048 × 4096 SITe chips, with a field 96 square at a scale of 014 pixel⁻¹. We used interference filters that pass lines of Hα, [S ii] λ6716, 6731, and [O iii] λ5007, plus red and green narrowband continuum filters so we could subtract the stars and produce pure emission-line images. Frames in each filter were dithered to enable automatic removal of cosmic rays and bad pixels. Further observational details are given in Table 1.

It is noteworthy that the Hα filter was quite narrow in bandwidth, 27 Å FWHM. Its transmission is 69% at the rest wavelength of Hα, but only 11% at 6548 Å and 16% at 6583 Å; hence, the [N ii] lines are greatly attenuated relative to Hα.⁷ This facilitates identification of SNRs based on their image-derived [S ii]:Hα ratios. Seeing throughout this run was about 1'' ≈ 32 pc at the distance of NGC 6946. While sufficient for identifying SNRs in all but the most crowded regions, this resolution was insufficient for obtaining properties such as diameter or morphology.

We used standard IRAF⁸ techniques for processing the images, including overscan correction, bias subtraction, and flat-fielding using dome flats. Procedures in the IRAF mscred package were then used to combine the data from the individual chips into a mosaic image for each frame, assigned a WCS for each using stars from the USNOB1 catalog (Monet et al. 2003). We then stacked all the images by filter onto an arbitrary standard coordinate system with a scale of 020 pixel⁻¹ and scaled and subtracted the continuum images from the emission-line ones (red from Hα and [S ii]; green from [O iii]) to give pure emission-line images with most of the stars and galaxy background light removed. Finally, we used observations of seven different spectrophotometric standard stars—all selected for their weak Hα absorption lines—from the catalog of Massey et al. (1988) to flux calibrate the emission-line images. Figure 1 shows a color version of the final images (R = Hα, G = [S ii], B = [O iii]), with the positions of historical SNe in NGC 6946 also indicated.

Zoom In Zoom Out Reset image size

To select SNR candidates, we used the DS9 display program to show the continuum-subtracted WIYN images in all three emission lines as well as a [S ii]:Hα ratio image and a continuum image (to discriminate stars or stellar subtraction residuals from point-like nebulae). We then visually inspected these to select SNR candidates based on a high [S ii]:Hα ratio. The initial inspection was carried out by Middlebury undergraduate Marc DeLaney; subsequently two of us (WPB and PFW) compiled independent candidate lists; we then conferred to agree on a consensus list of 147 candidates, including the 27 from MF97. The positions of all 147 candidates are shown in Figure 2. The vast majority of the SNR candidates appear within the spiral arms or clustered on the outskirts of large complexes of H ii emission and star formation.

Zoom In Zoom Out Reset image size

Figure 3 shows an example for a small region ∼15 south of the nucleus (see Figure 1) containing two of the MF97 objects and several new SNR candidates. In this example, we have combined the three continuum-subtracted emission-line images into a single color panel that shows how the stronger [S ii] and/or [O iii] emission from the SNR candidates makes them stand out. The [S ii]:Hα ratio image was a key diagnostic for drawing our eyes to potential objects of interest. Then an assessment of the integrity of the candidate as an emission object was made by inspection of the actual images. This was done to separate candidates from stellar residuals or false regions of higher ratio in the ratio map that were due to noise.

Zoom In Zoom Out Reset image size

Having performed similar SNR searches in other galaxies such as M33 and M83, it is worth noting some differences for NGC 6946. M33 is of course much closer, and the nebulae of interest are almost always resolved. M83 is much more distant (4.6 Mpc) than M33, but our search there was aided by the exceptional 05 seeing conditions we obtained at the Magellan telescope (Blair et al. 2012). Many objects were resolved, but others extended down to the limits of what even HST could resolve (∼1–2 pc) (Blair et al. 2014). Our WIYN survey of NGC 6946 did not have exceptional seeing, and the distance is some 50% larger than for M83. Hence, relatively few of the nebulae of interest are resolved, and with variations in the complex galaxy background, it is much more difficult to perform a systematic search. The use of the [S ii]:Hα ratio image was particularly helpful for NGC 6946, which is the most distant galaxy for which we have performed this kind of ground-based SNR survey. Nonetheless, while we have expanded the SNR candidate list very substantially, clearly completeness has not been achieved. Our list of 147 SNR candidates falls far short of M83 for example (with 300+), even though the SN rate is larger in NGC 6946. Higher-resolution data (better seeing and/or HST imaging) would no doubt help substantially, but the greater distance for NGC 6946 is still a limiting factor.

In addition, we initially selected 51 emission nebulae with relatively high [O iii]:Hα ratios (using an [O iii]:Hα ratio image in the display). Most of these nebulae were expected to be H ii regions but we hoped that one or more might be a young, ejecta-dominated SNR, similar to Cas A in our Galaxy. With 10 SNe in the last century and a high incidence of massive stars, one might expect a number of young, ejecta-dominated SNRs to be present. Though none of the O-selected candidates had ratios as extreme as seen for Cas A or 1E0102-72.3 in the Small Magellanic Cloud, we nevertheless selected some of them for follow-up spectroscopy. None of these nebulae for which we obtained spectra have any indication of ejecta in their spectra or the broad lines would expect from a very young SNR. All are H ii regions with somewhat enhanced ionization state, and so these form the bulk of the H ii sample we use for comparison with the SNR candidates sample below. Finally, we inspected the positions of the nine⁹ historical SNe in NGC 6946 for evidence of nebular emission; we only detected line emission at the positions of SN1980K and SN2004et, both of which we targeted for follow-up spectroscopy. (See Section 4 below.)

The SNR candidates are listed in R.A. order in Table 2. For each candidate, we provide (1) a source name, (2,3) the position (J2000), (4) the deprojected galactocentric distance (GCD), (5) the Hα flux as derived from the emission-line images, (6) the [S ii]:Hα ratio measured from the images, (7) the spectrum we used to confirm the imaging ratios (see below), (8) whether or not the object has a spectroscopic confirmation that the [S ii]:Hα ratio is ≥0.4 (see below) and (9) other names for the source.

We used the Gemini Multi-Object Spectrograph (GMOS) on the 8.2 m Gemini-North telescope to obtain all the spectra reported here, during queue-scheduled programs in semesters 2014B (program GN-2014B-Q-83) and 2015B (program GN-2015B-Q-91). For the 2014B program, we designed six custom masks, each with 20–30 slitlets targeting SNR candidates whose positions we determined from our 2011 WIYN images, together with short R-band pre-images of several NGC 6946 fields taken with GMOS earlier in 2014 as part of the spectroscopy preparation program. We used two additional masks (which we refer to as masks 7 and 8 for simplicity) for the 2015B program. Slitlets in one or more of our eight masks were placed on 102 distinct SNR candidates, including ones with a range of sizes, GCDs, and ISM environments (locations in arms and in inter-arm regions). In addition to the SNR candidates, we also placed a number of slitlets on H ii regions for comparison purposes in both 2014 and 2015.

We used the 600 line mm⁻¹ grating designated G5307 and a GG455 cut-off filter to block second-order spectra. The detector in both years was a mosaic of three e2v deep-depletion CCD chips, binned by two in the spatial direction (for a scale of 0146 pixel⁻¹) and by four in the dispersion direction. The dispersion was 1.84 Å pixel⁻¹ (binned), resulting in coverage of the spectral range from at least Hβ through [S ii] λ6716, 6731 for virtually all the objects.¹⁰ Our masks had slitlet widths from 125 to 175, with wider slits used for the larger objects, and lengths of 6'' or longer to permit local background and sky subtraction.

With each mask, we took spectra at three or four slightly different grating tilts, to cover wavelength gaps between chips and to gain somewhat more total spectral range.¹¹ At each wavelength setting, we obtained two or more identical exposures to minimize the effects of cosmic rays. For calibration, we programmed quartz flats and CuAr arc frames immediately before or after the science exposures with each mask and grating setting. A journal of all the science observations from both 2014 and 2015 appears in Table 3. The SNR candidates for which we obtained spectra are indicated by the small red boxes in Figure 2.

The data were processed using standard procedures from the gemini package in IRAF for bias subtraction, flat-fielding, wavelength calibration, and combination of spectra with different grating tilts to provide the final results. Flux calibration was based on baseline GMOS observations of a few spectrophotometric standard stars, carried out in the same semester as part of standard GMOS operations.

During the processing, the two-dimensional (2D) spectra from different slitlets were separated to give individual 2D spectra from each slitlet. We examined each of these individually and selected the object region, as well as one or more sky background regions, stripped out 1D spectra of each, and then subtracted the sky spectrum from the corresponding object to obtain the final background-subtracted object spectra. Many of the objects are located in regions with bright surrounding galactic background (both continuum and emission lines) from NGC 6946, so the selection of a representative local background was done on a best-effort basis. In addition to the targeted SNR candidates and H ii regions, we extracted spectra from other H ii regions that appeared by chance in the slits when this was possible. This allowed us to increase our sample of H ii regions from 24 to 65. We then performed fits to obtain emission-line fluxes from the spectra, assuming Gaussian profiles, for the following lines and line complexes: Hβ alone, the [O iii] doublet, the [O i] doublet, the Hα-[N ii] region, and the [S ii] doublet. For the fits, we assumed that the background varied linearly with wavelength around each line, and that the FWHM of all lines in each complex was the same.

Representative examples of the spectra that were obtained are shown in Figure 4. The three SNRs were selected primarily to show how the quality of the spectra changes as a function of brightness. The spectroscopically obtained [S ii]:Hα ratio for L19-048 was 0.45, just above the value for spectroscopic confirmation, while those for the previously known bright object L19-097 = MF-15 and the faint candidate L19-096 are higher. All three SNR candidates show evidence of emission from [O i], which is another indicator that the emission we see arises from shock-heated gas.

Zoom In Zoom Out Reset image size

Table 4 lists the information we obtained for the SNR candidates for which we obtained spectra. Specifically we list (1) the source name, (2) the extracted Hα flux, (3–9) ratios of various emission lines to Hα [taken to be 300], (10) the total [S ii]:Hα ratio and (11) the measured FWHM of the lines in the Hα-[N ii] complex. For doublets, where the line ratio is constrained by atomic physics, that is [O iii], [O i], and [N ii], we have listed only the stronger line. We visually inspected all of the spectra and the fits to them; values that we judge to be more uncertain are indicated with tildes in the table. No allowance has been made for additional errors associated with difficulties in background subtraction. A number of the objects were observed with more than one mask, sometimes with different slit orientations. In these cases, we used the spectrum that we judged to be the most accurate and reported it for reference in the "Spectrum" column in Table 2, where (for example) 05.18 should be interpreted as "mask 5 slitlet number 18".

MF97 identified 27 SNR candidates in NGC 6946. All of these objects have [S ii]:Hα ratios in our WIYN images that exceed 0.4. MF97 obtained spectra of six of their candidates. We have obtained spectra of 23 of the MF97 objects, including new spectra of four objects for which MF97 had spectra—MF-03, MF-21, MF-22, and MF-26. All of these indeed have measured spectroscopic ratios that exceed 0.4. We note that MF97 used a relatively conservative value of [S ii]:Hα ≥ 0.45 to establish their catalog (to avoid the issue of errors in the determined ratio affecting objects just above or below the normal 0.4 criterion). Hence, it is perhaps not surprising that the previous objects are strongly confirmed here. MF97 report "typical 15 seeing" for their work. (MF97 looked at a number of galaxies and they do not give a specific value of seeing for NGC 6946.) Hence, it is also clear that MF97 were only able to find a combination of objects that were relatively bright (so not smeared out by seeing) and/or were relatively isolated from contaminating emission. They estimated that at least four times as many SNRs were likely present in NGC 6946, and our current survey has far surpassed that estimate.

Even though we have identified many more candidates than MF97, it is clear that with its better seeing and higher sensitivity, our survey is still limited and likely to be significantly incomplete. For example, Figure 3 shows two objects, L19-067 and L19-075, both in close proximity to H ii contamination, that we were able to identify as candidates while MF97 could not. However, it is not hard to imagine any number of additional objects in the many even more confused emission regions that our survey would have missed.

The object MF-16, listed here as L19-098, deserves separate mention as it is far away and the brightest object in our catalog. Originally thought to be a possible example of an exceedingly bright (and possibly multiple) SNR (Blair & Fesen 1994; Blair et al. 2001), similar to the extraordinary SNR in NGC 4449 (Blair et al. 1983; Milisavljevic & Fesen 2008), X-ray variability was subsequently established that clearly indicates the presence of an accreting black hole binary within the nebular complex (Roberts & Colbert 2003; Fridriksson et al. 2008; Rao et al. 2010). Most recent analyses (Kaaret et al. 2010; Berghea & Dudik 2012) model the system based on the ULX binary only, but clearly, the system involves some combination of shock-heated and X-ray photoionized emission. Dunne et al. (2000) show resolved line profiles on the bright emission lines indicating kinematic motions of order 250–400 km s⁻¹, and while Roberts & Colbert (2003) show the bulk of the X-ray emission is likely due to the ULX, they estimate the SNR component could be as bright as $\sim 2.5\times {10}^{38}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$ in X-ray, which is quite substantial for a SNR. Although jets are often invoked for accreting ULX binaries, the HST images of the nebula (Blair et al. 2001) are not obviously consistent with this idea; the morphology shows a multiple loop structure, and the likely optical counterpart of the ULX is not centered in the smallest, brightest loop. Hence, the idea that the complex involves something more complicated than a single SN that created the ULX binary may still be relevant to consider in understanding the overall characteristics of this intriguing object.

As shown in Figure 6, the density-sensitive [S ii] ratio λ6716:λ6731 clusters around the low-density limit of 1.4 for the SNR candidates, and the fact that about as many objects have non-physical ratios above 1.4 as below suggests that (a) the ratios for some faint objects have significant errors (not unexpected), and (b) likely almost all the objects are close to this limit. This is in contrast to the situation in M83 (Winkler et al. 2017) and to a lesser extent in M33 (L18) where a significant number of the SNRs show evidence of high densities, especially for smaller diameter objects. In the absence of good SNR diameters here, we cannot search for trends with diameter, but upcoming Hubble Space Telescope observations should provide accurate diameter information for many of these objects. These HST images could also reveal the presence of very small SNR candidates, perhaps with high densities, which eluded detection in our ground-based images.

Zoom In Zoom Out Reset image size

Figure 7 shows the observed Hβ:Hα ratios of the SNR candidates with spectra as a function of GCD. Nearly all of the SNRs show significant reddening, as one would expect since, at 12° from the Galactic plane, foreground reddening from within our Galaxy along the line of sight is expected to be $E(B-V)=0.29$ (Schlafly & Finkbeiner 2011). There is clearly substantial internal and differential reddening within NGC 6946 as well, as there is a very evident trend for objects near the center of NGC 6946 to be more reddened than those more distant from the nucleus.

Zoom In Zoom Out Reset image size

Our SNR line ratios show a general decrease with increasing GCD as seen in Figure 8, although the dispersion at any particular GCD distance is large. The trend could well be indicative of abundance gradients in nitrogen and sulfur, with the dispersion being due to varying shock conditions and/or or local abundance variations. However, the trend stands in contrast to SNRs in M33, where both the [N ii]:Hα and [S ii]:Hα ratios have a large range and do not decrease systematically with increasing GCD (although the line ratios in the H ii region sample appear well behaved—see L18 Figure 8). There is a good correlation between the [N ii]:Hα and [S ii]:Hα ratios of the various objects, as shown in Figure 9, as also seen in other galaxy samples (see Winkler et al. 2017, Figure 9).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Line ratios in SNR spectra are expected to vary both as a function of shock conditions and metallicity. To see where the SNRs in NGC 6946 lie, we have compared the line ratios calculated from models by Allen et al. (2008) using the MAPPINGS III code for a range of shock velocities (100–1000 km s⁻¹) and pre-shock magnetic fields (10⁻⁴–10 μG). The results are shown in Figure 10. The results both for the ratio of [O iii] λ5007:Hβ and for [S ii]:Hα fall squarely into the region of the solar metallicity models. This is consistent with expectations for previous abundance studies of NGC 6946 such as Cedrés et al. (2012), Gusev et al. (2013), and references therein, depending of course on the adopted method of determining H ii region abundances from strong-line data only. Both of these papers also show a very modest abundance gradient in the H ii regions of NGC 6946, with considerable scatter about the mean, although very few H ii regions are sampled in the inner portion of the galaxy (see Cedrés et al. Figure 18).

Zoom In Zoom Out Reset image size

Searches for SNRs in NGC 6946 have been carried out in several other wavelength ranges. In particular, Lacey & Duric (2001) identified 35 radio sources as SNR candidates on the criterion that these sources had nonthermal spectral indices and were positionally coincident with Hα emission. These objects are located mostly in the spiral arms of NGC 6946 where there is active star formation and where one might expect SNe to explode. There are seven emission nebulae in our list of SNR candidates that lie within 2'' of radio SNR candidates; this is five more than had been identified previously, but a small fraction of the total number of radio objects. Lacey & Duric (2001) argued that the reason that few radio SNRs are detected optically is that the bright Hα emission from H ii regions makes optical searches for SNRs less sensitive in the spiral arms than in the rest of the galaxy. The fact that more of the optical SNRs are not detected at radio wavelengths is most likely a question of sensitivity. All of the SNR candidates identified by Lacey & Duric (2001) have radio luminosities of at least one-tenth of Cas A, much greater than the bulk of SNRs known in the Galaxy. By contrast, in M33, where White et al. (2019) have recently conducted a very deep radio survey with the Jansky Very Large Array, more than three-quarters of the optically identified SNRs have been detected at radio wavelengths.

An alternative diagnostic (to the [S ii]:Hα ratio) for identifying SNRs in at least some external galaxies is emission in the [Fe ii] 1.64 μm line. Since Fe⁺ is so easily ionized further, H ii regions are expected to have little if any [Fe ii] while [Fe ii] should, like [S ii], be elevated in the cooling tail behind SNR shocks. Hence, detection of an emission nebula with [Fe ii] is a strong indication of shock heating. In M83, where HST WFC3 IR imaging in [Fe ii] is available (Blair et al. 2014), about 40% of the optical SNRs in the observed region were detected in [Fe ii], and a handful of compact [Fe ii] nebulae in particularly dusty regions are strong SNR candidates whose emission is too highly absorbed to be detected optically. This raises the possibility that [Fe ii] might be valuable not only to help confirm optical SNR candidates, but also to help to obtain a more complete sample in heavily reddened regions.

In NGC 6946, Bruursema et al. (2014) carried out ground-based interference filter imagery in the light of [Fe ii] 1.64 μm. Ground-based [Fe ii] imaging is difficult due to sky contamination, as noted by Bruursema et al. (2014), but they were able to identify 48 candidate objects that they felt were above the noise. Interestingly, only three of these objects align with SNR candidates in our sample: L19-076, L19-095 = MF-14, and the exceedingly bright ULX MF-16 = L19-098 (discussed above). We are thus left wondering whether the other [Fe ii] objects are possible SNR candidates or whether the data quality issues are responsible for the large difference between NGC 6946 and M83. We can say, however, that the bulk of the Bruursema et al. (2014) candidates are not seen in projection onto the dustiest regions. M33, for which Morel et al. (2002) clearly detected a solid handful of optical SNRs in ground-based [Fe ii] observations, appears to be intermediate between the extremes of NGC 6946 and M83. High spatial resolution HST WFC3 IR observations would make a large impact in clarifying the situation for a galaxy as distant and as highly absorbed as NGC 6946.

SNRs are also X-ray sources, and therefore we have looked to see how many X-ray sources in NGC 6946 could be found in our candidate lists. The most detailed X-ray study of NGC 6946 to date was carried out using Chandra by Fridriksson et al. (2008), who constructed a catalog of 90 point sources, of which 25 appeared to be time variable (and hence likely X-ray binaries or background AGN). Of the 90 point sources, there are eight that are positionally coincident with objects in our sample. X-ray hardness ratios reported by Fridriksson et al. (2008) show that most of these have relatively soft X-ray spectra, consistent what is expected for thermal emission from a SNR. The main exception is F08-08, coincident with L19-029, which has a hardness ratio that is more typical of X-ray binaries and background galaxies. Not surprisingly, one of the X-ray sources coincident with L19 objects is the ULX L19-098 = MF-16. These two are also the only X-ray sources coincident with SNR candidates that also show evidence of (long-term) variability, according to Fridriksson et al. (2008).

Higher percentages of the optically identified sources have been X-ray-detected in M33 (112/200, Long et al. 2018) and M83 (67/225, Long et al. 2014). Of these, M83 is the more relevant for comparison. M83, a nearly face-on grand-design spiral, has a star formation rate of 3–4 M_☉ yr⁻¹ (Boissier et al. 2005), similar to NGC 6946, but lies a distance of 4.61 Mpc (Sahu et al. 2006) compared to 6.7 Mpc for NGC 6946. M83 was observed for 725 ks with Chandra, compared to a total of about 175 ks for NGC 6946. In addition, NGC 6946 is relatively close to the Galactic plane and as a result foreground absorption reduces the X-ray sensitivity, particular below 1 keV. Indeed, the hydrogen column density along the line of sight to M83 is $4\times {10}^{20}\,{\mathrm{cm}}^{-2}$, whereas for NGC 6946 it is $1.8\times {10}^{21}\,{\mathrm{cm}}^{-2}$ (Kalberla et al. 2005). For a thermal plasma with an effective temperature of 0.6 keV, the combination of greater distance and higher absorption implies that a typical SNR in NGC 6946 would have only about one-third the count rate of one in M83. Consequently, it is not surprising that we have detected fewer SNRs in X-rays in NGC 6946 than in M83.