New ASKAP Radio Supernova Remnants and Candidates in the Large Magellanic Cloud

These ASKAP radio data comprise a twelve-pointing mosaic taken of the LMC at a centre frequency of 888 MHz (bandwidth of 256 MHz, spatial resolution of 13.9$\times$12.1 arcsec² and position angle $-84\degr$). Details of the observations, data reduction and analysis of these data have been presented in Pennock et al. (2021). The image shown in Fig. 1 is affected by various artefacts and missing short spacing data. However, its robust maximum angular scale of 49 arcmin combined with improved sensitivity (Filipović et al., 2022) ensures a reasonable environment to search for new LMC SNR. At the time of these observations (April 2019) ASKAP didn’t have polarisation capability, and therefore, the ASKAP image used here is total intensity (Stokes I).

We observed MCSNR J0522–6543¹¹1To distinguish between SNR candidates and confirmed SNR, Maggi et al. (2016) established nomenclature where bona fide Magellanic Clouds (MCs) SNR are named with the prefix ‘MCSNR J’ and candidates with only ‘J’. with the Australia Telescope Compact Array (ATCA) on 30^th November 2019 and 23^rd February 2020 (project codes C3275 and C3292; see Appendix A.8). The observations were carried out in frequency switching mode (between 2100 MHz and 5500/9000 MHz) with 1-hour of integration over a 12-hour period using arrays 1.5C and EW367. The Compact Array Broadband Back-end (CABB) with its 2048 MHz bandwidth was used centred at $\nu$=5500 and 9000 MHz, totaling 117.62 minutes of integration. At the same time, we also used the $\nu$=2100 MHz band with a total of 112.2 minutes of integration. The primary (flux) calibrator was PKS B1934–638 ($S_{2100}$=11.651, $S_{5500}$=5.010 and $S_{9000}$=2.704 Jy) and secondary calibrator (phase) was PKS B0530–727 ($S_{2100}$=0.680, $S_{5500}$=0.585 and $S_{9000}$=0.695 Jy). For imaging, we used WSClean (Offringa et al., 2014) and flagged out data from the 6^th antenna for all observations to focus on the extended emission of this object at the expense of reducing the resolution. We achieved a resolution of 20.95$\times$16.60 arcsec² at 2100 MHz, 15.85$\times$12.79 arcsec² at 5500 MHz and 8.81$\times$7.05 arcsec² at 9000 MHz (all at P.A.=0 degrees) with a corresponding rms noise of 0.1, 0.05 and 0.025  mJy beam^-1, respectively.

The optical data used in this study consists of H$\alpha$, [Sii], [Oiii] images from the Magellanic Cloud Emission Line Survey (MCELS) survey; more details can be found in Points et al. (2019).

To gain insight into the type of stellar environment hosting the suspected progenitors of these objects, we use data sourced from the Magellanic Clouds Photometric Survey (MCPS; Zaritsky et al., 2004). To this end, we construct colour-magnitude diagrams and identify blue stars more massive than $\sim$8 M_☉ within a 100 pc (6.9 arcmin at the distance of the LMC; $\sim$10⁷ yr at 10 km s^-1) radius of each of the objects presented in this study. The size of SNR can well exceed 150 pc in diameter and therefore our chosen search area will give us a good view of the environment and possible origin of these objects as demonstrated in Bozzetto et al. (2017). This allows us to see the prevalence of early-type stars close to the candidate remnants.

We also consult various X-ray surveys including ROSAT (Haberl & Pietsch, 1999) and XMM-Newton (Haberl, 2019). Finally, we use the Spitzer infrared survey of the LMC, “Surveying the Agents of a Galaxy’s Evolution” (SAGE) (Meixner et al., 2006).

For atomic hydrogen, we used the ATCA & Parkes Hi survey data from Kim et al. (2003) with its angular resolution of 60 $\times$ 60 arcsec². We made the velocity channel maps and position-velocity ($p-v$) diagrams to constrain the velocity ranges of Hi clouds that physically interact with the SNR. The velocity channel maps provide a spatial correspondence between the SNR shell boundary and Hi clouds as tested by previous studies (e.g. Moriguchi et al., 2005). When the Hi clouds are associated with the SNR, they are expected to be located along with the shell. Likewise, Hi cavities in $p-v$ diagrams are also expected when supernova shocks and/or stellar winds from the progenitor evacuated (accelerated) the Hi gas.

In order to define the shock-interacting velocity range of Hi clouds, we first prepared velocity channel maps of Hi towards each SNR candidate. Because the shock-interacting clouds will be limb-brightened in the synchrotron radio continuum through the shock-cloud interactions with the magnetic field amplification, the velocity channel distribution of Hi is useful to identify the spatial correlation between the shocked Hi clouds and SNR candidate shells.

Here, we prepared a velocity channel distribution of Hi as a velocity step of 6.6 km s^-1 (see Appendix A). The map was centred at each SNR candidate with a size of $\sim$50 $\times$ 50 arcmin². Finally, we searched for expanding shell-like structures in the $p-v$ diagram by changing the integration range of R.A. When the expanding gas motion is identified, we selected the Hi velocity range which traces the inner edge of the cavity in the $p-v$ diagram.

We first investigate the present LMC SNR candidate sample of 20 objects. Yew et al. (2021) found that none of their 15 optically selected candidates can be detected in our ASKAP survey. In addition to these, Filipović et al. (2022) suggested that ASKAP J0624–6948 is an intergalactic SNR positioned in the outskirts of the LMC, but the true nature of this object remains mysterious.

This leaves four remaining LMC SNR candidates that are initially investigated in Bozzetto et al. (2017): J0506–6815 ([HP99] 635), J0507–7110 (DEM L81), J0538–6921 (MC 73) and J0539–7001 ([HP99] 1063)). We search for their radio continuum SNR signature in our ASKAP survey (see Fig. 2) as well as in other wavelengths.

We find that J0506–6815 ([HP99] 635) is a circular source in our ASKAP radio image (see Figs. 2 (top left panel) and 3) with an estimated flux densities of $S_{\rm 888\,MHz}$=78$\pm$2 mJy and $S_{\rm 1377\,MHz}$=64$\pm$4 mJy (from Filipović et al. (2021)) produce a spectral index $\alpha=-0.45\pm 0.24$. Together with a prominent central soft X-ray emission in the XMM-Newton survey and [Sii]/H$\alpha$$>$0.9, we now safely confirm MCSNR J0506–6815 as a bona fide SNR. Interestingly, the X-ray emission occupies central part of the SNR while radio and optical emission dominates at the edges. This anti-correlation indicates that MCSNR J0506–6815 could be very similar to the known iron-rich cores SNR such as MCSNR J0506–7025 and MCSNR J0527–7104 (Kavanagh et al., 2016, 2022).

SNR candidate J0507–7110 (DEM L81; Fig. 2 top right panel) shows some typical SNR morphological characteristics, but in our ASKAP LMC radio image this object is confused by the nearby emission and other sources. Therefore, we are keeping its status as a candidate SNR.

J0538–6921 (MC 73; Fig. 2 bottom left panel) is certainly a prototype radio SNR candidate based on its radio properties ($S_{\rm 888\,MHz}$=294$\pm$3 mJy; $S_{\rm 1377\,MHz}$=248$\pm$4 mJy; $S_{\rm 4800\,MHz}$=120$\pm$10 mJy; $\alpha=-0.56\pm 0.02$) but lack of X-ray or optical confirmation prevent us from a final classification.

We clearly see a point radio source close to the centre of X-ray LMC SNR candidate J0539–7001 ([HP99] 1063; Fig. 2 bottom right panel), but we cannot confirm its SNR nature as the positional displacement between radio and X-ray position of $\sim$25 arcsec is not negligible. Also, no other nearby and obvious fine-scale structure could be linked with the object’s possible SNR nature. However, we acknowledge that a compact type of SNR candidate such as J051327–6911 (Bojičić et al., 2007) is still a possibility.

We searched our new ASKAP images of the LMC for circular shaped objects (above 5$\sigma$ of local noise) with an enhanced ratio of radio continuum to H$\alpha$ emission. This method, introduced by Ye et al. (1991), successfully identifies objects in which non-thermal emission is dominant, i.e. SNR. Using this enhanced ratio, combined with the typically circular SNR morphology, we find 14 regions in which the presence of an SNR is plausible. We present a catalogue of radio images of these newly discovered SNR candidates in Fig. 4 and details in Appendix A.

We used the method described in Hurley-Walker et al. (2019, Section 2.4) and Filipović et al. (2022) to measure the position, extent and flux densities of all 14 selected LMC SNR candidates. In the same manner, we measure the three ATCA observations of MCSNR J0522–6543 (at 2100, 5500 and 9000 MHz). In short, we carefully selected proposed SNR regions that exclude all obvious point sources and then measured the total radio flux density, accounting for local background. We note the selected SNR candidates have low surface brightness and/or are sometimes embedded in complex environments (c.f. J0457–6823, J0459–7008b²²2Because of the close proximity (and similar position) to SNR N 186D, we use ‘b’ (as J0459–7008b) to distinguish between two possibly separate objects. and J0459–6757). This significantly influences the accuracy or may prevent any meaningful measurements. We estimate our flux density measurements have an error of $<$20 per cent. Also, Filipović et al. (2022, see their Figure 7) noted that a lack of short spacing data in observations³³3We also used the same 888 MHz image and the same ATCA array observations with the similar integration time. may result in under-estimates of extended flux density measurements. Filipović et al. (2022) suggest that spectral index estimates could be as much as $\sim$15–20 % (or $\alpha\sim$0.1) flatter.

In Table 1 we list the source name in Column 1, its central position (RA and DEC) in Columns 2 and 3, source angular extent as major and minor axis/diameter in Column 4 (in arcsec), arithmetic average of major and minor axes/diameter converted to parsecs for a distance of 50 kpc in Column 5, and position angle (PA) in Column 6, measured from north to east. Our flux density estimates at 888 MHz are listed in Column 7. As none of the here selected LMC SNR candidates (except for MCSNR J0522–6543) can be seen at other radio frequencies, we can not estimate their spectral index. To compare the surface brightness of our sample with established LMC SNR, we assumed a typical SNR spectral index of $\alpha=-0.5$ (Reynolds et al., 2012; Galvin & Filipovic, 2014; Bozzetto et al., 2017; Maggi et al., 2019; Filipović & Tothill, 2021b). However, we acknowledge that some of our SNR candidates may contain an old PWN which would have a somewhat flatter spectral index. Using this assumed spectral index allowed us to estimate the flux density and surface brightness of these sources at 1 GHz (Column 8; $\Sigma_{\rm 1\,GHz}$). In Column 9 we provide the number of massive OB stars found within a $\sim$100 pc radius of each of the objects’ central position. Lastly, in Columns 10 and 11, Y/N apply when optical and X-ray data are available and/or the source is detected.

These 14 new SNR candidates escaped previous detection because of their low surface brightness, which indicates an advanced age. They are most likely evolved and expanding in a rarefied environment, and we note that they occupy the bottom part of the SNR $\Sigma$–D diagram (Pavlović et al., 2018; Urošević, 2020). In fact, the arithmetic average of surface brightness ($\Sigma_{\rm 1\,GHz}$; Table 1, Column 8) from the sample of 40 LMC SNR (Bozzetto et al., 2017) is 7.9$\times 10^{-21}$ W m^-2 Hz^-1sr^-1 while from our sample of 14 new candidates is 4.5$\times 10^{-22}$ W m^-2 Hz^-1sr^-1. This order of magnitude difference suggests that we are discovering low surface brightness SNR in the LMC. At the same time, our sample SNR candidate diameters are fractionally larger but because of the large overlap with the sample of established LMC SNR ($D_{\rm av14SNR}$=51.2 pc and SD=16.9 vs. $D_{\rm av71SNR}$=44.9 pc and SD=24.9) we can not separate two samples.

The size distribution of the 14 new SNR candidates against the 71 (58+1+1+1+3+7) previously confirmed SNR from Bozzetto et al. (2017), Maggi et al. (2016); Maitra et al. (2019, 2021), Yew et al. (2021) and Kavanagh et al. (2022) can be seen in Fig. 5. Our new sample of 14 LMC SNR candidates are distributed across a large range of sizes from $\sim$30 to $\sim$96 pc. From the sample of 71 known LMC SNR, 45 objects have an estimated age, which spreads from the small size SN1987A of 35 yrs to the large 107 pc DEM L72 of 115000 yrs (Klimek et al., 2010). Bozzetto et al. (2017, see their fig. 18) showed that at age of $\sim$5000 yrs SNR such as N 49 (Park et al., 2012) or 30 Dor B (a.k.a. N 157B; Chen et al., 2006) can reach sizes of $>$30 pc in diameter which is mid-to-late Sedov phase. However, this is very much dependent on various factors such as the environmental density, initial progenitor type and its explosion energy. Nevertheless, this implies that our sample of 14 new objects belongs to a more evolved and mid-to-older ($>5000$ yrs) SNR population.

We also cross-matched various X-ray surveys and pointed observations including XMM-Newton and Chandra. However, one would not expect to detect X-ray emission from an SNR in the very late stage of evolution in such a rarefied environment, especially at the distance of 50 kpc, as shown in the above-mentioned case of J0624–6948 (Filipović et al., 2022). SNR X-ray emission depends on ISM density squared, while synchrotron emission scales linearly with ISM density (Duric, 2000a, b; Urošević, 2022). This implies that X-ray emission in less dense environments will fade much quicker than predominantly non-thermal (synchrotron) radio emission. In some cases (8/14) we can detect very weak optical emission (mainly H$\alpha$) which indicates active non-radiative shocks from the late Sedov and radiative phases, implying an older SNR.

Optical emission from SNR typically originate from highly compressed, thin, radiative shells. This may further indicate the eight optically-identified SNR candidates are not expanding in a low-density medium. In 6 candidates, we did not detect optical emission, which suggests that these SNR candidates are in the radiative phase of their evolution. However, in the radiative phase, there could be significant emission from various coolants, depending on temperature ($T$). Most notably, we may see [Oiii] emission from $T\sim$10⁵ K gas, as the X-ray emission of such shocks fades. However, for $T\sim$10⁵ K there is little line emission in the MCELS optical bands that we use here. This all implies that in many cases, the X-rays from SNR fade first followed by the optical emission and then the last standing emission before the remnant completely disappears would come from radio continuum (Vink, 2020). That suggests, the eight optically-detected candidates are perhaps not as old as those in which we detect only radio emission. We also detect one SNR candidate (J0542–6852) in the XMM-Newton survey which strengthens our SNR classification for this source.

We investigate the Hi properties of our SNR candidate sample (see Appendix A). As a result, $\sim$80% (11 out of 14) of them show possible evidence of expanding Hi shells, which were likely formed by a combination of shock waves and strong stellar winds from the massive progenitor (and possibly surrounding OB stars). As shown in the Appendix A, several SNR show a good spatial correspondence between the intensity peaks of the radio continuum and Hi, suggesting that shock-cloud interactions occurred. Indeed, our $p-v$ diagrams show cavities in which these SNR are almost freely expanding (Sano et al., 2020; Luken et al., 2020; Alsaberi et al., 2019). Some of these 14 LMC SNR candidates (such as J0451–6906, J0457–6823, J0459–6757, J0534–6720, J0534–6700, J0542–6852 and J0543–-6923 have reached the wind-cavity of Hi, while others (such as J0452–6638, J0504–6901, J0522–6543 ) are in the free expansion phase inside the wind bubble. This is very similar to Galactic SNR SN 1006 (Sano et al., 2022), G346.6–0.2 (Sano et al., 2021b) and the mixed-morphology W49B (Sano et al., 2021a) where a wind-blown bubble was found along the radio continuum shell with an expansion velocity of $\sim$10 km s^-1, which was likely formed by strong stellar winds from the high-mass progenitor of the SNR.

Finally, we investigate the stellar environment of each of our 14 SNR candidates (Fig. 6) using a 100 pc ($\sim$10⁷ yr at 10 km s^-1) search radius. The number of OB stars ($N_{\rm OB}$) in the environments of these objects is an indicator of whether they are more likely to be core-collapse rather than type Ia SNRs, as such short-lived stars are a direct tracers of star formation activity that must have occurred recently in the core-collapse scenario. This indicator was calibrated in the LMC in Maggi et al. (2016) using SNR types determined via other methods. Numbers of neighbouring OB stars less than 15 are associated with type Ia SNR while CC SNR had generally (much) more than 35 such stars around them. The overall number of OB stars in our candidates trend toward the lower end of what is observed in other confirmed LMC SNR with several in the ‘undecided’ range. The outlier here is J0534–6700 which has 325 blue early-type stars in the immediate vicinity of the remnant which point toward a most likely core-collapse scenario. However, in the SN-cavities we also expect a small number of neighbouring OB stars which suggests relatively low-mass (8-10 M_⊙) SN II-P progenitors. The local ratio of red supergiants to OB stars might corroborate such a hypothesis as well.