A New Microquasar Candidate in M83

UV/optical/infrared bands. Blair et al. (2014) and Dopita et al. (2010) carried out HST Wide Field Camera 3 (WFC3) imaging to search for SNRs in M83. The galaxy was covered by a mosaic of seven WFC3 fields, which were imaged in nine narrow-band and broadband filters. The first two fields (including the nuclear region) were observed in 2009 August and 2010 March (ID 11360; PI: R. O'Connell) and the other five were observed from 2012 July to 2012 September (ID 12513; PI: W. P. Blair). We refer the readers to Blair et al. (2014) for the details on the WFC3 observational setup (e.g., orientation of the seven fields, set of filters used for the survey, and exposure times). The HST imaging survey was combined with a ground-based imaging survey (Blair et al. 2012) with the Inamori Magellan Areal Camera and Spectrograph on the 6.5 m Magellan telescope; and by follow-up spectrophotometric observations (Winkler et al. 2017) with the Gemini Multiobject Spectrograph (GMOS) on the 8.2 m Gemini-South telescope.

X-ray band. We observed M83 with the Advanced CCD Imaging Spectrometer (ACIS) on board the Chandra X-ray Observatory, with ten visits spaced between 2010 December 23 and 2011 December 28 (Observing Cycle 12), for a total of 729 ks; in all those observations, the target galaxy was placed on the back-illuminated ACIS-S3 chip. We refer to Long et al. (2014) for a comprehensive discussion of the Chandra/ACIS setup and its astrometric calibration. In addition, we integrated the 2010–2011 dataset with the ACIS-S observations of 2000 April 29 (51 ks; Cycle 1) and 2001 August 4 (10 ks; Cycle 2), which provide higher sensitivity to the softest energies.

Radio band. We mapped M83 with the ATCA, with three sets of observations: with 3 × 12 hr in 2011, 2015, and 2017. The data were recorded simultaneously at central frequencies of 5.5 and 9.0 GHz, with a bandwidth of 2 GHz at each frequency band. The 2011 and 2015 observations were taken with the telescope in its most extended 6 km configuration, while the 2017 observations were taken with the telescope in its more compact 1.5 km configuration. The more compact configuration provided more short baselines to increase the sensitivity of our final radio map to diffuse emission. See Long et al. (2014) for a preliminary report on the results based on the 2011 ATCA observations alone. A full presentation of the radio results is in preparation (T. D. Russell et al. 2020, in preparation).

To create a list of objects for study, we inspected the images of SNRs in the HST data, looking for sources with distinct bipolar shape, hot spots, or other unusual asymmetrical morphology. At 4.6 Mpc, the 004 pixel scale of the WFC3-UVIS camera corresponds to just over 1 pc, providing excellent data for assessing the morphology of even the smallest objects expected to be of relevance. Based on this inspection, we selected nine objects for more detailed study, as listed in Table 1. This list is not necessarily exhaustive, as there are a number of other objects within the total SNR candidate sample in M83 that might also be considered as showing interesting and unusual morphologies. Here, we have chosen the clearest examples of SNRs and SNR candidates with peculiar morphologies in an attempt to understand the physical nature of such objects.

The list includes the two sources that were flagged as "worthy of special mention" by Dopita et al. (2010, their Section 3.1 and Table 2) because of their jet-like or bipolar morphology. Those two sources are listed as S5 and S6 in our Table 1 (presented in order of R.A.). Further out along the spiral arms, we identified two strong line emitters (S2 and S7 in our Table 1), with interesting elongated morphologies that appear to extend over ∼100 pc in length. We then selected five candidate SNRs with irregular morphology (S1, S3, S4, S8, and S9). Finally, for comparison, we also included the microquasar MQ1 in the table, which we reanalyzed in the same way as for the other nine sources. The spatial distribution of the selected objects projected on an Hα image of M83 is shown in Figure 1. Narrow-band and broadband HST images of the fields around each object are shown in Figures 2–4. Seven of the nine candidate SNRs have been detected in X-rays; those without a Chandra detection are S6 and S7.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is worth noting that what we define here as source S2 was identified by Blair et al. (2012) as two separate SNR candidates, B12-96 and B12-98. However, there is an elongated line-emission feature that connects the two optical sources (Figure 2), and there is a single, point-like Chandra counterpart (L14-139) located roughly in the middle between the two optical peaks. Therefore, for the purpose of this study, we consider the whole complex as a single astrophysical object with a peculiar, extended morphology.

It is also useful to clarify our definition of S7 (discussed in more details in Section 3.1). The field around this source contains a blue and a red star cluster separated by about 20 pc (Figure 4) along a north–south axis. The strong optical/IR line emission and radio emission coincide with the reddened star cluster, and that is what we define as the S7 source in our study. There is no X-ray source associated with the reddened star cluster; there is, instead, a Chandra source (consistent with an ordinary SNR) at the location of the blue cluster (dashed error circle in Figure 4) but that is not the object of this study.

HST. Linear combinations of WFC3 broadband filter images were used to create rescaled continua, which were subtracted from the corresponding narrow-band images, as described in Blair et al. (2014). For this work, we used the continuum-subtracted images already prepared for that earlier study. We used the imaging and photometry tool ds9, to select source and background regions for our target objects, and extract the instrumental count rates. We used the PHOTFLAM parameter for the relevant filters in WFC3-UVIS and WFC3-IR detectors to convert from count rates to flux densities, and the filter widths listed in the WFC3 Instrument Handbook to convert to total fluxes in each (narrow) band.

Chandra. We downloaded and reprocessed the data from the Chandra archives, for all the ACIS observations. We used standard tasks within the Chandra Interactive Analysis of Observations version 4.10 (Fruscione et al. 2006). First, we reprocessed the data with chandra_repro. Then, we applied the merge_obs task, which reprojects all the data sets to the same tangent point and merges them, creating a coadded, exposure-corrected image. Filtered images in different energy bands were obtained with the task dmcopy. We used specextract to build spectra and response files individually from each observation and combine them into a merged spectrum, because the X-ray flux of typical SNRs is too faint for meaningful spectral analysis of individual observations.

ATCA. The ATCA radio observations were flagged and calibrated following standard procedures within MIRIAD (Sault et al. 1995). Flux and phase calibration were done using PKS 1934−638 and 1313-333, respectively. The data were then imaged within CASA (McMullin et al. 2007). To help minimize sidelobes from the bright central region, images were created with a Briggs robust parameter of −1. Individual images were created at each central frequency (5.5 and 9 GHz), where data from each epoch were stacked to provide the highest sensitivity at each observing band. We also created a single image including both frequency bands from all epochs to increase sensitivity (T. D. Russell et al. 2020, in preparation).

Gemini spectroscopy. Eight of the nine SNRs candidates studied in this work were observed with GMOS. We used the results published by Winkler et al. (2017). In the context of our present work, we obtain three important pieces of information from the Gemini results, which enhance and integrate the information provided by the HST images: (i) an estimate on the optical extinction, from the observed ratio of Hα and Hβ fluxes (assuming a theoretical intrinsic ratio of 2.86); (ii) a means to distinguish the fractional contributions of Hα and [N ii]λλ6548, 6584 to the combined emission imaged by the HST/WFC3-UVIS camera in the F657N filter; (iii) a constraint on the electron density of the ionized gas, from the flux ratio of the [S ii]λλ6716, 6731 doublet.

We measured the observed continuum-subtracted optical fluxes in the four HST/WFC3 narrow-band filters: F502N, F657N, F673N, and F164N (Table 2). For all sources except S7, the flux ratio between the F673N and F657N bands (proxy for the standard diagnostic ratio of [S ii]λλ6716, 6731/Hα) is between ≈0.2 and 0.5. S7 stands out with a much lower ratio of ≈0.07. This is consistent with the Gemini/GMOS spectral classification of S7 as a photoionized H ii region rather than a shock-ionized SNR (Winkler et al. 2017).

Table 2.  Observed Fluxes/Count Rates in Various Bands: Data from Chandra, HST, and ATCA

ID	Count Rate (0.35–8 keV)	${F}_{[{\rm{O}}{\rm{III}}]}$	${F}_{{\rm{H}}\alpha +[{\rm{N}}{\rm{II}}]}$	${F}_{[{\rm{S}}{\rm{II}}]}$	${F}_{[\mathrm{Fe}{\rm{II}}]}$	${S}_{9\mathrm{GHz}}$	S5.5 GHz
	(10−3 ct s−1)	(10−15 CGS)	(10−15 CGS)	(10−15 CGS)	(10−15 CGS)	(μJy beam−1)	(μJy beam−1)
S1	0.154 ± 0.021	0.83 ± 0.08	3.8 ± 0.5	1.3 ± 0.1	0.50 ± 0.05	140 ± 18	200 ± 15
S2	0.455 ± 0.028	2.6 ± 0.9	23.6 ± 2.0	7.7 ± 1.1	⋯	71 ± 12 (East)	120 ± 20 (East)
						82 ± 12 (West)	130 ± 20 (West)
S3	0.186 ± 0.020	2.3 ± 0.2	3.1 ± 0.3	0.86 ± 0.08	0.29 ± 0.03	60 ± 10	59 ± 10
S4	0.327 ± 0.026	1.2 ± 0.1	5.1 ± 0.5	1.8 ± 0.2	0.40 ± 0.04	190 ± 20	280 ± 20
MQ1	1.823 ± 0.052	0.35 ± 0.04	5.2 ± 0.6	1.7 ± 0.3	4.4 ± 0.4	1000 ± 200	1800 ± 200
S5	0.234 ± 0.019	2.6 ± 1.0	8.0 ± 1.3	2.9 ± 0.7	0.81 ± 0.07	<80	82 ± 20
S6	<0.02	0.43 ± 0.28	4.3 ± 0.8	2.4 ± 0.5	0.4 ± 0.1	<90	110 ± 20
S7	<0.02	2.4 ± 0.9	50 ± 4	3.7 ± 1.2	1.2 ± 0.2	850 ± 15	1000 ± 15
S8	0.576 ± 0.030	1.2 ± 0.7	5.4 ± 1.2	1.7 ± 0.7	0.78 ± 0.08	52 ± 15	54 ± 10
S9	0.416 ± 0.025	1.7 ± 0.2	6.7 ± 0.6	1.4 ± 0.2	0.37 ± 0.04	<33	<48

Download table as:  ASCIITypeset image

The F502N, F673N, and F164N bands are dominated by [O iii]λ5007, [S ii]λλ6716, 6731, and [Fe ii]λ1.64 μm, respectively. Determining the Hα luminosity from F657N is approximate, because that filter is also passes [N ii]λλ6548,6584. Whenever possible (i.e., for eight of our nine targets, see Table 3), the relative contributions from Hα and [N ii] can be directly measured from the Gemini/GMOS spectra and the actual Hα intensity determined (Winkler et al. 2017). For the shock-ionized objects, the Hα contribution varies between 0.23 and 0.53 times the flux in the F657N band; the average of those values is ≈0.40. A larger contribution of 0.59 is seen in S7: this is again consistent with its classification as a photoionized H ii region (although this value by itself does not rule out a shock-ionized SNR buried within the bright photoionized region).

Table 3.  Line Fluxes, Line Ratios, Densities, and Extinction from the Gemini Spectra

ID	F[O iii]	FHα	FHα/FHβ	F[S ii]/FHα	${F}_{{\rm{H}}\alpha }/{F}_{{\rm{H}}\alpha +[{\rm{N}}{\rm{II}}]}$	F6716/F6731	ne	AV	Ionization
	(10−15 CGS)	(10−15 CGS)					(cm−3)	(mag)
S1	1.0	4.2	4.69	0.48	0.53	1.22	180	1.55	S
S2	1.6	5.6	6.67	0.81	0.44	1.33	84	2.66	S
S3	2.2	1.4	4.76	0.69	0.46	1.17	240	1.60	S
S4	0.66	0.73	5.45	1.62	0.23	1.02	460	2.03	S
S6	0.31	1.2	4.35	1.47	0.33	1.35	70	1.31	S
S7	3.8	72.1	6.00	0.19	0.59	1.15	260	2.32	P
S8	0.95	1.65	4.92	0.99	0.32	0.89	770	1.70	S
S9	1.8	3.0	3.70	0.58	0.49	1.06	390	0.81	S

Note. S5 and MQ1 were not observed with Gemini.

Download table as:  ASCIITypeset image

For the only target that does not have a Gemini/GMOS spectrum (S5), we assumed the average ratio of F(Hα) = 0.4F(F657N) derived above from the shock-ionized sources. Looking at the whole sample of about 140 SNR candidates with Gemini spectra, it is clear (Winkler et al. 2017, their Figures 8 and 9) that there is a large spread in the [N ii]/Hα flux ratios, corresponding to F(Hα)/F(F657N) spanning the whole range between ∼0.25 and 0.7, possibly as a function of shock velocity and local metal abundance. Thus, our choice of F(Hα) = 0.4F(F657N) is roughly the midpoint (on a log scale) of that range. (For comparison, Blair et al. 2014 assumed a slightly lower value of F(Hα) = 0.33F(F657N) for sources without direct line-ratio measurements, and F(Hα) = 0.5 F(F657N) was assumed by Soria et al. 2014 for MQ1.)

The optical spectra also provide direct information on the density of the emitting plasma, from the observed flux ratio of the sulfur doublet ([S ii]λ6716/[S ii]λ6731). The WFC3 filter F673N includes both lines, but the ratio comes from the Gemini spectra. We find (Table 3) a range of values between ≈0.9 and ≈1.35, which correspond to electron densities between ≈70 and 800 cm⁻³, assuming a temperature near ∼10⁴ K for the S⁺ zone (Osterbrock & Ferland 2006; Sanders et al. 2006).

Next, we corrected the observed fluxes for the effect of optical extinction. In addition to the Galactic line-of-sight value (A_V ≈ 0.18 mag), we need to take into account the local extinction in the star-forming disk of M83. For the eight objects with Gemini spectra, we determined how the observed ratio between the Hα and Hβ line fluxes differs from the canonical value of 2.86; we assumed that the reduced contribution from Hβ is due to extinction (A_Hα ≈ 0.818A_V; A_Hβ ≈ 1.164A_V). The observed Hα/Hβ flux ratios span the range ≈3.7–6.7 (Table 3). This corresponds to total extinctions (line-of-sight plus intrinsic) A_V from ≈0.8 mag (for S9) to ≈2.7 mag (for S2). We used the average of those eight values, $\langle {A}_{V}\rangle =1.7$ mag, as a plausible estimate for the unknown extinction of S5. Putting together the previous findings, and scaling to the distance of 4.6 Mpc, we determined the intrinsic luminosities (Table 4) in the [O iii], Hα, and [Fe ii]λ1.64 μm lines (for [Fe ii], in only eight of the nine targets). For comparison, we also recalculated the line emission from MQ1 (Table 4) using the same procedure as for the nine SNR candidates. For MQ1, we adopted a total extinction A_V = 3.9 mag, as discussed in Soria et al. (2014). The higher extinction for MQ1 is not surprising, given its proximity to the dusty nuclear starburst region.

Table 4.  Extinction-corrected Line Luminosities from the HST Data

ID	${L}_{[{\rm{O}}{\rm{III}}]}$	LHα	${L}_{[\mathrm{Fe}{\rm{II}}]}$
	(1037 erg s−1)	(1037 erg s−1)	(1037 erg s−1)
S1	1.0 ± 0.1	1.6 ± 0.2	0.12 ± 0.01
S2	10.3 ± 3.5	19.6 ± 1.7	⋯
S3	3.0 ± 0.3	1.2 ± 0.1	0.10 ± 0.01
S4	2.5 ± 0.3	1.4 ± 0.2	0.14 ± 0.02
S5	3.8 ± 1.5	2.9 ± 0.5	0.27 ± 0.03
S6	0.4 ± 0.3	1.0 ± 0.2	0.13 ± 0.03
S7	6.7 ± 2.5	43.1 ± 3.4	0.45 ± 0.07
S8	1.8 ± 1.0	1.6 ± 0.3	0.26 ± 0.03
S9	1.0 ± 0.1	1.5 ± 0.2	0.11 ± 0.01
MQ1	5.0 ± 0.5	10.0 ± 1.2	2.1 ± 0.2

Download table as:  ASCIITypeset image

[Fe ii]λ1.64 μm is often used as a tracer of cooler gas in star-forming regions and starburst galaxies (e.g., Oliva et al. 1989; Mouri et al. 2000; Alonso-Herrero et al. 2003; Labrie & Pritchet 2006). Fe⁺ has an ionization potential of only 16.2 eV; therefore, it is easily ionized to Fe⁺⁺ in H ii regions, with the result that the [Fe ii] line emission from photoionized gas is strongly suppressed. Instead, Fe⁺ survives in cooler, partially ionized and recombining regions, such as the cooling flows behind SNR shocks. In particular, [Fe ii] traces dense, collisionally excited gas at temperatures of ≈6000 K, and partially ionized, X-ray heated gas at temperatures ≈8000 K (Mouri et al. 2000). If we assume that most of the [Fe ii] emission comes from the SNR recombination zones, we can use the de-reddened flux ratio between [Fe ii]λ1.64 μm and Hβ as an independent check of our estimated values of extinction and Hα/[N ii] flux ratios.

Assuming the canonical value L_Hα ≈ 2.86L_Hβ, we infer from Table 4 that ${L}_{[\mathrm{Fe}{\rm{II}}]}/{L}_{{\rm{H}}\beta }\approx 0.20$–0.45 for seven (specifically, S1, S3, S4, S5, S6, S8, S9) of the eight SNR candidates for which an [Fe ii] measurement is available. This range of values is consistent with the model predictions (Allen et al. 2008) for shock-ionized gas with solar abundances, interstellar medium (ISM) densities in the range n_e ∼ 1–10 cm⁻³, equipartition magnetic field, and shock velocities ≈100–500 km s⁻¹. It is also consistent with the values observed from nearby SNRs (Oliva et al. 1989).

The only outlier in our sample is again the S7 nebula (Figure 5), which has ${L}_{[\mathrm{Fe}{\rm{II}}]}/{L}_{{\rm{H}}\beta }\approx 0.03$ (Table 4). This low value is inconsistent with any shock velocity ≳100 km s⁻¹ (Allen et al. 2008), but is consistent with the interpretation of S7 as an H ii region. Optical and infrared [Fe ii] emission lines are sometimes seen in H ii regions, for example in the Orion nebula (Osterbrock et al. 1992; Rodríguez 1992; Bautista & Pradhan 1998). The infrared [Fe ii] lines are generally attributed to collisional excitation in partially ionized zones close to the ionization front, at densities n_e ∼ 10²–10⁴ cm⁻³ (Bautista & Pradhan 1998; Marconi et al. 1998; Verner et al. 2000).

Given the exceptionally strong Hα (Table 4) and radio (Table 2) luminosity of the S7 nebula, and its peculiar properties compared with the rest of the sample, we investigated this source further. The broadband optical image (Figure 4) of the S7 field shows a highly reddened cluster of massive stars (identified as cluster NGC 5236-2-530 in the catalog of Silva-Villa & Larsen 2011), coincident with the peak of the Hα and [Fe ii] emission, as well as with the position of the unresolved radio emission (discussed later in Section 3.3); it also shows a bright cluster of blue stars located ∼20 pc to the north of the peak radio emission. A soft, faint (L_X ≈ 10³⁶ erg s⁻¹) thermal X-ray source (L14-275: Long et al. 2014) is centered at the location of the blue stars rather than at the position of the optical and radio nebula (Figure 4). From aperture photometry on the broadband HST images, we estimate that the integrated (and de-reddened) absolute brightness of the cluster at the central position of S7 is M_V ≈ −9.7 mag, M_B ≈ −10.0 mag. Using the stellar population code starburst99 (Leitherer et al. 1999), we find that this is consistent with an instantaneous burst of star formation, at an age ≲3 Myr and a stellar mass ≈(1.0–1.5) × 10⁴M_⊙. Such a population is expected to produce an ionizing photon flux Q(H0) ≈ (3.0–4.0) × 10⁵⁰ photons s⁻¹ above 13.6 eV, mostly from main-sequence O stars.

On the other hand, we have already measured the intrinsic Hα luminosity of S7 (Table 4), and we can infer its Hβ luminosity by applying the standard Balmer decrement of 2.86 for photoionized gas at 10,000 K (Osterbrock & Ferland 2006); we obtain L_Hβ ≈ 1.5 × 10³⁸ erg s⁻¹. We also know from atomic physics that the emission of one Hβ photon is triggered every ∼8.5 ionizing photons between 13.6 eV and ∼0.2 keV (Osterbrock & Ferland 2006). Hence, the observed Balmer-line emission implies an ionizing photon flux Q(H0) ≈ 3.1 × 10⁵⁰ photons s⁻¹, in good agreement with the brightness of the optical continuum. Interestingly, the other cluster of blue stars to the northeast of S7 is brighter in the blue band because it is dominated by blue supergiants at an age of ∼10 Myr, but it has an ionizing flux an order of magnitude lower than the highly reddened cluster coincident with S7.

It is interesting to compare our results (mostly based on line emission) with the findings of Williams et al. (2019), who fitted the HST/WFC3 broadband colors of the stars within a 50 pc radius of most of the SNR candidates in M83. From that modeling, Williams et al. (2019) determined the age distribution of the stellar populations and the corresponding zero-age main-sequence mass of the most massive stars currently present around each SNR (a proxy for the progenitor mass of the SNR). For eight of the nine SNR candidates in our target list (all except S7), the stellar progenitor mass spans a range between ≈8 M_⊙ (for S8) and ≈17 M_⊙ (for S3), which is the typical mass range of SN progenitors (Williams et al. 2019). By contrast, they find a zero-age main-sequence mass ≈40 M_⊙ for the most massive stars around S7 (region 207 in their catalog). This is consistent with our interpretation that the S7 optical nebula is mainly the result of photoionization by O stars. Note that the 50 pc source region used by Williams et al. (2019) includes not only the reddened star cluster but also most of the blue cluster.

Zoom In Zoom Out Reset image size

Seven of the nine candidate SNRs selected for this study have point-like X-ray counterparts (Table 1). Using the ciao task srcflux, we estimated and compared the count rates and fluxes of each source in each of the 12 Chandra/ACIS-S observations available in the archive, searching for variability between observations (the count rates are not high enough to constrain intra-observational variability). X-ray variability by an order of magnitude (Soria et al. 2014) was one of the main clues that suggested a microquasar interpretation for MQ1 (Chandra source L14-237). We find significant variability only in one SNR candidate: L14-139, counterpart of S2. The variability of L14-139 is less dramatic than that of MQ1; it spans a factor of a few in flux over the whole series of observations, with no clear systematic trend. In Table 5, we report the model-independent, observed fluxes in the 0.5–7 keV band ("broad" band in srcflux) for each of the 12 observations, and a model-dependent flux in the same band, based on an absorbed power law with photon index Γ = 1.5 and total absorbing column N_H = 10²¹ cm⁻².⁸ All other sources are consistent with constant fluxes throughout the 12 observations, as we expect from hot plasma in SNRs. The lack of an X-ray counterpart for the S7 nebula, down to a detection limit of ≈5 × 10³⁵ erg s⁻¹, supports the idea that S7 is an H ii region, photoionized by stellar sources.

Table 5.  Observed Flux from the Microquasar Candidate S2 from the Individual Chandra Observations

ObsID	MJD(start)	Date	${F}_{0.5-7}$	Model ${F}_{0.5-7}$
			(10−15 erg cm−2 s−1)	(10−15 erg cm−2 s−1)
793	51663.58	2000 Apr 29	${7.2}_{-1.8}^{+1.8}$	${8.8}_{-2.2}^{+2.2}$
2064	52157.00	2001 Sep 4	${14.5}_{-7.6}^{+10.4}$	${7.7}_{-4.1}^{+5.5}$
12995	55553.43	2010 Dec 23	${3.0}_{-1.0}^{+1.1}$	${5.0}_{-1.6}^{+1.9}$
13202	55555.69	2010 Dec 25	${6.7}_{-1.6}^{+1.5}$	${6.5}_{-1.5}^{+1.5}$
12993	55635.50	2011 Mar 11	${3.8}_{-1.1}^{+1.5}$	${6.1}_{-1.8}^{+2.3}$
13241	55638.90	2011 Mar 18	${5.7}_{-1.4}^{+1.3}$	${7.3}_{-1.7}^{+1.8}$
12994	55643.15	2011 Mar 23	${4.1}_{-0.8}^{+0.8}$	${6.4}_{-1.2}^{+1.3}$
12996	55649.67	2011 Mar 29	${4.8}_{-1.9}^{+2.5}$	${3.7}_{-1.5}^{+1.8}$
13248	55654.30	2011 Apr 3	${3.9}_{-1.4}^{+1.7}$	${4.4}_{-1.5}^{+1.9}$
14332	55802.77	2011 Aug 29	${1.7}_{-0.9}^{+1.1}$	${2.2}_{-1.1}^{+1.4}$
12992	55808.25	2011 Sep 4	${3.6}_{-1.2}^{+1.4}$	${3.6}_{-1.1}^{+1.4}$
14342	55923.41	2011 Dec 28	${6.0}_{-1.6}^{+1.9}$	${6.0}_{-1.6}^{+1.9}$

Note. Model-independent and model-dependent fluxes calculated with the ciao task srcflux. The fluxes are not corrected for absorption.

Download table as:  ASCIITypeset image

The second step of our X-ray analysis is an assessment of the average source colors. We used the soft (0.35–1.1 keV), medium (1.1–2.6 keV) and hard (2.6–8.0 keV) photon fluxes already calculated by Long et al. (2014), and plotted them in the traditional color–color diagram often used for X-ray source classifications (Prestwich et al. 2003). We find (Figure 6) that six of the sources are located in the band usually occupied by SNRs, or more generally by any source consisting of an optically thin thermal-plasma emission at (average) temperatures ∼0.5 keV; see, e.g., Long et al. (2014), Soria & Wu (2003), Prestwich et al. (2003) for other examples of the source classification in X-ray color–color diagrams of M83. Instead, L14-139 (S2) stands out for its harder colors, as does L14-237 = MQ1, just as both sources also stand out for variability. L14-139 is located in the sector of the diagram occupied by power-law sources with a hard spectrum (photon index Γ ∼ 1.5). L14-237 is located in the region of the diagram usually populated by X-ray binaries in the high/soft state (steep power law or disk blackbody); this is confirmed by detailed spectral analysis (Soria et al. 2014).

Zoom In Zoom Out Reset image size

The third step is spectral modeling of the stacked spectra from all Chandra/ACIS-S observations, for each source. As expected from our color–color analysis, six sources (L14-63 = S1, L14-183 = S3, L14-186 = S4, L14-256 = S5, L14-310 = S9, and L14-358 = S9) have spectra consistent with thermal plasma emission (mekal model in xspec). In most cases (Table 6, and Figures 7, 8), two temperature components are needed: one component with a temperature kT₁ ≲ 0.3 keV and the other with a temperature kT₂ ∼ 0.5–1.0 keV; only the lower-temperature component is statistically needed for the spectrum of L14-63 (S1).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 6.  Best-fitting Model Parameters and Luminosities of the X-Ray Counterparts

ID	$({N}_{{\rm{H}}})$ a	kT1	$({N}_{1})$ b	kT2	$({N}_{2})$ b	Γ	$({N}_{\mathrm{po}})$ c	${\chi }_{\nu }^{2}$	${F}_{0.3-10}$	${L}_{0.3-10}$
	(1022 cm−2)	(keV)	(10−6 cm−5)	(keV)	(10−6 cm−5)		(10−7)		(10−15 CGS)	(1037 CGS)
L14-63 (S1)	${0.71}_{-0.40}^{+0.22}$	${0.19}_{-0.04}^{+0.08}$	${21.0}_{-18.6}^{+86.3}$	⋯	⋯	⋯	⋯	0.41 (5.7/14)	${1.3}_{-0.2}^{+0.2}$	${10.8}_{-9.4}^{+41.6}$
L14-139 (S2)	${0.39}_{-0.23}^{+0.33}$	⋯	⋯	${0.52}_{-0.21}^{+0.12}$	${1.0}_{-0.6}^{+3.7}$	${1.27}_{-0.30}^{+0.31}$	${6.3}_{-1.9}^{+2.5}$	0.91 (21.8/24)	${6.7}_{-1.1}^{+1.3}$	${2.5}_{-0.6}^{+2.2}$
L14-183 (S3)	${0.13}_{-0.13}^{+0.36}$	$\lt 0.21$	${8.3}_{-8.2}^{+87.0}$	${0.58}_{-0.11}^{+0.09}$	${0.47}_{-0.18}^{+0.54}$	⋯	⋯	1.18 (8.28/7)	${1.1}_{-0.3}^{+0.7}$	${0.9}_{-0.6}^{+5.8}$
L14-186 (S4)	${0.18}_{-0.18}^{+0.55}$	$\lt 0.20$	${25}_{-24}^{+660}$	${0.75}_{-0.16}^{+0.15}$	${0.93}_{-0.36}^{+1.44}$	⋯	⋯	1.07 (22.52/21)	${2.0}_{-0.4}^{+0.8}$	${2.3}_{-1.8}^{+43.5}$
L14-256 (S5)	${0.07}_{-0.07}^{+0.56}$	${0.27}_{-0.08}^{+0.07}$	${0.58}_{-0.23}^{+11.9}$	${0.99}_{-0.21}^{+0.21}$	${0.36}_{-0.12}^{+0.28}$	⋯	⋯	0.82 (9.0/11)	${1.4}_{-0.2}^{+0.2}$	${0.52}_{-0.15}^{+6.0}$
L14-310 (S8)	$\lt 0.31$	${0.26}_{-0.08}^{+0.05}$	${0.89}_{-0.28}^{+0.25}$	${0.80}_{-0.12}^{+0.13}$	${0.87}_{-0.22}^{+0.24}$	⋯	⋯	0.88 (18.48/21)	${3.5}_{-0.3}^{+0.3}$	${1.0}_{-0.1}^{+2.6}$
L14-358 (S9)	${0.05}_{-0.05}^{+0.11}$	$\lt 0.12$	${11.9}_{-10.3}^{+4.3}$	${0.49}_{-0.05}^{+0.07}$	${0.85}_{-0.18}^{+0.27}$	⋯	⋯	0.91 (14.61/16)	${2.7}_{-0.4}^{+0.4}$	${1.4}_{-0.6}^{+1.3}$
L14-237 (MQ1)	See detailed spectral modeling in Soria et al. (2014)

Notes.

^aIntrinsic absorption column density. In addition, a fixed Galactic foreground column density ${N}_{{\rm{H}}}=4\times {10}^{20}$ cm⁻² was included in every model. ^bStandard mekal normalization $N={10}^{-14}/(4\pi {d}^{2})\,\int {n}_{{\rm{e}}}{n}_{{\rm{H}}}{dV}$, where n_e and n_H are the electron and hydrogen densities in the emitting plasma (units of cm⁻³), and d = 4.61 Mpc = $1.42\times {10}^{25}$ cm. ^cStandard power-law normalization in units of photons (keV)⁻¹ cm⁻² s⁻¹ at 1 keV.

Download table as:  ASCIITypeset image

The only source that stands out is L14-139 (S2), as already suspected from its X-ray colors. Spectral analysis confirms that L14-139 has a power-law dominated spectrum, with a photon index Γ = 1.3 ± 0.3, plus a soft excess consistent with thermal emission (Table 6). The soft thermal component contributes for ≈13% of the observed flux, and 28% of the de-absorbed luminosity. The hard nonthermal spectrum supports the interpretation of this source as an X-ray binary. In most of the 12 Chandra observations (in particular, in all the longer ones), the source was located several arcmins away from the aimpoint on the ACIS chip, to the effect that the point-spread function was somewhat degraded; thus, we are unable to determine whether the soft emission is more spatially extended than the hard component. We can only say that both the soft and the hard emission are consistent with a point-like source at that location, with a full width at half maximum ≲50 pc.

In the optical continuum, L14-139 has a point-like counterpart: a bright, red star with apparent magnitude m_555W = 25.69 ± 0.05 mag, m_814W = 21.46 ± 0.05 mag. Assuming the same extinction A_V = 2.66 mag inferred for the ionized nebula (Table 3), we obtain a de-reddened absolute brightness M_V ≈ −6.5 mag, M_I ≈ −8.4 mag, typical of a red supergiant (Figure 2). A somewhat different value of the extinction can be obtained from the best-fitting column density N_H (Table 6), converted to an optical extinction with the relation of Güver & Özel (2009); this corresponds to an extinction ${A}_{V}={2.0}_{-1.0}^{+1.5}$ mag (consistent with the nebular extinction within the errors), and an absolute brightness M_V ≈ −5.8 mag, M_I ≈ −8.0 mag (also consistent with a red supergiant).

In most of our sources, a model estimate of the de-absorbed X-ray luminosity is substantially affected by the large uncertainty on the absorbing column density (Table 6), and the degeneracy between absorption and normalization of the softest thermal plasma component. We used the xspec model cflux to estimate the 90% confidence limits of the absorbed and de-absorbed fluxes reported in Table 6. As an order-of-magnitude estimate, the six thermal-plasma sources in our sample have unabsorbed luminosities ∼10³⁷ erg s⁻¹, with the possible exception of L14-63 (S1), which may be as luminous as ≈10³⁸ erg s⁻¹ after we account for its intrinsic absorption. The X-ray binary candidate L14-139 has an average X-ray luminosity L_X ≈ 3 × 10³⁷ erg s⁻¹, consistent with the luminosity of the low/hard state.

By analogy with our discovery of MQ1 and of other microquasars, we are looking here for exceptionally luminous radio sources, and/or evidence of bipolar radio lobes. In this respect, the most interesting radio sources in our sample are S2 and S7.

S2 is resolved into two sources of similar flux density, S_{5.5 GHz} ≈ 0.2 mJy each, separated by ≈50 pc. Both sources have a similar radio spectral index, α ≈ −0.9 ± 0.6, consistent with optically thin synchrotron emission. The two radio sources roughly correspond to the two peaks of optical line emission identified by Blair et al. (2012) as the two separate SNR candidates B12-96 and B12-98. The hard, nonthermal X-ray source L14-139 is located approximately in between the two peaks of radio emission (Figure 2). The elongated region of optical line emission is also oriented along the same direction as the two radio sources, but extending well beyond them on either side. There are two possible interpretations for this configuration of radio, optical, and X-ray emission: either a chance alignment of multiple SNRs (two of which are also radio bright), with an unrelated X-ray binary in between them; or a microquasar with a core corresponding to the point-like X-ray source, and radio/optical lobes produced by the interaction of the jet with the ISM.

S7 is the brightest source in our sample, with a flux density S_{5.5 GHz} ≈ 1.0 mJy, and one of the brightest compact radio sources in M83 (only slightly fainter than MQ1). Its flux measured from our 2011–2017 ATCA data is consistent with the flux measured by Maddox et al. (2006) from their 1998 VLA observations. Our ATCA observations reveal a spectral index of α = −0.35 ± 0.28: this is consistent both with optically thin synchrotron, and with free–free emission. We have already argued (Section 3.1), based on its optical line ratios, that S7 is associated with a photoionized H ii region, rather than a collisionally ionized bubble. Standard relations between Balmer emission and free–free radio emission in star-forming regions (Caplan & Deharveng 1986; Condon 1992) predict that an Hβ luminosity ≈1.5 × 10³⁸ erg s⁻¹ should correspond to a 5.5 GHz flux density of ≈0.2 mJy at the distance of M83. In reality, the observed radio flux is five times higher. We do not have enough information to say whether S7 contains an additional radio synchrotron source (possibly a young SNR inside the H ii region), or whether the discrepancy is simply due to empirical scatter at the level of individual H ii regions.

The radio source associated with S8 is another interesting case: the peak of the radio emission is displaced by ≈07 south of the bright arc-like optical nebula (Figure 4), near the center point of the optical arc. The radio spectral index is consistent with flat (α ≈ 0), although it could also be steep or inverted (Table 2). A possible interpretation of those two findings is that the optical arc is part of the expanding SNR shell, strongly interacting with the ISM only on its northern side because of a density gradient. Alternatively, the location and radio luminosity (∼10³⁴ erg s⁻¹) of the radio source are consistent with a pulsar wind nebula, filling the central region of the SNR (Gaensler & Slane 2006). However, the associated X-ray source (L14-310) is centered somewhere between the optical and radio peaks (Figure 4), and is dominated by thermal plasma emission (Table 5). Hence, the interpretation of this source remains uncertain.

Finally, we reexamined the radio emission from MQ1 (the brightest non-nuclear radio source in M83), taking advantage of the 2015 and 2017 ATCA observations that took place subsequent to the study of Soria et al. (2014). From our analysis of the combined ATCA dataset, we obtain integrated flux densities S_{5.5 GHz} ≈ 1.8 mJy and S_{9 GHz} ≈ 1.0 mJy, respectively. At the distance of M83, this corresponds to a radio luminosity of ≈2.5 × 10³⁵ erg s⁻¹. This is much larger than the integrated 5.5 GHz radio luminosity of ≈5 × 10³³ erg s⁻¹ for the most powerful microquasar bubble in the Milky Way, SS 433/W50 (Dubner et al. 1998); it is also higher than the most radio-luminous Milky Way SNR, Cassiopeia A, with a 5.5 GHz luminosity of ≈5.5 × 10³⁴ erg s⁻¹ (Arias et al. 2018).

According to Chomiuk & Wilcots (2009a), the radio luminosity function of SNRs is a power law and the number of SNRs is proportional to the star formation rate. If that is the case, we expect the maximum radio luminosity of a SNR to be higher in M83 (with a star formation rate ≈3–5 M_⊙ yr⁻¹: Boissier et al. 2005) than in the Milky Way, with its more modest star formation rate (≈1.5–2 M_⊙ yr⁻¹: Licquia & Newman 2015). As a result, we expect a few SNRs in M83 with a radio luminosity comparable to MQ1 even though none are observed in the Galaxy. The radio luminosity function of microquasar bubbles (powered by compact objects accreting in the supercritical regime) is not known yet, because of the small number of such sources identified so far; however, we do know of other bubbles with 5 GHz luminosities ∼10³⁵ erg s⁻¹ (e.g., NGC 7793-S26: Soria et al. 2010; IC 342 X-1: Cseh et al. 2012). Although the radio luminosity and energy content of the most powerful radio SNRs and microquasar bubbles may be of the same order of magnitude, the two classes of sources differ in age and size: theory predicts (Sarbadhicary et al. 2017) that the most luminous radio SNRs have ages ∼10²–10³ yr and radii ≲a few pc (e.g., the bright SNR in NGC 4449: Chomiuk & Wilcots 2009b; Bietenholz et al. 2010), while accretion-powered bubbles can have ages ≳10⁵ yr (timescale of supercritical mass transfer from the donor star) and sizes ∼100 pc (Pakull & Mirioni 2002; Pakull et al. 2006, 2010).