SHOCKS, SEYFERTS, AND THE SUPERNOVA REMNANT CONNECTION: A CHANDRA OBSERVATION OF THE CIRCINUS GALAXY

The reduced and merged X-ray image we obtained is presented in Figure 1. Overlaid over the X-ray image are the 13 cm radio contours and the regions we used to extract the surface brightness profiles shown in Figures 3 and 4. This image shows several extended structures with different orientations (see also Curran et al. 2008). Our proposed geometry is detailed in Figure 2, which was drawn over the combined X-ray/radio/optical image, to reflect the scale, reach, and orientation of each component as accurately as possible. Our figure is similar to the picture proposed by Elmouttie et al. (1998a), but there are some differences, particularly in the interpretation of the different structures. The AGN and circumnuclear star-forming ring are at the center of the galaxy, the ring facing toward us. The galaxy's disk extends in the NE–SW direction, with an orientation of ∼60° relative to the plane of the sky (Jones et al. 1999; Curran et al. 2008), and shows some X-ray emission, though it is fainter than the other structures. The NW lobe is unobscured, on the near side of the disk, and the SE lobe is partly obscured, showing a clear dip in X-ray emission in the areas covered by the disk. Both lobes appear to be edge-brightened, though the NW lobe is much brighter and somewhat smaller. The orientation of the lobes is roughly perpendicular to that of the disk. There is a dip in X-ray luminosity on both lobes behind the bright edges and a rise toward the center of both, suggesting that there is further structure within the lobes. We extracted surface brightness profiles for both lobes, to quantitatively verify the correspondence between the X-ray and radio structures, as well as to confirm the edge-brightening in both cases. The results are illustrated in Figures 3 and 4. Near the AGN, coinciding with the base of the NW lobe, quite strong X-ray emission coincides with a region of highly ionized gas that is believed to be an ionization cone (Marconi et al. 1994; Elmouttie et al. 1998b; Wilson et al. 2000; Smith & Wilson 2001). The X-ray emission from the galaxy's halo is very faint and not directly discernible from the images shown.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The radio contours show the overlap between radio and X-ray emission (for a detailed analysis of the radio maps see Elmouttie et al. 1998a). There is faint radio emission in the NE–SW direction, coincident with the disk of the galaxy, while the radio lobes extend perpendicularly to it. The pattern of radio emission in the lobes matches the X-ray structures perfectly: the edge-brightened emission in the lobes, the dip in luminosity behind the edges, and the rise toward the center of the lobes are spatially coincident across both wavelengths, except in a few points. We tested different orientations and distances when extracting the radial profiles in Figures 3 and 4. The coincidence between the radio and X-ray edge peaks persists across most of both lobes, though it is clearer in some areas. We chose regions that covered all the appropriate structures: shells, lobe interior, and hotspots, while trying to avoid the main emission from the ionization cone, which has no radio counterpart, and the edges of the CCD.

The differences in the morphology of the emission between the two lobes can only be partly explained by orientation effects. The E lobe is larger and does show X-ray emission outside the confines of the radio lobe edges in the regions farthest from the AGN. The W lobe is smaller, more spherical, and its emission seems to be confined within what is seen in the radio maps. However, the edge of the CCD is just outside the edge of the W lobe in our long observations, and it is possible that we may be missing some of the outermost edge in this direction, in which case both lobes could exhibit a tip of X-ray emission outside the radio lobe. It is possible that the outflow may have encountered very different environments on its way to its current position, which could explain the asymmetry between the W and E lobes.

The W structure also shows the base of the ionization cone detected by Marconi et al. (1994), while no counterpart is visible in the E, most likely due to its being obscured by the galaxy's disk (see Figures 2 and 5). The Chandra images show three distinct tails, propagating from the AGN in the W direction. The two brighter tails to the north have [O iii] and Hα counterparts (see the right panel of Figure 5 and Smith & Wilson 2001; Wilson et al. 2000; Elmouttie et al. 1998b), while the southernmost edge only shows up in Hα + [N ii]. This southernmost edge could be obscured, as suggested by Elmouttie et al., or could be produced by a different mechanism. While the top tail is clearly one of the edges of the ionization cone, it is unclear whether the other edge is given by the central or the southernmost tail (Elmouttie et al. 1998b).

Zoom In Zoom Out Reset image size

As we discussed in Section 1, the nucleus of Circinus and its immediate surroundings have been studied in detail in the past. The objective of our analysis is to study the larger-scale X-ray emission associated with the radio lobes of Circinus.

We extracted X-ray spectra of the radio lobes from regions excluding the AGN and circumnuclear emission, as well as the (presumably) photoionized plume in the W lobe, to avoid contamination. Our W region also avoids most of the area covered by the part of the disk that lies behind the lobe, thus minimizing any possible contamination from it in the resulting spectrum. The spectra were binned to 15 counts per bin to allow Gaussian statistics to be used. We also binned the spectra to 20 counts per bin to check the robustness of the fits.

We used regions that cover most of the radio lobes, excluding the circumnuclear emission and any point sources, after verifying that the spectrum of the gas inside the lobes does not differ from those of the gas in the shells, presumably due to the low surface brightness emission from the shells contributing in these regions. We worked with a number of different source and background regions to minimize contamination. The spectra of the W lobe are fairly consistent regardless of our choice of source and background, but the E lobe was problematic in this respect: we found that the overall shape of the spectrum was independent of our choice of regions, but not entirely consistent with any simple single- or double-component model. As discussed below, this anomalous shape may be caused by a higher contribution of star-related emission from the galaxy's disk. The statistics on the E lobe are also poorer than those of the W counterpart, most likely due to the higher obscuration, making it difficult to mask out bright regions and still obtain a good fit.

We fitted to our spectra a model consisting of local Galactic absorption (wabs) and emission from a collisionally ionized gas (apec) (see Figures 6 and 7). We approached the analysis by fitting the model both to the individual spectra together and to the co-added spectrum (generated with the CIAO tool combine_spectra). Only the former approach is shown in the figures, but the results are consistent for both methods. Given that the abundance is not well constrained (especially for the E lobe, where we have problems with both multi-phase gas and covering absorbers), and if left free during the fitting the values tend to become implausibly low, we assumed a fixed abundance of 0.15, compatible with the lower limit set by the abundances in the intergalactic medium (see, e.g., Danforth et al. 2006). Abundances higher than 0.2 do not affect the estimated values for the temperature, but the fit statistics are much worse. See the last paragraph of Section 4.2 for a discussion on the implications of a different value of Z.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We found that the emission in the W lobe is very well fitted with this single-temperature model, although there are some residuals, especially at high energies, most likely caused by contamination from the AGN point-spread function. We obtained a best-fitting temperature of 0.74^+0.06_−0.05 keV and a reduced χ² of 1.09. See Figure 6 and Table 2 for details.

The E lobe, however, proved to be more problematic. This area is partially covered by the galaxy's disk, and the spectrum (see Figure 7) is not successfully modeled by a single-temperature model (reduced χ² ∼ 1.5). Adding a power law or a second thermal component does improve the fit, albeit slightly (reduced χ² ∼ 1.2), and most of the residuals persist. Adding an extra absorbing column does not improve the fit. The poor fit could be caused by emission from an unresolved X-ray binary population and massive stars in the disk, or unavoidable contamination from the nearby, strong AGN.

XSPEC determined a best-fitting temperature for the E lobe of kT = 1.6^+0.6_−0.2 keV, but it is not well constrained. A detailed analysis shows local minima when the model assumes kT = 0.8 and kT = 1.8 keV. Thus, we used these values as a lower and upper limit, respectively, for the shell calculations displayed in Table 2.

The disk over the E lobe is problematic because it may be not only introducing some contamination but also masking an E counterpart to the W ionization cone, which could also be contributing to our spectra. Given the different morphologies of the two lobes, it is also possible that the temperature structure may be different for the E structure. We will discuss the nature of the emission in the lobes and its structure in the following sections.

We also carried out a fit over the regions corresponding to the disk (NE–SW) and found that a single-temperature model works remarkably well, with kT = 0.69^+0.19_−0.17 keV. This temperature is very similar to the one we found for the W shell, but not so well constrained (mainly due to a higher fraction of contamination from the AGN).

To test whether there are any changes in the nature of the emission in the brighter areas within the lobes, we extracted spectra from these regions. A fit to a single power-law model gave very poor results, since the thermal emission from the shells is still present. We then fitted to the spectra a model equal to the one we used for each shell plus either a power law or a second thermal component. The addition of a second component improves the fit; however, the poor statistics (the brightest individual spectrum has just ∼100 counts) do not allow us to distinguish between a thermal and a non-thermal model in this case. Although the fit is slightly better with the non-thermal model, the difference is not statistically significant. We will discuss the nature of this emission in Section 4.3.

Given that the edges of the ionization cone coincide with the base of the W shell almost exactly (see Figure 5 and the discussion at the end of Section 3.1), it is possible that some of the line emission studied by Marconi et al. (1994) may be caused by collisional ionization rather than photoionization. However, without [O iii], [N ii], and Hα maps with a large field of view and good resolution to measure the intensities and ratios, it is hard to tell. Given the intensity of the AGN and the light from the starburst ring, photoionization must indeed play a role in the emission we see in this region. However, it is extremely unlikely that photoionization from the central AGN could explain the edge-brightened, large-scale emission we are observing farther out. We calculated the extent of photoionization caused by the AGN following the methods of Wang et al. (2009) and the models from Kallman & McCray (1982). Given that

$$\begin{equation} \centering \mathcal {\xi }=\frac{L}{n R^{2}}, \end{equation} \tag{ 1 }$$

where ξ is the ionizing parameter, which we assume to be between 10 and 100 (reasonable values for AGN outflows, as demonstrated by Ogle et al. 2003; see also Wang et al. 2009); n ∼ n_e is the particle density (which we assume to be between 10⁻² and 10⁻¹ cm⁻³); R is the distance (in cm); and L is the unabsorbed X-ray luminosity of the AGN required to produce the observed ionization factor (we assumed L = 1.6 × 10⁴² erg s⁻¹, after de Rosa et al. 2012). In the case of Circinus, this implies that the AGN would be able to weakly ionize (ξ ∼ 10) gas up to ∼700 pc away, but strong ionization (ξ ∼ 100) would only be produced within the inner ∼200 pc. Given the relatively low luminosity of the AGN (Yang et al. 2009) and the dense environment surrounding it, this is consistent with what is observed for the ionization cone but shows that the edge-brightened emission in the shells cannot be accounted for by photoionization. Given that the X-ray emission at the edges of the radio lobes extends up to 2–3 kpc in projection (for the W and E shell, respectively), even the possibility that the AGN was more powerful 10³–10⁴ yr ago makes photoionization unrealistic to explain the level of emission we are detecting at these distances, requiring densities an order of magnitude smaller than those we measure in these regions to achieve weak photoionization.

While there are no high-resolution, large-scale [O iii] or Hα maps that cover all of the ionization cone, the existing images (Elmouttie et al. 1998b; Wilson et al. 2000) do show other regions of the galaxy that coincide with part of the lobes, and there seems to be no enhancement in emission in these regions, matching the radio and X-ray data, while this enhancement is observed in the ionization cone. From a morphological perspective, since the base of the W shell coincides with the ionization cone (Marconi et al. 1994), and given the results above, we can infer that there is a transition between predominantly photoionized and shock-ionized gas at these distances, which could be further investigated with large-scale [O iii] and Hα maps and more X-ray observations.

For consistency, we extracted a spectrum of the two northernmost tails of the ionization cone (see the left panel of Figure 5) in our longer exposure, covering distances of 150–500 pc from the AGN. We fitted the resulting spectrum to a photemis model (part of WARMABS, an analytical XSPEC model implemented from XSTAR) and foreground Galactic absorption. We assumed a gas density of 10⁴ cm⁻³ and an ionizing power law with Γ ∼ 1.6 (de Rosa et al. 2012). We found that photoionized emission alone does not fully account for the spectral shape, yielding a reduced χ² ∼ 2.5. This may be due to the less-than-ideal background subtraction (a local background region that can compensate for the AGN contamination, without including the readout streak or emission from the star-forming ring, will still include thermal emission from the galaxy's disk) or to a more complex spectral structure derived from the large volume covered by our region (a requirement to obtain the necessary statistics for the fit). By testing the addition of different continua to improve the fit, we observed that the ionization parameter remained stable, at values of $\log (\mathcal {\xi }) \sim 1.7$–2.0, which supports our assumptions in the previous paragraph. We also found very low abundance constraints (∼10⁻²) on O, Mg, and Si and higher values (∼0.1–0.3) for Fe and Ne. We were unable to constrain the abundances for the other elements, but increasing them over our initial assumption (0.15☉) increases the χ².

We also studied the likelihood of detecting an identical structure in the E lobe, behind the disk of the galaxy. We estimated the count rate we would expect in our image from this structure with the PIMMS tool, assuming it to have the same unabsorbed flux as our W ionization cone spectrum (F_{0.3–10 keV, unabs} = 1.5 × 10⁻¹² erg cm⁻² s⁻¹) and a foreground absorbing column from the disk comparable to that of the Cen A dust lane (N_H ∼ 1.5 × 10²² cm⁻²). We found that we would only obtain ∼300 counts from this structure in our image, below the emission level from the disk of the galaxy itself, and thus we would not be able to detect it. Although the obscuration from the disk might be lower, it is also possible that the E ionization cone, if present, may be fainter than its W counterpart, if the environment through which the radiation must pass is denser. The possibility of an ionization cone in the E lobe, therefore, cannot be discarded.

The match between the X-ray and radio emission is striking in Circinus (see Figures 3 and 4). Unlike in other, more distant systems, where we lack the spatial resolution to resolve these structures in such detail, the proximity of Circinus allows us to observe that the emission from the shells coincides extremely well in radio and X-rays. In all the radio frequencies there is a dip in luminosity behind the shell and a rise in emission toward the center of the radio lobes. Given the edge-brightened nature of the emission, the abrupt jump in luminosity, and the spectral properties we derived for the gas, we consider that the most likely scenario is that we are observing shells of shocked gas around the radio lobes of Circinus. We will discuss below whether this shock may be starburst or jet driven. The dip and rise in luminosity behind the shells suggests further complexity in their structure, since we would expect the lobes to be uniformly filled with gas. We will discuss the nature of this emission for both the starburst and jet scenarios.

The fact that the X-ray emission does not appear to be outside the radio lobes (except at the tip of the E lobe, as discussed in Section 3.1) suggests that the same shock that is causing the X-ray emission in the shells (which, as detailed in Section 3.2, is thermal in origin) is also giving rise to enhanced synchrotron emission at radio frequencies. The apparent thickness of the shells, measured from the X-ray images, is ∼300 pc and appears constant across the entire structure.

To test the shock hypothesis, we assumed two ellipsoidal shells (the geometry that most closely follows the shells as depicted in Figure 2, with the smallest departure from spherical symmetry) with the two minor axes equal. The two-dimensional regions we used in our images have eccentricities of 0.74 and 0.52 for the E and W shells, respectively. The E structure is larger, with R_{int, maj} = 1.0 kpc, R_{ext, maj} = 1.4 kpc, R_{int, min} = 0.7 kpc, R_{ext, min} = 0.9 kpc. The W shell has R_{int, maj} = 0.8 kpc, R_{ext, maj} = 1.1 kpc, R_{int, min} = 0.7 kpc, R_{ext, min} = 0.9 kpc. These dimensions correspond to volumes of 7.8 × 10⁶⁴ cm³ and 6.1 × 10⁶⁴ cm³ for the E and W shells, respectively.

We derived the electron densities for both shells from the normalization of the thermal model (see Table 2), assuming a uniform filling factor. From these densities it is also possible to calculate the total mass of gas in the shells, the pressure, the total thermal energy, and the work available from the gas. We found the densities to be slightly higher than those we found for Mrk 6, but the total energy involved is lower by roughly two orders of magnitude, which is to be expected given the much lower radio power observed in Circinus.

In order to test our hypothesis of a shock scenario, we attempted to extract spectra of the gas immediately outside the lobes but were unable to fit a model to the data, due to the poor statistics. We then generated radial profiles for the longest observation from concentric annuli (r_min = 132 arcsec, r_max = 305 arcsec, δr = 8.7 arcsec), excluding the emission from the AGN, the lobes, and the galactic disk, and a suitable background. The surface brightness of the gas in the halo is extremely low, and it barely shows up in radial profiles (see Figure 8). The drop in luminosity at R ∼ 440 pixels (∼4.2 kpc) indicates the possible extent of the halo, but the errors are too large to consider this anything more than a qualitative indication.

Zoom In Zoom Out Reset image size

To better constrain the surface brightness, we estimated the count statistics for a circular region centered in the AGN in the longest observation, excluding the regions coincident with the lobes, the disk of the galaxy, any point sources, the readout streak, and any CCD gaps or adjacent chips in the process. We used two different radii for this circular region, r = 246 and 305 arcsec, to better assess the effect of the background, which was extracted from all the remaining S3 chip. We detected the halo at a 3.8σ level in the 0.3–7 keV energy range (1439 ± 382 and 884 ± 239 counts for the larger and smaller source regions, respectively, corresponding in both cases to a surface brightness of 0.011 ± 0.003 counts pixel⁻²) and at a 5.5σ level in the 0.5–2 keV range (1259 ± 245 and 852 ± 154 counts for the larger and smaller source regions, respectively, corresponding in both cases to a surface brightness of 0.010 ± 0.002 counts pixel⁻²). Since the orientation of the chips in the other two observations is different, it is not possible to apply the same regions to them; instead, we extrapolated the statistics from the first observation to the total observed time. We thus obtained a 1σ range for the surface brightness of the halo, in the 0.3–7 keV energy band, of 0.01–0.02 counts pixel⁻² (0.04–0.08 counts arcsec⁻²).

The constraints on the surface brightness allowed us to test a thermal model with different temperatures, assuming a suitable extraction region outside the radio lobes (a spherical shell centered in the AGN, with R_int = 3.68 kpc, R_ext = 5.24 kpc). Given the range of temperatures found in the halos of similar galaxies (see, e.g., Tüllmann et al. 2006b; Yamasaki et al. 2009), we derived the electron densities for kT = 0.1, 0.2, and 0.3 keV.

We used our low/high values for the surface brightness and the halo shell region to obtain estimated count rates for each temperature. We then derived the apec normalizations required to obtain these count rates and derived the electron densities and pressures from those normalizations. The results are displayed in Table 3. We find that the electron densities we derive are compatible with those estimated by Elmouttie et al. (1998a), based on observations of other nearby spiral galaxies. We also estimated the cooling time for the gas, obtaining results of (0.3–3.6) × 10⁸ yr for kT = 0.2–0.3 keV and as low as 6.2 Myr for kT = 0.1 keV. These times are very short, since line cooling is very important at these low temperatures, suggesting that the hot gas into which the lobes are expanding must have a comparatively recent origin, possibly in a starburst and/or supernova wind.

To verify our assumptions on the properties of the gas in the halo, we used the relations of Tüllmann et al. (2006a), which give the correlation between the X-ray luminosity of the halo and other parameters for a sample of star-forming spiral galaxies. Some of their correlations indicate a very strong dependence between the soft X-ray luminosity of the halo and parameters such as the FIR/Hα SFR and between the overall galactic X-ray luminosity and the mass of dust or H i in the galaxy.

Given how obscured and nearby Circinus is, most previous studies have focused just on the central star-forming ring or the supernova remnant (SNR), so that some of the parameters we could use to derive the luminosity of the halo are not known, but we were able to use the values from Curran et al. (2008), Jones et al. (1999), Elmouttie et al. (1998b), and For et al. (2012) (L_{B, bol} ∼ 2.9 × 10⁴³ erg s⁻¹, L_{UV, bol} ∼ 9.1 × 10⁴² erg s⁻¹, M_dust ∼ 10⁶ M_☉, M_stars ∼ 10¹¹ M_☉, SFR_FIR ∼ 3–8 M_☉ yr⁻¹) with Tüllmann et al.'s correlations. This allowed us to constrain the soft (0.3–2 keV) luminosity of the halo of Circinus to (1.6–6.3) × 10³⁹ erg s⁻¹ from the correlations of Tüllmann et al. Given that the extraction region we used in our analysis does not cover the full halo of the galaxy, we find that our results are consistent with what is expected from these correlations and the observations (see Table 3). The luminosity we thus constrained is also compatible with the count rates we estimated from our images.

Comparing the electron densities we derive for the shells and the external gas, we see that density ratios of ∼1.1–3 can be derived from kT_ext = 0.2 or kT_ext = 0.3 keV, and smaller values (0.6–0.9) for kT_ext = 0.1 keV. The small density contrast makes the latter temperature less consistent with our shock interpretation, given that the density ratio ρ_out/ρ_shell tends to 4 in the case of a strong shock. Based on observational results (see, e.g., Hummel et al. 1991; Beck et al. 1994), the electron densities we derived from kT = 0.2 and kT = 0.3 keV are compatible with those seen in the halos of spiral galaxies.

Although the fact that we do not directly detect the external gas is still a limitation, since it would allow us to rule out a slow shock into a cold, dense medium, or pressure conditions where a shock is not possible, the morphology of the X-ray emission, the observed properties, and the agreement of the extrapolated values for the external gas with what we would expect in our assumed scenario indicate that we are indeed observing a strong shock being driven into the ISM of Circinus.

Using the observed parameters, we can calculate the Mach number for the shock, as follows (Landau & Lifshitz):

$$\begin{equation} \centering \mathcal {M}=\sqrt{\frac{4(\Gamma +1)(T_{{\rm shell}}/T_{{\rm out}})+(\Gamma -1)}{2\Gamma }} , \end{equation} \tag{ 2 }$$

where Γ is the polytropic index of the gas, assumed to be 5/3. We find Mach numbers $\mathcal {M}\sim 2.7$–3.6 for the W shell (up to $\mathcal {M}=5.1$ if the external temperature is closer to kT = 0.1 keV) and $\mathcal {M}\sim 2.8$–5.3 for the E shell (up to $\mathcal {M}=7.6$ for kT = 0.1 keV). The sound speed is ∼264⁺⁵⁹₋₇₈ km s⁻¹, and the velocity of the shells is 915⁺⁴⁵₋₃₅ km s⁻¹ for the W shell and 950⁺⁴⁷⁰₋₁₃₀ km s⁻¹ for the E shell. The errors on these values reflect the temperature range of both the shell and the external gas; the uncertainty on the E shell is very large due to the range of possible shell gas temperatures, as indicated in Table 2. It is clear, in any case, that the velocity of the shells must be supersonic and close to 1000 km s⁻¹, in agreement with what is observed in similar systems (e.g., Mrk 6, NGC 3801: Mingo et al. 2011; Croston et al. 2007)

If we assume that the shells have been expanding with a constant velocity, equal to the one we have derived for the W structure, we can constrain their age to ∼10⁶ yr. The total energy (enthalpy of the cavity and kinetic energy from the shock) involved in creating both shells is ∼2 × 10⁵⁵ erg, the equivalent of 10⁴ supernova explosions with individual energy 10⁵¹ erg, assuming a 100% efficiency.

While the spectral fits of the ionization cone (see Section 4.1) seem to support our initial assumption of Z = 0.15☉, this value is still poorly constrained. It is very likely that the areas of the galaxy covered by our regions contain a multi-phase gas and a range of abundances, but these cannot be studied in detail without deeper observations. The gas density we derive from our models depends on both the abundance and the emission measure, so that, assuming that there are no abrupt variations in the abundance between the shells and the regions immediately outside them, for Z = 0.30☉ the electron density in the shells would be ∼25% lower, while that of the external gas would go down by ∼20%. The ratio between both densities, however, only depends weakly on Z, so even if our assumptions on the abundance were wrong and the densities were smaller, our conclusions would stand. The Mach number and the shock velocity would not change, but the total energy involved in the process would be slightly smaller.

Given that Circinus is a spiral galaxy, with active star formation, it is possible that the bubbles that we are observing perpendicular to the plane of the galaxy may be caused by a starburst or supernova-driven wind. As pointed out by Elmouttie et al. (1998a), however, the particle component within starburst-driven "bubbles" is expected to be a mixture of thermal and non-thermal electrons, with the thermal material causing depolarization. The W "bubble" in Circinus does not show a high degree of depolarization. In the E "bubble" the higher depolarization is most likely caused by the gas in the disk, which is obscuring part of the structure (for full details on the radio polarization of Circinus see Elmouttie et al. 1998a).

As we saw in Section 4.2, the total energy involved in the creation of both "bubbles" in Circinus is ∼2 × 10⁵⁵ erg. This value is similar to the one Matsushita et al. (2005) calculated for the superbubble observed in M82 ((0.5–2) × 10⁵⁵ erg), one of the most powerful examples of superbubbles. The efficiency with which supernova explosions transfer energy to the ISM is quite low (3%–20%, depending on the density of the ISM and time lapse between subsequent explosions). Elmouttie et al. (1998a) inferred a supernova rate of 0.02 yr⁻¹ from their radio observations. To check this value, we used the far-infrared data from Genzel et al. (1998) (L_FIR = 7 × 10⁹ L_☉) and the correlations from Condon (1992) to estimate a massive SFR (${\rm SFR}_{M\, \ge \, 5\ M_{\odot }} \sim 0.64$ M_☉ yr⁻¹) and derived a supernova rate of ∼0.026 yr⁻¹. With either of these values, the possibility of a supernova-driven wind is energetically viable in Circinus. However, the enhanced emission at the center of the "bubbles" is not easy to explain, morphologically, with this model. Multiple generations of "bubbles," within the 10⁶ yr age estimated for the larger structure, seem unlikely.

Some consideration must be given to how the radio "bubbles" are related to the structures closer to the AGN. Elmouttie et al. (1998b) suggest that there is an ongoing outflow of gas in the W ionization cone (with a total kinetic energy of 5 × 10⁵³ erg), which could be powered by the AGN or a starburst wind. They suggest that the similar sizes of the Hα central ring and the base of the central radio emission indicate that these structures are related, and that an ongoing starburst wind is the cause of both the inner outflow and the enhanced radio emission at the center of the "bubbles." Emission line ratios in the cone, however, suggest that the emission in the cone may be caused by AGN-driven photoionization. For Circinus, Elmouttie et al. also point out that the morphologies of the ionization cone and the radio "plume" are quite different, indicating that these structures may not have the same origin. As suggested by Schulz (1988) for NGC 4151, it is possible that, as the shock that created the larger structures in Circinus made its way through the halo gas, it created a low-density region through which ionizing radiation, from the AGN or a starburst wind, could escape, giving rise to the cone. It is thus entirely possible that there is an ongoing starburst wind in the central regions of Circinus, feeding material into the ionization cone, but that this is not powerful enough to be the cause behind the large-scale emission in the shells and at the center of the "bubbles."

The strongest argument against the starburst-driven scenario as an explanation for the large-scale emission, however, comes from the variations in the radio spectral index we observe across the "bubbles." Variations in the spectral index are a sign of a change in the electron population; a lower value of α may indicate a region where particle acceleration is occurring. An increasing value of α indicates radiative losses and an aging electron population. We calculated the fluxes for the shells, the areas immediately behind the "bubbles," and the regions with enhanced radio and X-ray emission inside them (which we labeled as hotspots), as illustrated in Figure 9. We convolved the 6 cm, 13 cm, and 20 cm ATCA radio maps so that the spatial resolution was the same in all of them (20.5 × 20.5 arcsec), thus ensuring that we are measuring the emission from the same structures in all cases. To best estimate the extent of the rms noise, we chose a background region over the galaxy's disk, where there are flux variations, rather than a region off-source. We then derived spectral index values for each structure from the 6 cm/13 cm and the 13 cm/20 cm flux measurements. The results are illustrated in Table 4.

Zoom In Zoom Out Reset image size

Although our results are limited by the very low flux levels and by the fact that at 6 cm the structures we are observing are slightly different, it is clear that the spectral index does vary between these three regions, and consistently so for both the western and eastern structures, in the sense that there is flattening (alpha decreases) in both hotspots and steepening (alpha increases) in both shells, even if the individual values are slightly different (this may be attributed to low spatial resolution, disk obscuration of the E lobe, and/or different spectral indices in the pre-shock electrons for both shells). These results are also consistent with the spectral index map from Elmouttie et al. (1998a). As stated above, this symmetrical flattening of the spectral index at the center of the "bubbles" is most easily explained if there is ongoing particle acceleration. The simplest scenario to explain this is that these regions are points where a radio jet interacts with the ISM and thus the equivalents of the hotspots observed in more powerful systems (see Section 4.4 for more details). It is natural to assume that this same jet is the one that provided the energy injection that caused the shock we are observing. This is further supported by the morphology of the emission, which is coincident in X-rays and radio and symmetrical for both lobes, roughly equidistant from the nucleus of the galaxy.

In light of these results, we find that the scenario that best describes the data is that which is typically assumed for radio galaxies (see, e.g., Scheuer 1974; Kaiser & Alexander 1997). In this scenario the jet emanating from the central AGN makes its way into the ISM of the host galaxy, transferring part of its energy to the surrounding gas at the termination points (or hotspots). This energy transfer heats up the ISM gas, creating a supersonic bow shock. The material that has flowed up the jet is observed in the form of radio lobes.

Hotspots in X-rays are characterized by a non-thermal power-law spectrum, though they are generally very faint, given that synchrotron losses are very high at these energies. As we stated in Section 3.2, the statistics of the spectra of the hotspot regions are quite poor, and they do not allow us to distinguish between thermal and non-thermal emission. If the emission were non-thermal, it would be unlikely to have its origin in widespread inverse-Compton scattering, since the contribution of such a component for low-power radio galaxies is negligible. We estimate the IC luminosity density at 1 keV to be ∼6 × 10³⁴ erg s⁻¹ keV⁻¹ for the case of Circinus (equivalent to (2–5) × 10⁻⁵ counts s⁻¹ in Chandra ACIS-S, depending on the IC power-law photon index), using the assumptions on the B field detailed in Section 4.5. Moreover, if there was a detectable contribution from inverse Compton, it would appear uniformly across both lobes. We calculated the spectral index between 6 cm and 1 keV for the W hotspot and found it to be ∼1. This value is in the range of radio/X-ray ratios seen in FR I jets (see, e.g., Harwood & Hardcastle 2012), suggesting that synchrotron emission could account for at least part of the X-ray emission we are observing.

Although our images do not show an active jet (at radio, optical, or X-ray frequencies), we know that this does not rule out the possibility that there is one, from what is observed in powerful FR II radio galaxies, where the jets are rarely seen, yet the presence of hotspots clearly indicates their presence (e.g., Kharb et al. 2008). The jet's rest-frame emissivity may be very low in Circinus. Moreover, the orientation of the lobes suggests that the jet should be at a large angle to the line of sight, so that relativistic beaming may be reducing the visibility of a jet that would intrinsically be visible. Doppler-dimmed jets and hotspots have been found in other Seyfert galaxies with low radio luminosities (see, e.g., Kukula et al. 1999), illustrating how the interactions of the jet with different, complex environments can change the observed radio properties of these sources. This might also explain why the X-ray emission in the shells extends farther than the radio along the (presumed) direction of propagation of the radio jet. This geometry is expected from some jet propagation simulations (see, e.g., Sutherland & Bicknell 2007).

In this context, to further verify our shock scenario, we calculated the minimum internal pressure of the radio lobes under equipartition conditions from the radio data, fitting a broken power-law electron energy spectrum with p = 2 at low energies, steepening to p = 3 at the electron energy that gives the best fit to the data. We assumed ellipsoidal geometry, no protons, and γ_min = 10. We obtained a value of P ∼ (6–9) × 10⁻¹⁴ Pa, consistent with the results obtained by Elmouttie et al. (1998a), and roughly two orders of magnitude lower than the pressures we derive from our X-ray data. This departure from minimum energy is often found in FR I radio galaxies (e.g., Morganti et al. 1988); we also found this effect in NGC 3801, Mrk 6, and Cen A. These results imply that there is some additional contribution to the internal pressure, caused by a large deviation from equipartition conditions, a high fraction of non-radiating particles, such as thermal or relativistic protons originating from the interaction of the galaxy's gas with the jet, or a low filling factor. Of these, the second explanation is the most plausible one (see Hardcastle et al. 2007; Croston et al. 2008a).

From the age of the shells and the total energy we can infer the kinetic luminosity of the (invisible) jet, which is ∼10⁴¹ erg s⁻¹. Comparing these results to those we obtained for Mrk 6 (Mingo et al. 2011), NGC 3801 (Croston et al. 2007), or Cen A (Kraft et al. 2003) shows that these parameters scale down with the radio power of the parent AGN (see Table 5).

The fact that the X-ray emission in the shells is thermal is expected, if our shock scenario is correct. As pointed out by Croston et al. (2009), it is likely that fast shocks are required to produce in situ particle acceleration and X-ray synchrotron emission in the shocked shells, as observed in Cen A. The velocities we derive, smaller than those found in Cen A, and the unequivocally thermal nature of the emission we observe suggest that the case of Circinus is more similar to that of NGC 3801 or Mrk 6.

The fact that the radio emission and X-ray emission are spatially coincident in the shells of Circinus resembles the morphology that is observed in some SNRs and poses the question of how similar the underlying physical processes are. Given that we do not observe non-thermal emission in the X-rays in the shells of Circinus, but we do see edge-brightening in radio, we investigate whether this enhancement of the radio synchrotron emission in Circinus can be fully accounted for through a shock-driven enhancement of the B field strength, similar to what has been observed in ongoing cluster mergers (e.g., Markevitch et al. 2005; van Weeren et al. 2010). The configuration of the magnetic field in Circinus is not very clear. The argument of Elmouttie et al. (1998a), stating that the dominant component of the magnetic field in the shells of Circinus seems to be predominantly radial (perpendicular to the shock front, parallel to its direction of propagation), is based on polarization maps and is not very strong. Other systems such as NGC 1068 (Wilson & Ulvestad 1987; Simpson et al. 2002) and NGC 3079 (Duric & Seaquist 1988) indicate that fields tangential to the shock front (perpendicular to the direction of propagation of the shock) are more often found in radio galaxies. Both radial and tangential B field configurations can be found on SNRs (see, e.g., Milne 1987), but only the component that is tangential to the shock front is compressed by it. Compressed magnetic fields have been found around the leading edges of FR I radio galaxies (Guidetti et al. 2011, 2012) from rotation measure analysis, suggesting that, given the right conditions (a component to the magnetic field tangential to the front and a strong shock), this phenomenon may be common to many low-power systems.

To investigate whether this shock-driven B enhancement could be enough to account for the enhancement in radio emission, we must calculate whether the jump in emissivity (of a factor of at least 40) observed between the shells and the gas just outside them can be explained by this mechanism. The synchrotron emissivity of a power-law distribution of electrons is given by

$$\begin{equation} \centering J(\nu)=C N_{0} \nu ^{-{(p-1)}/{2}} B^{({p+1})/{2}}, \end{equation} \tag{ 3 }$$

where C can be assumed constant, N₀ is the normalization of the relativistic electron spectrum, and (p − 1)/2 = α. If we assume that B is tangential to the shock front, applying the Rankine–Hugoniot conditions for a strong shock results in B_shells/B_ext∝ρ_shells/ρ_ext, which, substituting the values from Tables 2 and 3, results in B_shells ∼ 1.1–3B_ext. In Equation (3) N₀ is proportional to the electron density, but given that the electron density depends on the energy density, which scales with the temperature in the shock, we have that N_{0, shells}/N_{0, ext}∝(ρ_shells/ρ_ext)(T_shells/T_ext)^{p − 1}. Substituting the values from Tables 2, 3, and 4, we can calculate the emissivity jump. Taking only the values for the W shell, which are better constrained, we see that α ∼ 0.70–0.94 and T_shell ∼ 0.69–0.81 keV. If T_ext = 0.2 keV, using the densities from Tables 2 and 3, we have (ρ_shells/ρ_ext) ∼ 1.1–1.7, (T_shells/T_ext) ∼ 3.5–4.1, and thus (J_shells/J_ext) ∼ 7–68. If T_ext = 0.3 keV, we have (ρ_shells/ρ_ext) ∼ 1.9–3.0, (T_shells/T_ext) ∼ 2.3–2.7, and thus (J_shells/J_ext) ∼ 18–164.

These results show that shock compression alone could be enough to fully account for the enhancement in radio synchrotron emission if the conditions are favorable (a strong shock and a magnetic field tangential to the shock front). It is likely that, even in the best conditions, there is a component of the magnetic field that is perpendicular to the shock front. Our results do not preclude local particle acceleration—it is in fact needed if the conditions at the shock are not ideal—but we know that any particle reacceleration is not efficient enough to power synchrotron emission in the X-rays.

When making the parallel between the shells of Circinus and what is observed in SNRs, some caution must be exercised. In young SNRs where B field amplification is observed, it is mediated by efficient particle acceleration (Völk et al. 2005). It seems to be clear that, in most cases, field compression alone cannot account for the acceleration of cosmic rays that is observed in SNRs, which are in fact the main source of Galactic cosmic rays. The field amplification in SNRs is much larger than the enhancement expected from shock compression alone, given that the values of B in SNRs tend to be of the order of mG (Vink & Laming 2003), 2–3 orders of magnitude higher than the values found in spiral galaxies, where typical values for the ISM of the disk are ∼10 μG (see, e.g., Beck et al. 1996, 1994; Vallée 2011; Hummel et al. 1991). Synchrotron losses also become substantial at the values of B found in most SNRs. Rayleigh–Taylor instabilities at the shock front (Guo et al. 2012) seem to be present in most SNRs. While Rayleigh–Taylor instabilities may be behind some of the B enhancement in shock shells around radio galaxy lobes, strong instabilities are not expected to develop until the jet has switched off and the radio lobes have become buoyant (see, e.g., Brüggen & Kaiser 2001).

The fact that the spectral index in the hotspots of Circinus is flatter than that of the lobes suggests that there is spectral aging of the synchrotron emission, evolving radially outward from the hotspots. In the model we discuss here the electron population in the shells, however, must come from electrons from the external medium swept up in the shock. These electrons are already relativistic, as we know from synchrotron observations of the halos of spiral galaxies (see, e.g., Beck et al. 1996). We cannot directly measure the spectral index of the emission outside the shells, since it is too faint. However, we studied the spectral index in the disk and found a value α ∼ 1. This value is similar to that found in the shells, hinting that this is the spectral index of the electron population outside the shells as well. This provides additional support to the idea that the halo electrons swept up in the shock are not strongly reaccelerated, since that would flatten their spectral index, and that the main mechanism behind the enhanced synchrotron emission in the shells we observe in the radio maps is indeed a B field enhancement. The thermal nature of the X-ray emission in the shells also supports the hypothesis of shock compression as the dominant mechanism behind the radio emission enhancement.

We carried out equipartition calculations for the shells and found that, under those conditions, the pressures we derive are over an order of magnitude below our measurements. This suggests that the shells are not in equipartition and that the contribution to the total energy density from the synchrotron-emitting electrons is not the dominant one. This is to be expected, since the spatial coincidence between the X-ray-emitting hot gas and the synchrotron-emitting plasma implies that they coexist in this region, and the former component is expected to dominate the energy density.

As mentioned above, compressed magnetic fields may be common to low-power radio galaxies. Given that field enhancement through shock compression does not preempt particle acceleration, it is likely that in most systems a combination of both is responsible for the observed emission. An accurate measurement of the spectral index and emissivity of the electrons outside the lobes and better X-ray statistics for both the shells and the external environment would allow us to estimate the contributions from both processes in Circinus and in other systems. In the case of systems more similar to Cen A, where particle acceleration seems to be causing synchrotron emission in the shells, all the way to the X-rays, constraints on the magnetic field in the shock, e.g., via rotation measure analysis, as in Guidetti et al. (2011), could be used for estimations on the contribution to the emission caused by magnetic field compression.

Our results suggest that the parallel between the jet-driven shocks of low-power FR I galaxies and those of SNRs, where a similar morphology and some of the same physical processes can be expected, is more common than we previously thought. The fact that it has only been observed in Circinus is mainly due to the small distance between this galaxy and our own, which allows us to spatially resolve these structures and observe their characteristics with the level of sensitivity of our current instruments. This is particularly relevant for more distant low-power radio galaxies in general and those with late-type hosts in particular. It is also significant for the edge-brightening in the radio lobes of Mrk 6 (Kharb et al. 2006; Mingo et al. 2011), which could be explained through this mechanism. As pointed out in Section 4.2, the shock results suggest that we are facing a scenario similar to the one we found in NGC 3801, suggesting that some of our conclusions may apply to this object, as well as, possibly, to NGC 6764.