Elliptical Galaxy in the Making: The Dual Active Galactic Nuclei and Metal-enriched Halo of Mrk 273

The 4–6 keV image and the corresponding radial profiles in the Nuclear Region are shown in Figure 4. The peak of the X-ray image overlaps with the SW nucleus seen in the HST NICMOS H-band (F160W) image. A possible secondary peak can also be seen by eye in the X-ray image, which is associated with the NE nucleus seen in the H-band image. In addition, the radial profiles of the surface brightness along the northeast (NE) and southwest (SW) directions are compared. The profiles are centered on the peak of the X-ray image. The radial profile along the northeast direction confirms the secondary peak identified by eye.

Zoom In Zoom Out Reset image size

Previous studies have revealed AGN activity in the central region of Mrk 273 (e.g., Colina et al. 1999; Scoville et al. 2000; Iwasawa et al. 2011a), and a possible dual AGN was suggested by Iwasawa et al. (2011a, 2018) and U et al. (2013). The analysis of hard X-ray image in Section 3.2.1 also suggests the possibility of dual AGN activity. In order to further explore the nature of the two nuclei, spectra of both SW nucleus and NE nucleus were extracted from apertures defined in Figure 5.

Zoom In Zoom Out Reset image size

For the SW nucleus, the aperture was centered on the peak of the hard X-ray flux. It was chosen to include the majority of the emission from the SW nucleus revealed by the NICMOS/NIC2 F160W image, and the radius of the aperture was 075. The original spectra of the SW nucleus extracted from the 2000 and 2016 data are shown in Figure 6. In general, the spectra from each individual observation of the 2016 data remain similar to each other. The spectrum from the 2000 data, however, shows clearly stronger emission in the hard X-ray band.

Zoom In Zoom Out Reset image size

In order to improve the S/N, all of the spectra from the 2016 data were combined into one, while the spectrum from the 2000 data was left alone. A comparison of the two spectra in the energy range of 3–8 keV is shown in Figure 7. Again, it is clear that the hard X-ray emission has decreased significantly from the year 2000 to 2016–2017. By fitting the spectra in the 3–8 keV band from the 2000 and 2016 data separately, we find that both the absorption column density and the intrinsic luminosity of the AGN vary over the years (N_H = ${3.62}_{-0.72}^{+1.83}$ ×10²³ cm⁻² and 4–8 keV flux = ${9.98}_{-1.36}^{+2.73}$ × 10⁻¹³ erg s⁻¹ cm⁻² for the 2000 data, and N_H = ${1.22}_{-0.29}^{+0.43}$ × 10²³ cm⁻² and 4–8 keV flux = ${1.31}_{-0.10}^{+0.12}$ × 10⁻¹³ erg s⁻¹ cm⁻² for the 2016 data, respectively.¹⁷ ). In the following, we search for the origin of this drop of flux in the hard X-ray.

Zoom In Zoom Out Reset image size

We first assume that only the intrinsic luminosity of the AGN has decreased over the years. Following this scenario, both spectra from the 2000 data and the 2016 data were fitted in 3–8 keV energy range simultaneously, with a model made up of one absorbed power-law component and an Fe Kα line. All parameters for the two epochs were tied together, except the normalizations for the power-law component and the Fe Kα line. As a first trial, the photon index of the power-law component was set as a free parameter. The fit is acceptable (reduced ${\chi }^{2}$ = 1.50), except that it gives a power-law photon index of ∼0.6, which is unreasonably small for an AGN. For example, this value is well below the measured lower limit (Γ = 1.4; see Figure 8 in Ueda et al. 2014) for the Swift/BAT hard X-ray-selected AGNs. The photon indices derived in the same paper are Γ = 1.94 with a standard deviation of 0.09 for Type 1 AGNs and Γ = 1.84 with a standard deviation of 0.15 for Type 2 AGNs. Therefore, we fixed the power-law index to the standard value of Γ = 1.9 (e.g., see Piconcelli et al. 2005; Ishibashi & Courvoisier 2010; Ueda et al. 2014, and references therein). The best-fit model gave a reduced ${\chi }^{2}$ of 1.51 (i.e., it is virtually the same as when Γ was left as a free parameter). The results are shown in Figure 7, and the best-fit parameters are summarized in Table 2. Although not statistically significant, there are possible residuals corresponding to other iron lines with energy higher than 6.4 keV in the rest frame. From year 2000 to year 2016–2017, the total flux in 3–8 keV without absorption correction has decreased from 3.10 × 10⁻¹³ erg s⁻¹ cm⁻² to 1.21 × 10⁻¹³ erg s⁻¹ cm⁻². This is consistent with the variability on a scale of several years seen from observations at different epochs as discussed in Xia et al. (2002), Teng et al. (2009, 2015), and Iwasawa et al. (2018).

Table 2.  Results from Fits of the Individual SW and NE Nuclei

Model Componenta	Parameters	SW Nucleus		NE Nucleus
		2000 Datab	2016 Datab	2000 + 2016 Datac
PL	NHd	${2.79}_{-0.22}^{+0.30}$ (t)	${2.79}_{-0.22}^{+0.30}$ (t)	${6.78}_{-3.31}^{+4.99}$	6.78(fixed)
	Γ	1.9(fixed)	1.9(fixed)	1.9(fixed)	1.9(fixed)
Fe Line	Elinee	${6.36}_{-0.04}^{+0.04}$ (t)	${6.36}_{-0.04}^{+0.04}$ (t)	N/A	${6.64}_{-0.11}^{+0.11}$
	widthe	${0.11}_{-0.02}^{+0.02}$ (t)	${0.11}_{-0.02}^{+0.02}$ (t)	N/A	${0.20}_{-0.07}^{+0.07}$
	EWe	${0.23}_{-0.05}^{+0.06}$	${0.95}_{-0.12}^{+0.05}$	N/A	${0.70}_{-0.29}^{+0.99}$
	fluxf	${3.67}_{-2.35}^{+2.35}$	${5.18}_{-0.07}^{+0.07}$	N/A	${0.98}_{-0.58}^{+0.62}$
MEKAL	NHd	N/A	N/A	0.0009(fixed)	0.0009(fixed)
	kTg	N/A	N/A	${7.21}_{-2.29}^{+4.47}$	7.21(fixed)
	Z/Z⊙	N/A	N/A	${2.31}_{-1.46}^{+1.16}$ h	0(fixed)
Flux (4–8 keV)i		${2.94}_{-0.50}^{+0.50}$	${1.16}_{-0.24}^{+0.24}$	${0.39}_{-0.22}^{+0.01}$ j
Flux (4–8 keV, absorption-corrected)i		${5.74}_{-0.95}^{+0.95}$	${2.29}_{-0.51}^{+0.51}$	${1.60}_{-0.85}^{+0.04}$ j
Luminosity (4–8 keV, absorption-corrected)k		${1.22}_{-0.20}^{+0.20}$	${0.49}_{-0.11}^{+0.11}$	${0.35}_{-0.18}^{+0.01}$ j
${\chi }_{\nu }^{2}$ (DOF)l		1.5(52)	1.5(52)	⋯	⋯
CSTAT (DOF)m		⋯	⋯	158.2(201)	156.7(201)

Notes.

^aDifferent model components used in the fitting. They are the absorbed power-law component (PL), the Fe line with Gaussian profile, as well as the thermal, hot diffuse gas component (MEKAL). "N/A" means that the component is not included in the corresponding model. All of the errors listed correspond to a confidence range of 90%, or ∼±1.6σ. For the data fitted with the CSTAT statistic (i.e., the results of the NE nucleus), the errors are derived from the default MCMC method implemented in XSPEC. ^bThe results of the simultaneous fitting to the spectra of the SW nucleus from the 2000 data and 2016 data, assuming that the change of hard X-ray flux is caused by the intrinsic variability of the central engine. Label (t) means the corresponding parameters are tied together in the fitting. ^cThe spectra from the 2000 data and 2016 data are fitted simultaneously. ^dAbsorption column density. In units of 10²³ cm⁻². ^eIn the rest frame and in units of keV. ^fIn units of 10⁻⁶ photons s⁻¹ cm⁻². ^gIn units of keV. ^hAbundance of Fe in units of solar abundance. All other elements are fixed at solar abundance. ⁱIn units of 10⁻¹³ erg cm⁻² s⁻¹. ^jThe results of the four fits are combined together for simplicity, where the errors represent the full range of the fluxes and luminosities obtained for the four fits. ^kIn units of 10⁴³ erg s⁻¹. ^lThe reduced ${\chi }^{2}$ (${\chi }_{\nu }^{2}$) from the fits and the corresponding degrees of freedom (DOF). ^mThe CSTAT statistics from the fits and the corresponding degrees of freedom (DOF).

Download table as:  ASCIITypeset image

Alternatively, an increase in the column density of the absorbing material in front of the central engine alone can also lead to the observed decrease of the hard X-ray flux. A model with tied power-law photon indices and normalizations but independent absorption column densities was fitted to the data. However, the reduced ${\chi }^{2}$ (2.7) from this fit is poor. Therefore, these results seem to favor the first scenario where the decrease of the hard X-ray luminosity is caused by the fading of the central engine. However, we cannot formally rule out the possibility that both scenarios could be at work simultaneously, because we have to fix the photon indices of the power laws in our fits due to the degeneracies among the parameters, which might bias our results.

The measured hard X-ray (4–8 keV) continuum fluxes from the SW nucleus are ∼3.2 × 10⁻¹³ erg s⁻¹ cm⁻² in the 2013 XMM-Newton data (with the contribution of the NE nucleus subtracted; Iwasawa et al. 2018) and ∼7.2 × 10⁻¹⁴ erg s⁻¹ cm⁻² measured in the 2016 Chandra data (here we assumed that the continuum flux of the NE nucleus in the 2013 XMM-Newton data was the same as that measured in the 2016 Chandra data, given that the flux of the NE nucleus remained the same in 2000 and 2016–2017). The corresponding Fe Kα line fluxes of the SW nucleus are ∼3.6 × 10⁻⁶ photons cm⁻² s⁻¹ for the 2013 XMM-Newton data (Iwasawa et al. 2018) and ∼5.2 × 10⁻⁶ photons cm⁻² s⁻¹ for the 2016 Chandra data. Over these ∼3 years, the continuum flux has therefore dropped by a factor of ∼4.4, while the Fe Kα line flux has increased by a factor of ∼1.4. Therefore, the line emission has not followed the continuum precisely. This sets a lower limit on the physical scale of the inner edge of the torus (∼1 pc) if the Fe Kα line arises from that region. Note that this argument only relies on the fact that the fluxes of the Fe line and the continuum have changed differently over the years; the exact change in flux does not matter here.

For the NE nucleus, the aperture was chosen to include all of the hard X-ray emission in the vicinity of the secondary peak seen in the 4–6 keV image. The radius of the aperture was chosen to be 1'', in order to contain as much hard X-ray emission associated with the NE nucleus as possible, but minimize the overlap with the aperture used for the SW nucleus.

As shown in Figure 6, the hard X-ray flux of the SW nucleus has dropped by ∼60% from 2000 to 2016, making the hard X-ray emission from the NE nucleus less contaminated by emission from the SW nucleus. As a result, the spectra of the NE nucleus from the 2016 data are less affected by the hard X-ray emission from the SW nucleus. All spectra of the NE nucleus from the 2016 data were combined for this analysis.

The raw spectra are shown in Figure 8, and the combined spectrum from the 2016 data is shown in Figure 9. Due to the limited counts obtained, the spectral features critical for the spectral fitting are washed out if the spectra are binned to 15 counts bin⁻¹. Unbinned spectra, in this case, have bins with zero counts and will therefore bias the results (see footnote 15). Therefore, the spectra were binned mildly to 1 count bin⁻¹, and the CSTAT statistic was used for the fits. The spectra were fitted in the energy range of 3–8 keV, with a model consisting of a thermal gas component with variable Fe abundance (Mewe–Kaastra–Liedahl or MEKAL; see Liedahl et al. 1995 and references therein) and Galactic absorption (with column density of 9 × 10¹⁹ cm⁻²; see Kalberla et al. 2005), as well as a power-law component with absorption. The photon index of the power-law component was fixed at 1.9, as adopted for the SW nucleus. The best-fit models are shown in Figure 9. The best-fit results are listed in Table 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We have also fitted the spectra of the NE nucleus in the full energy range (0.4–8 keV) with the same model, and the results of the best fit are consistent with those obtained from the fit in the energy range of 3–8 keV. Specifically, a temperature of kT = ${6.24}_{-2.37}^{+3.86}$ keV for the thermal component and an absorption column density of N_H = ${9.00}_{-3.22}^{+4.98}\,\times$ 10²³ cm⁻² for the power-law component were obtained. They are omitted in Table 2 to avoid redundancy. These results suggest the coexistence of a very hot, Fe xxv line-emitting gas and a heavily absorbed AGN.

In order to gain a better knowledge of the Fe emission line from the NE nucleus, we also fitted the spectra with a model consisting of an absorbed power law, a zero-metallicity thermal gas component, and an Fe line with a Gaussian profile. The fit failed to converge when the absorption column density for the power law and the temperature of the thermal gas component were set as free parameters. We have thus fixed those parameters to their best-fit values obtained from the fits with the power-law plus MEKAL model described in the last paragraph. The results of the new fit are summarized in Table 2. Although the Fe line seen in the SW nucleus is Fe Kα with a rest-frame energy of 6.4 keV, the dominant Fe line seen in the NE nucleus is instead Fe xxv with a rest-frame energy of ∼6.7 keV. The broad width of the Fe xxv line suggests that the line is blended with other highly ionized Fe lines.

Another evidence of AGN activity in the NE nucleus comes from the shape of the hard X-ray continuum. As shown in Figure 10, eight parallel pseudoslits were used to extract spectra across the Nuclear Region. All of the slits are perpendicular to the vector connecting the SW and NE nucleus. The spectra are shown in Figure 11.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The flux (in units of counts s⁻¹ keV⁻¹) in the two energy bins, 4–5 keV and 6.5–8 keV in the observer's frame, was used to monitor the variation of the continuum across the pseudoslits. The Fe Kα 6.4 keV line and highly ionized Fe xxv 6.7 keV line (possibly blended with other highly ionized Fe lines) are located in the 6–6.3 keV and 6.3–6.5 keV energy bins in the observer's frame separately. Therefore, the line emission does not contribute to the 4–5 keV and 6.5–8 keV bins.

At the position of the SW nucleus (slits 4 and 5), the 4–8 keV flux peaks, but the flux in both the 6.5–8 keV and 4–5 keV bins is relatively low. At the position of the NE nucleus (slits 1 and 2), the fluxes in both the 6.5–8 keV and 4–5 keV bins are larger than those in the SW nucleus. First, the higher flux in the 6.5–8 keV bin of the NE nucleus suggests a harder continuum than that of the SW nucleus. This perhaps implies the existence of a more heavily obscured AGN in the NE nucleus, compared to the one in the SW nucleus. Next, the higher flux in the 4–5 keV bin of the NE nucleus, together with the strong Fe xxv 6.7 keV line seen at the same location, suggests a significant contribution from a hot gas component.

The absorption-corrected flux of the NE nucleus in the 4–8 keV band is ∼1.6 × 10⁻¹³ erg s⁻¹ cm⁻², and it is lower than the 5.6 × 10⁻¹³ erg s⁻¹ cm⁻² estimated in Iwasawa et al. (2018). This difference might be due partly to the fact that the former is calculated from an aperture of 1'' in radius while the latter is estimated from the decomposition of the NuSTAR spectrum (extracted from an aperture with a radius of 08). The value estimated from the NuSTAR data thus sets an upper limit to the flux of the NE nucleus.

The heavily absorbed nature of the AGN in the NE nucleus is also consistent with observations at other wavelengths. The high absorption column density is supported by the previous infrared and radio observations. The NE nucleus contains most of the molecular gas in the system within an extremely compact core (radius ∼120 pc; see Downes & Solomon 1998), and the NE nucleus is the source of most of the mid-infrared luminosity (Soifer et al. 2000). Near-infrared integral field spectroscopy of the inner kiloparsec of Mrk 273 associates high-ionization coronal line emission ([Si vi] 1.964 $\mu {\rm{m}}$) with the SE radio source (radio component to the south of the NE nucleus; see footnote 14 for more information), which is likely caused by photoionization from the AGN in the NE nucleus (U et al. 2013).

In addition, current radio data suggest that the radio emission of the NE nucleus arises primarily from multiple radio supernovae and/or supernova remnants, and it is thus dominated by starburst (e.g., Carilli & Taylor 2000; Bondi et al. 2005). The X-ray data agree with these radio observations that the starburst dominates here, as the X-ray emission in the soft band (<2 keV) is consistent with emission from hot, diffuse gas most likely heated by the starburst activity. However, possible radio emission from an obscured nucleus (if present) cannot be ruled out. Indeed, there is also tentative evidence that one of the compact radio components associated with the NE nucleus may be the radio counterpart of an AGN (Bondi et al. 2005). This is consistent with our result that a heavily obscured AGN exists in the NE nucleus based on the X-ray data.

Overall, the multiwavelength data suggest the coexistence of a heavily absorbed AGN and a hot gas component heated primarily by the starburst in the NE nucleus.

To get a global view of the Nuclear Region, we extracted spectra from Region 1 in Figure 3 for analysis. All of the spectra from the 2016 data were combined as a single spectrum for fitting, while the spectrum from the 2000 data was left alone. The AGN emission and thermal emission from the hot gas within the Nuclear Region should be treated separately due to their different physical origins. Therefore, spectra with energy range of 0.4–2 keV and those with energy range of 2–7 keV were fitted separately.

There is no sign of contribution from the scattering/reflection emission of the AGN in the 0.4–2 keV band. This is supported by the fact that no point-like peak is seen in the image in the soft X-ray band. We therefore conclude that a model with thermal gas component(s) is enough to describe the data. In order to constrain the temperature and metal abundance of the thermal gas component in the Nuclear Region, simultaneous fitting of both the 2000 data and the 2016 data was carried over the 0.4–2.0 keV energy range. A single-temperature, hot diffuse gas (MEKAL) model could not describe the data (${\chi }_{\nu }^{2}$ = 3.1).

A model made up of two thermal gas components with variable metal abundance for individual elements (hereafter called VMEKAL) was then used in the simultaneous fitting. As stated above, there is intense starburst activity happening in the Nuclear Region. It is therefore natural to expect that the metallicity pattern of the hot, X-ray-emitting gas in this region is similar to that of the yield of SNe II (hereafter called SNe II metallicity pattern). On the other hand, a deviation from the SNe II metallicity pattern is also possible if there is significant amount of gas not enriched by the SNe II. Therefore, three metallicity patterns were tested in the fits for completeness. For Pattern A, both gas components had the SNe II metallicity pattern used in Iwasawa et al. (2011b). For Pattern B, the Z(Si), Z(O), and Z(Fe) of both gas components were set as free parameters, and Z(Ne) = Z(Si) = Z(Mg). Besides, all other elements were fixed at solar values. For Pattern C, the SNe II metallicity pattern used in Pattern A was adopted for the hotter gas component, and the metallicity pattern used in Pattern B was adopted for the cooler gas component.

The best-fit temperatures of the two gas components are ${0.75}_{-0.08}^{+0.07}$ keV and ${2.21}_{-0.73}^{+3.83}$ keV, respectively. For the abundance ratios, all three fitting approaches adopted above agree that supersolar α/Fe ratios are measured in both gas components, while the absolute values for the individual element abundances are not exactly the same. The median values of the best-fit model parameters and their associated errors for the cooler gas component obtained from the fits above are adopted for further analysis (see Table 3). The α/Fe ratio of the hotter gas component is SNe II-like and is omitted in Table 3 for simplicity.

Table 3.  Comparison among the Physical Properties of the Thermal Gas Component in Different Regions

Model Parametera	Nuclear Regionb	Host Galaxy Region	Ionization Cone	Southern Nebula		NE-extended Nebula
				2016 Data	2000 + 2016 Data
NHc	${0.23}_{-0.05}^{+0.07}$	${0.03}_{-0.006}^{+0.005}$	0.009(fix)	0.009(fix)	0.009(fix)	0.009(fix)
kT (keV)	${0.75}_{-0.08}^{+0.07}$	${0.73}_{-0.10}^{+0.08}$	${0.82}_{-0.08}^{+0.09}$	${0.60}_{-0.10}^{+0.10}$	${0.57}_{-0.07}^{+0.06}$	${0.31}_{-0.01}^{+0.07}$
ned	24	21	16	3	3	2
O/O⊙	${0.59}_{-0.34}^{+0.48}$	${0.48}_{-0.21}^{+0.40}$	${0.44}_{-0.24}^{+0.45}$	${0.79}_{-0.40}^{+1.14}$	${0.83}_{-0.22}^{+1.05}$	${0.16}_{-0.05}^{+0.63}$
Si/Si⊙	${0.53}_{-0.22}^{+0.20}$	${0.17}_{-0.15}^{+0.14}$	${0.37}_{-0.22}^{+0.14}$	${1.00}_{-0.04}^{+2.13}$	${1.32}_{-0.16}^{+1.74}$	${0.64}_{-0.30}^{+1.25}$
Fe/Fe⊙	${0.08}_{-0.03}^{+0.03}$	${0.10}_{-0.05}^{+0.03}$	${0.11}_{-0.04}^{+0.04}$	${0.22}_{-0.06}^{+0.56}$	${0.30}_{-0.05}^{+0.30}$	${0.31}_{-0.11}^{+0.14}$
α/Fee	${6.6}_{-3.7}^{+3.5}$	${1.7}_{-1.6}^{+1.6}$	${3.3}_{-2.6}^{+4.2}$	${4.6}_{-2.3}^{+7.1}$	${4.5}_{-1.9}^{+3.1}$	${2.1}_{-1.8}^{+5.9}$
flux (0.4–2 keV)f	${7.96}_{-0.22}^{+0.49}$	${3.75}_{-0.18}^{+0.17}$	${1.00}_{-0.08}^{+0.06}$	${1.88}_{-0.16}^{+0.20}$	${1.78}_{-0.12}^{+0.13}$	${0.54}_{-0.08}^{+0.08}$
luminosity (0.4–2 keV)g	${2.63}_{-0.07}^{+0.16}$	${1.24}_{-0.06}^{+0.06}$	${0.33}_{-0.03}^{+0.02}$	${0.62}_{-0.05}^{+0.07}$	${0.59}_{-0.04}^{+0.04}$	${0.18}_{-0.03}^{+0.03}$
${\chi }_{\nu }^{2}$(DOF)h	1.4(93)	0.9(67)	⋯	⋯	⋯	⋯
CSTAT (DOF)i	⋯	⋯	142.8(178)	103.8(72)	124.9(121)	45.6(41)

Notes.

^aFitting Model: for the Nuclear Region, the model is two hot gas components with absorption (i.e., phabs(VMEKAL)+phabs(VMEKAL) in XSPEC). For all other regions, the model is just one hot gas component with absorption (i.e., phabs(VMEKAL) in XSPEC). The absorption column densities are fixed to the Galactic value of 9 × 10¹⁹ cm⁻² when necessary. All other elemental abundances not included in the table are fixed at solar values. All the results are from the simultaneous fitting to both the 2000 data and 2016 data. All the errors listed correspond to a confidence range of 90%, or ∼±1.6σ. For the data fitted with the CSTAT statistic, the errors are derived from the default MCMC method implemented in XSPEC. ^bThe median values obtained for the cooler gas component are listed. See Section 3.2.4 for more details. ^cThe absorption column density, in units of 10²² cm⁻². ^dElectron number density, in units of 10⁻³ cm⁻³. Spherical volumes with a filling factor of 1 are assumed in the calculation except for the Ionization Cone Region, where a bi-cone with an opening angle of 75° and a filling factor of 1 is assumed. ^eAbundance ratios of α elements to Fe, determined by the ratio of the Si (row 4) and Fe (row 5) abundance. The values are rounded to one decimal. For the data fitted with the ${\chi }^{2}$ statistic, the errors were obtained assuming standard error propagation: ${\sigma }_{a/b}=\sqrt{{({\sigma }_{a}/b)}^{2}+{({\sigma }_{b}a/{b}^{2})}^{2}}$. For the data fitted with the CSTAT statistic, the errors were derived from the default MCMC method in XSPEC. ^fIn units of 10⁻¹⁴ erg s⁻¹ cm⁻², absorption-corrected. ^gIn units of 10⁴¹ erg s⁻¹, absorption-corrected. ^hThe reduced ${\chi }^{2}$ (${\chi }_{\nu }^{2}$) from the fits and the corresponding degrees of freedom (DOF). ⁱThe CSTAT statistics and the corresponding degrees of freedom (DOF). The spectra were binned to 1 count bin⁻¹.

Download table as:  ASCIITypeset image

For the spectrum with energy range of 2–7 keV, the fit was carried out on 2016 data. As a first trial, a model made up of a single absorbed power-law component and an Fe line was used. As shown in the left panel of Figure 12, it hardly describes the data (${\chi }_{\nu }^{2}$ = 3.0). Second, a hot thermal gas component (MEKAL) was added to the model. As shown in the middle panel of Figure 12, the model matches the data better (${\chi }_{\nu }^{2}$ = 0.95). The best-fit temperature of the thermal gas component is ${2.18}_{-0.52}^{+1.80}$ keV, which is consistent with that of the hotter gas component used in the fit of 0.4–2 keV spectra (${2.21}_{-0.73}^{+3.83}$ keV). However, the flux of the best-fit model tends to be systematically lower than that of the data beyond 6.5 keV in the observer's frame, suggesting the possible existence of another hard X-ray component.

Zoom In Zoom Out Reset image size

Another heavily absorbed power-law component was thus added to the model, as shown in the right panel of Figure 12. The best-fit model gives a reduced ${\chi }^{2}$ of 0.8, and the absorption column densities of the two power-law components are ∼2.1 × 10²³ cm⁻² and ∼1 × 10²⁴ cm⁻², respectively. The error for the latter is relatively large, which is on the order of 5 × 10²³ cm⁻². This is probably caused by the degeneracies among the fit parameters, most likely the degeneracy between the absorption column density and the temperature of the hot gas. These results agree with those obtained in the fits of the SW and NE nuclei separately (Table 2). They are also broadly consistent with the values listed in Iwasawa et al. (2018; ${2.6}_{-1.1}^{+0.9}$ × 10²³ cm⁻² and ${1.4}_{-0.4}^{+0.7}$ × 10²⁴ cm⁻², respectively), which are obtained from the fit of NuSTAR data.

For the absorption-corrected flux of the two power-law components, the value of the less absorbed power-law component (${1.97}_{-0.09}^{+0.34}\times {10}^{-13}$ erg cm⁻² s⁻¹) agrees with that obtained from the 075 aperture spectrum of the SW nucleus listed in Table 2, while the value for the more absorbed one (corresponding to the NE nucleus) is not well constrained (<2 × 10⁻¹² erg cm⁻² s⁻¹). Together with the results presented in Sections 3.2.2 and 3.2.3, our results of the two heavily absorbed AGNs are in general consistent with those obtained in Iwasawa et al. (2018).

In order to study the spatially extended halo emission, an image in flux units is needed. To obtain such an image, the dependence of the collecting area on the energy and position as well as the effective exposure in different regions of the detector needs to be taken into account properly. The image of the Southern Nebula in flux units was produced with the CIAO script flux_image for each observation. In order to obtain the spectral weights needed to produce the exposure map, preliminary spectra were extracted from an elliptical region (Region 3 in Figure 3) for each observation. All of the spectra from the 2016 data were combined into one single spectrum. The combined 2016 spectrum and 2000 spectrum were then fitted with a single-temperature gas model (MEKAL) simultaneously, and the best-fit model spectra were used to generate the spectrum weights for exposure maps of each observation separately. Finally, all the images were combined.

The final product, the image of the Southern Nebula in flux units, is shown in the right panel of Figure 2 and in Figure 3. Soft X-ray emission is seen extending up to ∼40 kpc in the directions both perpendicular to and along the tidal tail. Apparently, the X-ray emission is suppressed in the tidal tail region seen in the optical image, where a decrease of ∼23% in the surface brightness is measured, compared to that of the whole Southern Nebula. This suggests further that the X-ray emission is not exclusively caused by the tidal effects, which is discussed further in Section 4 (see also Iwasawa et al. 2011a).

The spectra of both the 2000 and the 2016 data were extracted from Region 3 in Figure 3. As the 2000 data were taken 16 years before the 2016 data, the instrument response of ACIS-S in the soft X-ray has changed significantly. As shown in Table 1, only ∼680 counts were obtained in 2016–2017 for the Southern Nebula. Although the spectrum from the 2016 data seems noisier, it does not mean that the data are of lower quality compared to the 2000 data. In Figure 16, the 2000 and 2016 data were binned with the same method, i.e., at least 5 counts bin⁻¹ for visualization. As more counts were obtained in the 2016 spectrum, the bin size of the 2016 spectrum (in units of keV) is much smaller than that of the 2000 spectrum (e.g., this can be seen most clearly in the 1.6–2.0 keV data in Figure 16). This is why the 2016 spectrum appear more "noisy" than they actually are. In order to get robust measurements from the spectra of the Southern Nebula region, the spectral fitting was carried out as follows.

Zoom In Zoom Out Reset image size

We first checked whether a model with a fixed, solar abundance can fit the data. For this exercise, the spectra were binned to 15 counts bin⁻¹ so that the ${\chi }^{2}$ statistics and F-test could be applied. The fits gave a ${\chi }^{2}$ of 87.5 with a DOF of 55 in the case of the fixed solar abundance, and a ${\chi }^{2}$ of 49.2 with a DOF of 52 for the variable abundance model. The p-value of the F-test is very low (∼1.2 × 10⁻⁶), implying that the fixed abundance model is not sufficient to properly describe the data.

We then turned to the model with variable abundances. First, only spectra from the 2016 data were combined and fitted. In agreement with Iwasawa et al. (2011a), emission features of Mg xi 1.34 keV and Si xiii 1.85 keV are seen by eye in the spectrum. Such line emission suggests the existence of abundant α elements in the nebula. The spectrum was binned to 1 count bin⁻¹, and the CSTAT statistic was used for the fit. The spectrum was fitted with a single-temperature gas model (VMEKAL) with Galactic absorption. Specifically, the abundances of Si, Mg, and Ne were tied together, while those of O and Fe were left as independent variables. The abundance for all other elements not mentioned above were fixed at solar values. The corresponding results from the fits are summarized in Table 3.

The robustness of the results to the change of the abundance assumption was investigated. This was explored by setting the abundance of all other elements free in the fitting, but they had not changed the results from the fits to a noticeable extent. This is expected, since no line features of those elements are strong enough to affect the results of the fit in the spectral range of 0.4–2 keV. As a result, all those elements were fixed at solar values in the fits below.

Second, simultaneous fitting to both the 2000 and the 2016 data was carried out. Again, the spectra were binned to 1 count bin⁻¹, and the CSTAT statistic was used for the fit. The model and the parameter settings were the same as the one used when the 2016 data alone was fitted. The results are summarized in Table 3. In general, the two fits give results that are consistent with each other. Taking into account the errors, both fits give the same temperature for the nebula. The largest difference between the two fits lies in the absolute metallicity, but both results still broadly agree with each other.

As stated at the end of the Section 2.1, the uncertainty of the α/Fe ratio was obtained through the default MCMC method implemented in XSPEC when the CSTAT statistic was used. The probability contours of the Si (Ne, Mg) and Fe abundances are shown in Figure 17, based on the simultaneous fitting of the 2000 and 2016 data. It can be seen that, although the absolute values of Si (Ne, Mg) and Fe abundance are uncertain, the ratio of α/Fe is constrained relatively well by the data (α/Fe = ${4.5}_{-1.9}^{+3.1}$ for ∼±1.6σ, or α/Fe = ${4.5}_{-2.4}^{+5.5}$ for ∼±3σ). The detection of a supersolar α/Fe ratio in the Southern Nebula is therefore robust (>3σ).

Zoom In Zoom Out Reset image size

The spatial variations of physical properties within the Southern Nebula may help reveal the origin of the hot gas. The possible spatial variations of the temperature and metallicity of the hot gas were explored.

Spectra were extracted from regions on both the east and west sides of the tidal tail for inspection. The temperatures of the western and eastern parts of the Southern Nebula agree with each other and are similar to the value of the whole nebula (kT ∼ 0.57 keV), taking the errors into consideration. The relative metal abundance ratios of the two parts also show similar values to those of the whole nebula (α/Fe = ${4.5}_{-1.9}^{+3.1}$). There is a hint that the absolute metal abundance in the western part of the nebula is systematically larger than that in the eastern part and that in the whole nebula. However, the uncertainties in the abundance are quite large (a factor of ≳2), and this result is thus not conclusive. Moreover, the difference between the northern and southern parts of the Southern Nebula was also explored. Again, no clear difference was seen in either the temperature or the metallicity pattern.

However, a detailed analysis of the narrowband images centered on the Mg xi 1.34 keV line and Si xiii 1.85 keV line has revealed tentative spatial variations within the Southern Nebula. As shown in Figure 18, there is a separate Mg xi 1.34 keV bright region (Mg-bright region) and Si xiii 1.85 keV bright region (Si-bright region) within the Southern Nebula. The spectra of both regions as well as the spectrum for the whole Southern Nebula are shown in Figure 19, and similar temperatures (kT ∼ 0.57 keV) were obtained for all spectra by fitting a thermal gas model to them. The spectra demonstrate the stronger Mg xi 1.34 keV line and Si xiii 1.85 keV line emission in the corresponding regions, respectively. As Si and Mg are both α elements, the relative abundance of these two elements should not vary much in different regions. One possible explanation for this variation is that the ionization states in the two regions are different, where the Si-bright region is more highly ionized than the Mg-bright region. The low S/N (≲3) of these emission features prevent us from carrying out a more quantitative modeling of this apparent variation.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As shown in Table 3, the measured temperature of the nebula is ${0.57}_{-0.07}^{+0.06}$ keV. Assuming the virialization of the hot gas and an NFW density profile for the dark matter halo (Navarro et al. 1997), the gas virial temperature can be written as follows:

$$\begin{eqnarray*}&&{{kT}}_{\mathrm{vir}}=\displaystyle \frac{1}{2}\mu {m}_{p}\displaystyle \frac{{{GM}}_{\mathrm{vir}}}{{r}_{\mathrm{vir}}}\simeq 0.135\left(\displaystyle \frac{{M}_{\mathrm{vir}}}{{10}^{12}{M}_{\odot }}\right){\left(\displaystyle \frac{{r}_{\mathrm{vir}}}{100\mathrm{kpc}}\right)}^{-1}\,\mathrm{keV}.\end{eqnarray*}$$

By further assuming the relations between the virial radius and both the halo mass and redshift presented in Mo & White (2002), the dark matter halo mass can be obtained through the equation ${{kT}}_{\mathrm{vir}}\sim 0.042\,{M}_{12}^{2/3}(1+z)$ keV, where M₁₂ is the virial mass expressed in units of 10¹² ${M}_{\odot }$. Similar to what is suggested for NGC 6240 in Nardini et al. (2013), a possible evolutionary scenario consistent with the temperature (∼0.57 keV) of the nebula is that of a cold dark matter halo with a total virial mass of ∼10¹³ ${M}_{\odot }$ formed at z ∼ 2.

This is almost an order of magnitude more massive than that of the Milky Way (∼1.3 × 10¹² ${M}_{\odot }$; see McMillan 2011 for details.) In addition, this value distinguishes group-scale halos from galaxy-scale halos (see Humphrey et al. 2006). The mass of the dark matter halo estimated here is unreasonably high, given the stellar mass of Mrk 273 (∼10¹¹ ${M}_{\odot }$; see Veilleux et al. 2002; Rodríguez Zaurín et al. 2009; U et al. 2012) and its dynamical mass (∼10¹² ${M}_{\odot }$, based on the velocity dispersion of the CO absorption features in the near-infrared, and assuming a King model with tidal radius to core radius of ∼50; Tacconi et al. 2002) of Mrk 273. Therefore, extra heating mechanism for the hot gas is at work.

On the contrary, the NE-extended Nebula shows a lower temperature of ${0.31}_{-0.01}^{+0.07}$ keV. Following the same procedure used for the Southern Nebula, a dark matter halo mass of ∼3.5 × 10¹² ${M}_{\odot }$ is obtained, which is a more reasonable value for the halo mass of Mrk 273. Compared to the Southern Nebula, the NE-extended Nebula is thus more likely to be the pre-existing, virialized gaseous halo.

Both galactic outflows generated from the nuclear/circumnuclear region and the merging activity can heat the gas in the nebula. It is thus necessary to discuss their relative importance.

One way to estimate it is to look at the required shock velocity. Assuming that the gas is heated up after being swept by an adiabatic shock, the temperature of the gas can be derived as ${{kT}}_{{\rm{s}}}=3\mu {m}_{p}{v}_{{\rm{s}}}^{2}/16$, where v_s is the speed of the shock front. Adopting kT = 0.57 keV (see Table 3), v_s will be 710 km s⁻¹, which is too high for the relative orbit speeds within a merging system like Mrk 273 (e.g., the inferred orbital velocity of the two nuclei of NGC 6240 is just 155 km s⁻¹; see Tecza et al. 2000 for details). In addition, the maximum speed of the non-outflowing ionized gas only reaches a maximum of 350 km s⁻¹, as is measured through the narrow component of the Hα emission in Rupke & Veilleux (2013). Therefore, the merging activity alone does not seem capable of causing shocks with the required v_s.

As is presented in Section 3.4, hot gas (kT ∼ 0.8 keV; see Table 3) associated with the ionization cones and outflowing gas extends to a scale of ≳5 kpc × 5 kpc. This outflowing hot gas might be the heating source for the Southern Nebula after it moves outward to a larger scale. Similar cases of hot gas heated by the AGN- and/or starburst-triggered outflow have been observed in several other galaxies (e.g., Wang et al. 2014; Tombesi et al. 2016, 2017). In addition, high-velocity outflow in Mrk 273 has been detected in the ionized and neutral gas phases (∼4 kpc; Rupke & Veilleux 2013; Rodríguez Zaurín et al. 2014), as well as the molecular gas phase (∼550 pc; U et al. 2013; Veilleux et al. 2013; Cicone et al. 2014; Aladro et al. 2018). The ionized outflow has been observed to reach a speed as high as ∼1500 km s⁻¹ in projection, while the molecular outflow reaches a maximum speed of ∼900 km s⁻¹ in projection. These high-speed outflows may be capable of causing adiabatic shocks with v_s ∼ 710 km s⁻¹ in the nebula, if there is no significant decrease in the outflow speed before they reach the nebula.

Another independent piece of evidence regarding the origin of nebulae comes from the α/Fe abundance ratio of the hot gas. Iron is nearly entirely produced from Type Ia supernovae (SNe; i.e., exploded white dwarfs in close binary systems), while the majority of α elements originate from SNe II (i.e., core-collapse massive stars). Based on the prediction of nucleosynthesis models for SNe II (e.g., Nomoto et al. 2006, 2013; Heger & Woosley 2010), Si/Fe ratios can reach as high as ∼3×–5× solar value, while ratios of ∼0.5 solar value are expected from SNe Ia (e.g., Nagataki & Sato 1998; Seitenzahl et al. 2013). Therefore, the α/Fe ratios can provide information about how the gas was enriched.

The high α/Fe ratios in both the Nuclear Region (Region 1; ${6.6}_{-3.7}^{+3.5}$) and the Southern Nebula (Region 3; ${4.5}_{-1.9}^{+3.1}$) might indicate a physical connection between the two, perhaps through the outflow generated in the Nuclear Region. Similar evidence for the transport of α-enhanced material out to a scale ≲10 kpc by galactic winds has been seen in other nearby galaxies (e.g., M82; Tsuru et al. 2007; Ranalli et al. 2008; Konami et al. 2011, and references therein).

If indeed the Southern Nebula is heated and enriched through the outflow, there are several reasons to believe that multiple outflow events are involved.

First of all, the maximum silicon yield of an SNe II is ∼0.1–0.3 ${M}_{\odot }$ for a massive-star progenitor with Z ≤ 0.02 (e.g., Nomoto et al. 2013). The total amount of silicon within the Southern Nebula is estimated to be ∼2 × 10⁶ ${M}_{\odot }$, adopting a derived gas mass of ∼1.5 × 10⁹ ${M}_{\odot }$ and assuming a 1.5× solar abundance for silicon estimated from the spectra. This can be translated into 1 × 10⁷ SNe II. Given the current SFR¹⁸ of ∼160 M_⊙ yr⁻¹, a continuous star formation activity of ≳10⁷ yr is required. It is thus very unlikely that the current star formation contributes solely to the enrichment. Past star-forming and/or starburst activities over a much longer timescale are required. Detailed spectral analyses of the circumnuclear region have indeed revealed stellar populations both young (3–60 Myr) and old (0.7–12.5 Gyr; see Raimann et al. 2003; Rodríguez Zaurín et al. 2009, for details).

Second, the characteristic timescale needed for the outflow to affect the Southern Nebula on a spatial scale of ∼40 kpc can be estimated. The average outflow speed of the warm ionized gas in projection is ∼1000 km s⁻¹ (measured from the average V_98% of the Hα emission; see column (7) of Table 3 in Rupke & Veilleux 2013). Assuming that the outflow travels with a constant speed of 1000 km s⁻¹, a timescale of 40 Myr is needed for it to reach the full extent of the Southern Nebula.

Moreover, as no significant spatial variations of the α/Fe ratio are seen within the Nebula, the whole duration of the enrichment might be as long as the dynamical timescale of ∼0.1 Gyr, which is measured from the sound-crossing time, $D/{c}_{s}=D{(5{kT}/3\mu {m}_{p})}^{-1/2}$. The age of the current outflow in warm ionized phase is ≲10 Myr, based on its velocity and spatial extent (Rupke & Veilleux 2013; Rodríguez Zaurín et al. 2014). As a result, the current outflow alone cannot be responsible for the α-element enrichment already observed in the Southern Nebula.

In all, multiple outflow events in the past, on a timescale of ≲0.1 Gyr, are required to explain the supersolar α/Fe ratio in the Southern Nebula. The merger event might have helped further erase spatial fluctuations in the α/Fe ratio within the nebula (similar to the case of Mrk 231; Veilleux et al. 2014).

As shown in the bottom-right panel of Figure 1, the brighter portion of the extended Hα emission south of the galaxy (T ∼ 10⁴ K) resembles the Southern Nebula in X-ray (T ∼ 7 × 10⁶ K). The eastern part of the Southern Nebula in the X-ray seems to be more luminous and extended than the Hα nebula, although this may be caused by the different sensitivity of the X-ray and Hα observations. Spence et al. (2016) argue that the southern Hα nebula is not related to the large-scale outflow, based on the moderate velocity shifts ($| {\rm{\Delta }}V|$ ≲ 250 km s⁻¹) and line width (FWHM ≲ 350 km s⁻¹) measured in the emission lines. However, these results do not formally rule out the possibility that the Hα nebula represents the accumulated reservoir of the outflowing gas if the turbulence caused by the outflows dissipated away as the material cooled down from ∼7 × 10⁶ K to ∼10⁴ K. The high temperature and high α/Fe ratio of the Southern Nebula favor an outflow-related origin.