VERY LONG BASELINE INTERFEROMETRY OBSERVATIONS OF NGC 6240: RESOLVING THE DOUBLE NUCLEI AND RADIO SUPERNOVAE

The EVN has been used to observe the nuclear region of NGC 6240 in order to investigate the nature of the AGNs and nuclear starbursts in the galaxy. The stations used for these observations at two epochs in 2003 and 2009 were Effelsberg (Ef), Hartebeesthoek (Hh), the 76 m Lovell telescope (Jb) at Jodrell Bank, Medicina (Mc), Noto (Nt), Onsala (On), Shanghai (Sh), Torun (Tr), Urumqi (Ur), and the Westerbork (Wb) phased-array. A summary of the observational details is presented in Table 1. All observations were conducted in phase-referencing mode at all frequency bands using the nearest calibration source J1651+0129 (centered at R.A. (J2000) = 16^h51^m036620, decl. (J2000) = 01°29'23458) at 101 away from NGC 6240. Only the Lovell telescope did not participate in the phase-referencing part of the observations due to its lower slewing rate.

The first epoch observations at 1.6 GHz and 5.0 GHz were made on 2003 October 30 and November 10 and utilized a 32 MHz total bandwidth with 2 bit sampling in dual-circular polarization mode for a total data recording rate of 256 Mbit s⁻¹. Observations of 2.5–3 minutes on the target were interleaved with observations of 1.5–2 minutes on the phase-referencing source, which gave about 4.2 hr total on-source time.

Second epoch observations at 5.0 GHz, 2.2 GHz, and 8.4 GHz were made on 2009 June 16 and 17. The observations at 5.0 GHz were made in dual-circular polarization mode utilizing 64 MHz total bandwidth with 2 bit sampling for a total data recording rate of 512 Mbit s⁻¹. Three-minute scans on the target were alternated with one-minute scans on the phase-referencing source, which resulted in 5.7 hr total on-source time. The observations at 2.2 and 8.4 GHz were made in single polarization mode with the S/X dual-frequency receiving system with a bandwidth of 128 MHz at each frequency and 2 bit sampling for a total recording rate of 512 Mbit s⁻¹. Three-minute scans on the target and one-minute scans on the phase-referencing source were alternated during the observation, which resulted in 2.9 hr total on-source time at each frequency. All observations were made with the same reference position near the southern nucleus of NGC 6240 (R.A. (J2000) = 16^h52^m588903, decl. (J2000) = 02°24'03339).

The data were correlated using the EVN MkIV correlator at the Joint Institute for VLBI in Europe (JIVE) and the output visibility data were averaged to 1–2 s. Sixteen frequency channels of 500 kHz were used across each 8 MHz band. Data calibration and imaging were carried out in the NRAO Astronomical Image Processing System (AIPS). After initial data editing, the delay error corrections resulting from ionospheric effects were applied to all visibility data. A priori amplitude calibration was performed using the system temperature and gain information provided for each telescope. The delays and fringe rates for the calibration sources 3C 273 or 3C 345, and J1651+0129 were applied to the target source, NGC 6240. Using the measured flux densities of the phase-referencing source J1651+0129 at two epochs as a basis, the relative accuracy of the target flux densities is estimated to be ∼10%. This error estimate includes the effects of bandwidth and time smearing and ionospheric delays, which are estimated to be several percent for these EVN observations.

Cleaned images of 512 × 512 pixels were produced, one centered near the southern nucleus and one at the intensity peak position of the northern nucleus offset by 05051 in right ascension and 14279 in declination from the phase center. The noise levels of one to two times theoretical noise and synthesized beams from CLEAN-ed images are summarized in Table 2. Self-calibration could not be applied successfully because of the low signal-to-noise ratio (S/N) of the images. Final images have been phase-referenced to the nearby calibrator J1651+0129. Because of residual errors in the calibration of the phase-reference sources, the associated structure at the lowest contours (≲ 3σ or 4σ) in the source images may not be reliable.

The positions of the radio components and their errors at the peak brightness at each frequency and both epochs are listed in Table 3. These composite position errors have been estimated from the theoretical thermal noise errors of interferometer phase (e.g., Thompson et al. 1986), the systematic errors of the phase-referencing VLBI observations (σ_phr), and the errors of the absolute position of the phase-referencing source (σ_ref), using the quadrature equation: $\sqrt{\vphantom{A^A}\smash{\hbox{${(\theta /2{\rm (S/N)})^2 + \sigma _{\rm phr}^2 + \sigma _{\rm ref}^2}$}}}$. Here, θ represents the synthesized beam size and S/N represents the signal-to-noise ratio. The values of σ_phr (Δα cosδ = 0.060 mas, Δδ = 0.020 mas) are obtained from the simulated astrometric accuracy of the EVN array (Pradel et al. 2006), adopting a separation between calibrator and target of 10 at a position angle of 1522 and at an approximate source declination of 0°. The value of σ_ref is 0.58 (mas) from the VLBA Calibrator List (Kovalev et al. 2007). Table 4 lists the flux densities of the radio components at each observing frequency, which also includes two weaker components appearing in the earlier VLBA observations (see Figures 2–4 in Gallimore & Beswick 2004). Source sizes of these components in Table 5 are measured from the final images using the AIPS task JMFIT. Assuming Gaussian distributions for sources, the upper limits of the source sizes have estimated accuracies of 3%–30% depending on observing frequency and epoch. Table 6 presents the properties of the radio sources detected in our observations.

The new VLBI maps obtained from 1.6 to 8.4 GHz display four compact high-brightness emission components as identified in Figure 1: the nuclei, N1 (south) and N2 (north) identified in earlier interferometric observations (Beswick et al. 2001; Hagiwara et al. 2003) and two new components, RS1 and RS2, near the southern nucleus. The radio source RS1 was first identified southwest of the southern nucleus at 5.0 GHz in 2003 October and it was detected at 5.0 and 8.4 GHz in 2009 June (see Figure 2). The other radio source, RS2, was identified northeast of the southern nucleus at Epoch 1 and was more prominent at 1.6 GHz than at 5.0 GHz (see Figure 4). At Epoch 2, source RS2 was clearly detected at 5.0 GHz but not at 2.2 GHz with an upper limit flux density of 2.75 mJy beam⁻¹. All diffuse emission in the central 2'' region, observed in earlier VLA observations (e.g., Colbert et al. 1994; Baan et al. 2007), has been resolved out.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The VLBA observation from 1999 August at 8.4 GHz did not detect the two nuclei at a 5σ upper limit of 0.65 mJy beam⁻¹ (Gallimore & Beswick 2004), but the northern nucleus was detected at 8.4 GHz in 2009 June with similar sensitivity, which suggests some intensity variability of the northern nucleus. Both nuclei were detected at 5 GHz at both epochs. Their peak positions in the EVN data (Table 3) agree at all frequencies with earlier VLBA measurements using estimated uncertainties of 0.6 mas in right ascension and 1.5 mas in declination (Gallimore & Beswick 2004). The 5.0 GHz maps in Figures 2 and 3 display the two nuclear components, N1 and N2, obtained at Epoch 1 and Epoch 2, respectively. The southeastern nucleus, N1, was not detected at the highest frequency of 8.4 GHz with an upper limit of 0.55 mJy beam⁻¹ (Figure 2). Individual source maps indicate that sources N1 and N2 are resolved but that the minor axis of RS2 northeast of N1 is unresolved (Figure 4). The peak of RS1 is separated from N1 by 23.3 mas at 5.0 GHz at a position angle of ∼35°, corresponding to 11.1 pc as the distance to NGC 6240. At the lowest frequency of 1.6 GHz, N1 and RS1 are not spatially separated. Also, RS1 was not detected at 2.2 GHz at Epoch 2 possibly because of the lower S/N.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The source RS2 is located northeast of the southern nucleus in our 1.6 GHz EVN map of Figure 4. It appears that the peak position of RS2 at 1.6 and 5.0 GHz lies close to the compact component S1 at 03 northeast of N1 in the 1.7 and 2.4 GHz VLBA observations (Gallimore & Beswick 2004). On the other hand, RS2 has an offset of ≈10–15 mas at 1.6 GHz or 5.0 GHz relative to S1 at 1.7 GHz or 2.4 GHz, which is not within the estimated positional errors (Table 3). Therefore, RS2 is a different source from S1 in the earlier VLBA data. We interpret both components to be part of the circumnuclear star-forming region that extends over 10 mas or ∼5 pc around N1.

The structure of the northeastern nucleus N2 at 5.0 GHz (Figure 3) shows a northwest extension of ≈002 (≈9.5 pc) which is consistent with the earlier VLBA maps at 1.7 and 2.4 GHz (Gallimore & Beswick 2004). Although this structure at lower contours may not be reliable because of incomplete phase and amplitude calibration, an east–west structure may be explained as a core-jet structure typically seen in high-luminosity Seyfert nuclei. The northern nucleus itself remains unresolved at both 1.6 and 2.2 GHz. Only at the highest frequency of 8.4 GHz, does N2 have an east–west elongation consistent with the earlier VLBA data (Figure 3).

The northern nucleus N2 shows an inverted spectrum rather than a steep spectrum which would be more typical of a radio jet (Figure 5). The structure of N2 at 5.0 GHz is partially resolved and shows an east–west extension at its lower contours that still remains questionable because of insufficient amplitude calibration (Figure 3). Changes in source structure may also result from a different (u,v) coverage for the two epochs. The detection of the X-ray emission toward N2 (Komossa et al. 2003) is less convincing than the definite detection toward N1. Nevertheless, the radio morphology of N2 resembles that of a core-jet structure seen in Seyfert nuclei, such as NGC 1068 (Gallimore et al. 2004), but this requires further verification. By contrast with Arp 299, the VLA observations of the H₂O maser in NGC 6240 have not detected any emission toward N2 during seven years (Hagiwara et al. 2003; Hagiwara 2010).

Zoom In Zoom Out Reset image size

The southern nucleus N1 appears partially resolved in an east–west direction (Figures 2(b) and (c)) but its flux density at 5.0 GHz has not changed significantly for six years. Contrary to the spectrum of N2 (see the discussion below), the spectrum of N1 with missing Epoch 2 detections at 2.2 and 8.4 GHz may not be inverted (Figure 5), although its spectral shape (including upper limits) strongly indicates a spectral turnover around 2.0 GHz. The non-detection of N1 at 8.4 GHz suggests a relatively flat AGN-like spectrum and a lower limit for a spectral index α < 0.55 between 5.0 and 8.4 GHz, using S_ν ∝ ν^α with S_ν being the flux density at frequency ν. In addition, the radio brightness temperature > 10⁶ K (Table 6) and the detection of strong hard X-ray emission and neutral iron line emission would argue for the presence of an AGN in N1. Considering the available evidence, the southern nucleus may host both an AGN and a (circum)nuclear starburst region.

The H₂O maser features in NGC 6240 nearly coincide with the continuum peak of the southern nucleus (Hagiwara et al. 2003; Hagiwara 2010). The narrow H₂O maser lines are redshifted by ∼ 200–300 km s⁻¹ with respect to the systemic velocity of N1 (e.g., Hagiwara 2010; Baan et al. 2007) and may originate in the receding side of a compact rotating molecular disk at the nucleus (e.g., Miyoshi et al. 1995), which would also support the presence of an AGN. Alternatively, these narrow lines distributed over 120 km s⁻¹ could be explained by wind maser emission as observed in the Circinus galaxy (Greenhill et al. 2003).

The southern nucleus N1 of NGC 6240 has characteristics that are comparable to those of component A of Arp 299 (Neff et al. 2004; Ulvestad 2009). The 8.4 GHz radio power of N1 is P_N1 = 3.7 × 10²² W Hz⁻¹ (Colbert et al. 1994) and it is only a factor of two more than the 8.4 GHz power of nucleus A of Arp 299 (P_A = 1.8 × 10²² W Hz⁻¹ and P_B1 = 2.9 × 10²¹ W Hz⁻¹) (Neff et al. 2004). The compact source A of Arp 299 has been resolved into discrete sources at a milliarcsecond scale by recent EVN observations and one of these sources (A1) is an AGN candidate with a flat spectral index of −0.13 ± 0.11 (Pérez-Torres et al. 2010). Five bright and compact radio sources lie within 10 pc and are candidates of Type II young radio supernovae (RSNe; Neff et al. 2004). Four of these five radio sources have flat or inverted radio spectra between 2.2 and 8.4 GHz, very similar to RS1 and RS2 near N1 in NGC 6240. The luminous H₂O maser emission in Arp 299 is found toward both nuclei A and B1 (Henkel et al. 2005) and at A it coincides with the continuum peak at VLA resolution (Tarchi et al. 2010). While this spatial coincidence may also suggest an AGN association (e.g., Hagiwara 2007), the maser in Arp 299 is blueshifted relative to the systemic velocity and is likely associated with an outflow in a star-forming environment.

In order to understand the observed spectra of the southern and northern nuclei, model fitting may be used to explain the frequency turnover at lower frequencies using a power-law plus pure FFA or SSA (e.g., Kameno et al. 2000). First, the FFA model for a nuclear spectrum is described by

$$\begin{equation} S_\nu =S_{\rm 0}(\nu /1.0)^{\alpha _0} \exp \lbrace {-\tau _{\rm ff}(\nu /1.0)^{-2.1}\rbrace }, \end{equation} \tag{ 1 }$$

where ν is the frequency in GHz, S₀ is the unobscured synchrotron flux density in mJy, α₀ is the optically thin non-thermal spectral index, and τ_ff is the opacity at 1.0 GHz. This model describes all FFA in the foreground of the synchrotron emission source and assumes that all source components are subject to the same foreground opacity (e.g., Parra et al. 2007). The second model of pure SSA is described as follows:

$$\begin{equation} S_\nu =S_{\rm 0}\nu ^{2.5} [1-\exp \lbrace {-\tau _{s}\nu ^{\alpha _0-2.5}\rbrace }], \end{equation} \tag{ 2 }$$

where ν is the frequency in GHz, S₀τ_s is close to the flux density if the SSA coefficient τ_s ≪ 1, and α₀ is the optically thin non-thermal spectral index.

The parameters of the FFA and SSA spectral fits for N1 and N2 and the reduced χ² values (per degree of freedom) have been summarized in Table 7, and the fitted curves are shown in Figure 5. Both the FFA and SSA models produce reliable fits for N1 with low values (< 1.5) for the reduced χ² with the FFA model being slightly worse but do not make good fits for N2. However, since there is no compact source in N1 or N2 having a high brightness temperature, any SSA in these nuclei is unlikely (Kellermann & Pauliny-Toth 1969). Using conventional energy equipartition arguments at lower radio frequencies, a turnover frequency of 2.0–3.0 GHz for a self-absorbed source requires an equipartition brightness temperature of ∼10^11.95 K (Readhead 1994). The observed brightness temperatures for our sources are only ∼10⁶–10⁷ K. Better constraints of the nuclear properties of N1 in terms of model fitting require more complete flux density measurements.

While for N2 neither the pure FFA nor the pure SSA model can perfectly explain the spectral bending at 2–3 GHz (see Table 7 and Figure 5), the observed bending does indicate that free–free self-absorption by ionized foreground gas in a starburst environment is relevant. Since there is also no hint of self-absorption in the MERLIN spectrum (Beswick et al. 2001), the size of such an absorbing medium must be very small and between about 9.5 and 25 pc in linear size (Gallimore & Beswick 2004). Similarly, the presence of an AGN in N2 cannot be confirmed from our EVN data, except that the core-jet-like structure, as also seen in the 1.7 and 2.4 GHz VLBA images, could support the presence of an active nucleus.

The radio source RS1 has a radio power at 8.4 GHz of 4.1 × 10²¹ W Hz⁻¹ and a spectral index of α_5.0–8.4 ≈ 0.34 in our Epoch 2 observations. The radio power of RS2 at 5.0 GHz (α_1.6–5.0 ≈ −0.91) is 7.5 × 10²⁰ W Hz⁻¹. The radio power of RS2 is comparable to those of the VLBI sources associated with the nucleus A in Arp 299 while the power of RS1 is nearly a factor of 10 higher, which makes it 1250 times the 8.4 GHz radio power of Cas A of 6 × 10¹⁷ W Hz⁻¹ (Colbert et al. 1994). Both RS1 and RS2 display a radio light curve with a long-term rise (Figure 6). The upper limit for the source size of RS1 at the highest frequency 8.4 GHz is 1.5 × 0.6 pc (Table 5) and the source is not sufficiently resolved by our EVN synthesized beam. Also, different EVN beams resulting from different (u,v) coverage at each epoch make it difficult to make a reliable comparison of the intrinsic source sizes.

Zoom In Zoom Out Reset image size

Speculation that RS1 could be the true southern nucleus, instead of N1, may be ruled out because RS1 was not detected in the 8.4 GHz VLBA observations (1999–2001) and only appeared in 2003. Also its location is significantly offset from the positions of N1 in the 2.4 and 1.7 GHz VLBA data. Similarly, the possibility that RS1 represents ejecta from the active nucleus N1 may be ruled out because its offset (11 pc) from N1 remained unchanged for six years and because RS1 is far from the known outflow structures seen in H i and OH southwest of N1 (Baan et al. 2007). The positional difference of 10 mas (≈50 pc; see Table 3) between RS2 at 1.6 and 5.0 GHz and the source S1 at 1.7 and 2.3 GHz in VLBA observations at epochs in 2000 and 2001 suggests that they are also different sources.

The radio sources RS1 and RS2 have remained detectable for more than six years since 2003. An important question is whether these sources are SNe (interacting with the circumstellar matter) or supernova remnants (SNRs, interacting with the dense interstellar medium). There is no evidence for expanding SN shells with a highest beam size of ∼6 mas (∼3 pc) in either RS1 or RS2 between the two epochs from our data (see Table 5). The range of radio luminosity of 10^20.8–10^21.6 W Hz⁻¹ of RS1 and RS2 is similar to the radio luminosity obtained for observed Type Ib/c or Type II RSNe (Weiler et al. 2002). Thus, it is possible that RS1 and RS2 are SNe or SN–SNR transition objects. Our data are insufficient to distinguish between the two cases and further sensitive VLBI monitoring is required. The VLBA data (1999–2001) also display two similar but weaker RSN or SNR candidates RS3 and RS4, the latter of which has a spectral index of α_1.7–2.4 ≈ 0.35.

Many nearby galaxies, such as the LIRG M 82 and ULIRGs Arp 220 and Arp 299, display compact radio sources that are evidence of ongoing (circum)nuclear star formation (Smith et al. 1998; McDonald et al. 2001; Neff et al. 2004; Parra et al. 2007; Conway et al. 2007; Ulvestad 2009). VLBI observations of the merging galaxy Arp 220 have detected a total of 18 compact radio sources within the western nucleus of the galaxy, over half of which have radio properties that are consistent with Type II SNe interacting with the surrounding medium (Parra et al. 2007). Likewise, the nuclei in Arp 299 show 30 compact radio sources, 25 of which are associated with the northeastern nucleus A (with an AGN candidate) and are spread over a region of 30 pc (Neff et al. 2004; Ulvestad 2009).

A comparison of the SN–SNR sources in NGC 6240 with those found in other sources shows that the radio powers of RS1 and RS2 are equivalent to those of the most powerful sources found in M 82 (see Fenech et al. 2008) although they are less powerful than those in Arp 220. As a result the upper limits for the sizes of RS1 and RS2 lie in the upper range of the general relation between radio luminosity and diameter observed for Galactic and extragalactic SNR sources (Berezhko & Völk 2004; Batejat et al. 2010). This would suggest that the environmental conditions in the LIRG NGC 6240, and possibly the star formation initial mass function, cannot yet be distinguished from those of the most luminous FIR galaxies.