Outflows in the Radio-Intermediate Quasar III Zw 2: A Polarization Study with the EVLA & uGMRT

Firstly, the model of a polarized calibrator was set manually using the task $\tt{SETJY}$ in $\tt{CASA}$. For this, we provided the model parameters such as the reference frequency, Stokes I flux density at the reference frequency, the spectral index, and the coefficients of the Taylor expansion of fractional polarization and polarization angle as a function of frequency about the reference frequency. The coefficients were estimated by fitting a first-order polynomial to the values of fractional polarization and polarization angle as function of frequency, which were obtained from the NRAO VLA observing guide³³3https://science.nrao.edu/facilities/vla/docs/manuals/obsguide/modes/pol.

Polarization calibration was carried out in three steps: (i) the cross$-$hand (RL, LR) delays, resulting from residual delay difference between R and L signals, were solved using a polarized calibrator that has strong cross-polarization. This was carried out using the task $\tt{GAINCAL}$ with $\tt{gaintype=KCROSS}$ in $\tt{CASA}$. (ii) the instrumental polarization (i.e, the frequency dependent leakage terms or ‘D-terms’), was solved using either an unpolarized calibrator or a polarized calibrator with good parallactic angle coverage. This corrected for the polarization leakage between the feeds owing to their imperfections and non-orthogonality (for e.g. the R/X$-$polarized feed picks up L/Y-polarized emission, and vice versa). We used the task $\tt{POLCAL}$ in $\tt{CASA}$ to solve for instrumental polarization with $\tt{poltype=Df}$ while using the unpolarized calibrator and with $\tt{poltype=Df+QU}$ while using the polarized calibrator. This task uses the Stokes I, Q, and U values (Q and U being zero for unpolarized calibrator) in the model data to derive the leakage solutions. Also, when $\tt{poltype=Df+QU}$ is used, this task solves for both source polarization and instrumental polarization simultaneously based on their differential rotation with the parallactic angle.

The unpolarized calibrator 3C84 was used for leakage calibration in all the projects. The average value of the D-term amplitude turned out to be $\sim 10\%$ (typically $\leq 20\%$) for the uGMRT antennas (Project DDTC130). For the VLA antennas, this value turned out to be $\sim 7\%$ for Project 20A-182 and $\sim 5\%$ for Project 17A-027 (typically $\leq 10\%$ in both cases). (iii) the frequency-dependent polarization angle was solved using a polarized calibrator with known polarization angle. This corrected for the R-L phase offset which arises from the difference between the right and left gain phases for the reference antenna. This was carried out using the task $\tt{POLCAL}$ in $\tt{CASA}$ with $\tt{poltype=Xf}$.

The calibration solutions were then applied to the multi-source dataset. The $\tt{CASA}$ task $\tt{SPLIT}$ was used to extract the calibrated visibility data for III Zw 2 from the multi-source dataset while averaging the spectral channels such that the bandwidth smearing effects were negligible. The total intensity or the Stokes I image of III Zw 2 was created using the multiterm-multifrequency synthesis (MT-MFS; Rau & Cornwell, 2011) algorithm of $\tt{TCLEAN}$ task in $\tt{CASA}$. Three rounds of phase-only self-calibration followed by one round of amplitude and phase self-calibration were carried out for the VLA data whereas, four rounds of phase-only self-calibration followed by four rounds of amplitude and phase self-calibration were carried out for the uGMRT data. Stokes Q and U images were created from the last self-calibrated visibility data, using the same input parameters as for the Stokes I image but with lesser number of iterations. The rms noise of the Stokes (I, Q, U) images is (19, 10, 11) $\mu$Jy beam^-1 for uGMRT at 685 MHz, (66, none, none) $\mu$Jy beam^-1 for VLA at 1.5 GHz, (15, 13, 13) $\mu$Jy beam^-1 for VLA at 5 GHz and (278, 47, 48) $\mu$Jy beam^-1 for VLA at 34 GHz.⁴⁴4Stokes Q and U images appear to have been overcleaned at 34 GHz compared to Stokes I image. However, potential discrepancies have been accounted for by blanking pixels with Stokes Q or U signal-to-noise ratio less than 3, greater than 10$\degr$ error, and greater than 10% error, while creating the linear polarization, polarization angle, and fractional polarization images, respectively.

Linear polarized intensity ($P=\sqrt{Q^{2}+U^{2}}$) and polarization angle ($\chi$ = 0.5 tan${}^{-1}(U/Q)$) images were obtained by combining the Stokes Q and U images using the $\tt{AIPS}$⁵⁵5Astronomical Image Processing System; Wells (1985) task $\tt{COMB}$ with $\tt{opcode=POLC}$ (which corrects for Ricean bias) and $\tt{POLA}$ respectively. Regions with intensity values less than 3 times the rms noise in the $P$ image and with values greater than 10$\degr$ error in the $\chi$ image were blanked while using $\tt{COMB}$. Fractional polarization ($F=P/I$) images were obtained in $\tt{COMB}$ with $\tt{opcode=DIV}$. Regions with fractional polarization errors $\gtrsim$10% were blanked in the image.

The Stokes Q and U images for the VLA 5 GHz data were created from two sub-bands with a bandwidth of 1024 MHz each, obtained from the original data having 16 spectral windows (spw) with a bandwidth of 128 MHz each, using the task $\tt{TCLEAN}$ in $\tt{CASA}$. These were combined to create the polarization angle ($\tt{PANG}$) images using the $\tt{AIPS}$ task $\tt{COMB}$. The mean polarization angle of the “core” was estimated from the VLA sub-band $\tt{PANG}$ images as well as the VLA 34 GHz $\tt{PANG}$ image using the task $\tt{IMSTAT}$ in $\tt{AIPS}$. The rotation measure (RM) of the core was estimated by fitting the relation: $\chi$ = $\chi_{\circ}$ + RM $\lambda^{2}$ ($\chi_{\circ}$ and $\chi$ being the intrinsic and observed polarization angles respectively) to the core polarization angles at the 3 frequencies: 5 GHz, 6 GHz and 34 GHz (the former two being the central frequencies of the VLA sub-band datasets), using $\tt{PYTHON}$ fitting procedure - $\tt{CURVE\_FIT}$. The uGMRT core polarization angle was excluded from this analysis since the core polarization structure in the uGMRT image was not clearly resolved from the jet, unlike the VLA images. We note that the calibrated data was not divided into four sub-bands since the signal-to-noise ratio in the individual sub-bands was too low for a reliable RM estimation.

In order to estimate the RM for the south-western (SW) lobe, we created another set of $\tt{PANG}$ images from the VLA sub-band datasets using the Stokes Q and U images that were convolved with an identical circular beam corresponding to the uGMRT resolution (4$\arcsec~{}\times$ 4$\arcsec$). We also created the uGMRT $\tt{PANG}$ image from the Stokes Q and U images convolved with this beam. The RM for the lobe was estimated by fitting the $\chi-\lambda^{2}$ relation to the mean polarization angles of the lobe at the 3 frequencies: 685 MHz, 5 GHz and 6 GHz using the $\tt{CURVE\_FIT}$ procedure.

Stokes I images of VLA 1.5 GHz and uGMRT 685 MHz data, convolved with an identical circular beam (6.5$\arcsec~{}\times$ 6.5$\arcsec$; which is the beam minor of the uv -tapered uGMRT image; see Section 3.1 ahead), were created using the task $\tt{TCLEAN}$ in $\tt{CASA}$. These images were made spatially coincident using the task $\tt{OGEOM}$ in $\tt{AIPS}$, and combined using $\tt{opcode=SPIX}$ in $\tt{AIPS}$ task $\tt{COMB}$. A ‘two-point’ spectral index image was created using the relation: $\alpha$ = log ($\mathrm{S}_{1}$/$\mathrm{S}_{2}$)/log($\nu_{1}/\nu_{2}$) where $\mathrm{S}_{1}$ and $\mathrm{S}_{2}$ are flux densities at frequencies $\nu_{1}$ and $\nu_{2}$ (here $\nu_{1}$=685 MHz and $\nu_{2}$=1.5 GHz). Regions with flux densities below 3 times the rms noise were blanked while using $\tt{COMB}$.

A triple radio source with a “hotspot"-core-“hotspot" structure and a curved jet, along with a misaligned lobe, are observed in III Zw 2. The jet in III Zw 2 does not terminate in compact hotspots similar to those observed in FRII radio galaxies but rather in “bow-shock-like” radio features (see Section 4.1 ahead). Figure 1 presents the uv-tapered uGMRT 685 MHz total intensity contour image of III Zw 2 superimposed over Sloan Digital Sky Survey (SDSS; Ahumada et al., 2020) g-band optical image, with an inset on the top right presenting the VLA 5 GHz total intensity contour image superimposed over the HST WFC3 F547M optical image of the SW jet and lobe displaying the bow-shock-like morphology of the “hotspot” region. The final resolution of the uv-tapered uGMRT image is 7.0$\arcsec~{}\times$ 6.5$\arcsec$ with a PA = 88$\degr$. The uv-tapering was carried out to detect the diffuse lobe emission, especially in the misaligned lobe, better.

As evident from the figure, the host galaxy is an interacting galaxy and part of an interconnected triplet of compact galaxies. While we cannot rule out contributions to the radio flux density from the companion galaxies seen in projection along the misaligned lobe, we find that the brightest part of diffuse radio emission in the misaligned lobe is not spatially coincident with the second largest of the companion galaxies to the south. Moreover, both the radio flux density and the spectral index of the misaligned lobe provide no clear indications that it has additional contribution from companion galaxies.

Figures 2, 3 and 4 present total intensity contour images of III Zw 2 with electric fractional polarization vectors superimposed in red using uGMRT at 685 MHz, VLA at 5 GHz in B-array and 34 GHz in C-array respectively. The left panel of Figure 5 presents a zoom-out of the inset of Figure 1, showing the full radio structure of III Zw 2, including the emission from the north-eastern (NE) lobe. From the figure, it appears that the tidal arm towards the north in the optical image may have interacted with the radio jet, but failed to disrupt it. The white curve superimposed over the SW lobe represents the bow-shock-like feature at the jet termination point.

Table  2 provides the values of total flux density, polarized flux density, fractional polarization and polarization angle for different radio components of III Zw 2 from different projects. The errors in the total flux density of the cores were estimated from the calibration uncertainties (assuming $\sim$10%) and rms noise of the images. Flux density values for compact components like the core were estimated using the Gaussian-fitting $\tt{AIPS}$ task $\tt{JMFIT}$, whereas for extended emission, the $\tt{AIPS}$ tasks $\tt{TVWIN}$ and $\tt{IMSTAT}$ were used. The rms noise values were also obtained using $\tt{TVWIN}$ and $\tt{IMSTAT}$. The polarized flux density, fractional polarization and polarization angle values were estimated as the mean values for different components from the polarized intensity, fractional polarization and polarization angle images respectively, using the task $\tt{IMSTAT}$ or $\tt{TVSTAT}$ in $\tt{AIPS}$. The respective errors were estimated from the same regions of the corresponding noise images. All the lengths and widths used in the calculations ahead, were measured using the task $\tt{TVDIST}$ in $\tt{AIPS}$.

The distance from the core to “hotspot” in the SW lobe is $\sim$ 24 kpc and to the “hotspot” in the NE lobe is $\sim$ 33 kpc. The jet is one-sided which could be explained as a result of the Doppler boosting and dimming effects; the brighter SW jet is the approaching one. Also, the jets are asymmetric in length, which could in principle arise from an asymmetry in the environment around the lobes. The medium to the south-west of III Zw 2 is denser than the medium to the north-east; the closest neighbouring galaxy to the south-west of III Zw 2 is only $\sim$30 arcsec away (see Figure 1; see also Brunthaler et al., 2005). We also detect prominently a diffuse lobe to the south-east that is misaligned with the primary lobes (see Figures 1 and  2). In the 34 GHz VLA image, we detect only the radio core.

The misaligned lobe in III Zw 2 was briefly discussed by Silpa et al. (2020), where we had presented a shorter exposure uGMRT image at 685 MHz, obtained on December 7, 2018. In that work we had obtained an integrated flux density in the radio core of $\sim$60 mJy and in the current work, it has dropped to $\sim$40 mJy. The $\approx$30% drop in the core flux density lies within the known observed range in the literature. The integrated flux density of the SW and NE lobes agree within errors between the two epochs. The spectral index image of III Zw 2 created using our new uGMRT data and similar resolution 1.5 GHz VLA data, indicates a mean spectral index of $-0.7$ and $-0.6$ in the SW and misaligned lobes, respectively (Figure 6).

The spectral index image shows a flat spectrum region in the misaligned lobe. This however is not spatially coincident with the southernmost companion galaxy, with the offset being $-6.4$ arcsec in the north and $+1.5$ arcsec in the east. We also find an ultra-steep ($\alpha<-$1) region to the south of the SW primary lobe. This may be plasma left behind by the jet head that has subsequently aged. Some of this lobe material may also be pushed towards a region of lower ambient density by the bouyant forces from the medium surrounding the SW lobe, giving rise to the misaligned lobe (see Sections 3.2.2 and  4.1 ahead).

The curved kpc-scale jet terminates in a bow-shock-like region in the SW lobe (see Figure 5). A similar structure is also observed at the end of the eastern lobe. A compact terminal hotspot is not observed in III Zw 2. The inferred B-fields are aligned with the lobe edges. The shape and magnetic field structure resemble that observed at the end of the eastern jet in the episodic radio galaxy, Pictor A (Perley et al., 1997; Saxton et al., 2002; Tingay et al., 2008), but without a preceding hotspot in III Zw 2. In fact the shape and magnetic field structures observed at the lobe edge in III Zw 2 are similar to what is observed at the lobe edges of Seyfert galaxies (Kharb et al., 2006; Sebastian et al., 2019b, 2020).

This radio morphology is also observed in Fig. 1e in the restarted jet simulation work of Clarke & Burns (1991). These authors find that when a restarted jet is launched into a region of thermalized plasma from the previous episode of jet activity, and therefore is overdense compared to the old jet material through which it passes, a protrusion is seen at the leading edge of the lobe (their Fig. 1e), similar to the feature observed in III Zw 2. The morphology of the SW lobe in III Zw 2 may therefore be indicative of restarted jet activity, which is also supported by the evidence of intermittent jet activity on pc-scales (Brunthaler et al., 2003; Brunthaler et al., 2005; Li et al., 2010). It is possible that the intermittency has a shorter duty cycle compared to RL AGN. The terminal lobe region could also indicate a sharp jet bend which may have resulted when the jet encountered a density inhomogeneity in the intergalactic medium. However, this cannot explain why a similar structure is observed on the counter-jet side.

Similar spectral indices and electron lifetimes for both SW and misaligned lobes (see Figure 6 and Table 3) suggest that the misaligned emission is not due to a previous AGN activity episode. It appears that the outflow, which is likely a combination of a jet and a wind (see Section 4.2 ahead), moves away from the jet termination region into a lower density region, that is, in the direction of least resistance. This would be consistent with our finding that $\rho_{a}$ for the misaligned lobe is lower than for the SW lobe (see Section 3.2.2).

The primary and the misaligned radio lobes of III Zw 2 exhibit a mean spectral index between $-$0.6 and $-$0.7 and no clear spectral steepening with distance from the core. These lobe characteristics are similar to those revealed in the “sputtering” AGN, NGC 3998 (Sridhar et al., 2020). These results could suggest that the primary lobes are being actively fed by the nucleus by intermittent fuelling and that the plasma in the misaligned lobe is being constantly energized and re-accelerated in the presence of turbulence, instabilities and shocks arising from galactic merging activity.

The transverse B-fields in the core are suggestive of a toroidal B-field component at the base of the outflow, which continues all the way up to the edge of the SW lobe as seen in the uGMRT 685 MHz image. The transverse B-fields in the uGMRT jet could either be interpreted as a series of transverse shocks that amplify and order B-fields by compression as seen in BL Lac jets (Gabuzda et al., 1994; Lister et al., 1998), or could represent a toroidal component of a large-scale helical B-field associated with the jet (Pushkarev et al., 2017). The parallel B-fields in the jet component J1 from the VLA 5 GHz image could represent a poloidal component of the helical jet B-field. Overall, the polarization structures observed in III Zw 2 are consistent with the work of Begelman et al. (1984), where the authors suggest that with increasing distance from the core, the toroidal component of the B-field becomes more dominant and the poloidal component decays faster since the poloidal field varies as r^-2 and the toroidal field varies as r^-1v^-1.

The depolarization of the jet in the 5 GHz VLA image could be explained as an “inverse depolarization” effect. This effect has been previously observed by Homan et al. (2002); Hovatta et al. (2012) in optically thin jet features. This effect could originate from a combination of structural inhomogeneities in the jet B-field that could cancel the polarization along the line of sight and internal Faraday rotation that increases with the square of the wavelength (Sokoloff et al., 1998; Homan, 2012, 2014). Such structural inhomogeneities could naturally arise in helical B-fields and random B-fields that are tangled over long length scales. Internal Faraday rotation could align the polarization from the far side of the jet with that from the near side of the jet. This reduces the cancellation between the vectors and increases the net polarization at longer wavelengths.

However, the works of Mehdipour & Costantini (2019) and Miller et al. (2012) suggest that toroidal B-fields could be associated with magnetically-driven AGN winds and poloidal B-fields could be associated with jets. Therefore, one cannot rule out the possibility of a magnetized accretion disk wind or the outer layers of a broadened jet (like a jet sheath) threaded by toroidal B-fields, on larger spatial scales than sampled by the VLA observations and poloidal B-fields threading the spine of the jet. The inherent complexities in the models, data and numerical simulations make it difficult to differentiate between MHD winds and the outer layers of a broadened jet. A co-axial jet and wind outflow model where the jet is embedded inside the wind, would appear co-spatial in projection. Moreover, such a model would be similar to that revealed in the protostellar system HH 212 comprising of a jet and MHD disk wind (Lee et al., 2021).

In order to confirm if the model of Miller et al. (2012) and Mehdipour & Costantini (2019) is valid for III Zw 2, one would require highly sensitive observations at higher frequency and higher resolution that can substantially sample poloidal B-fields in the spine of the jet (for which we already have a marginal detection), in conjunction with emission-line data that can provide complete information on the multi-phase outflow.

By modelling the X-ray reflection spectrum of III Zw 2 from joint XMM-Newton and NuSTAR observations, Chamani et al. (2020) find that the spin parameter for III Zw 2 is $\geq$0.98. A positive spin parameter suggests a prograde rotation i.e. a co-rotation of the black hole and the accretion disk (Reynolds, 2019); the high value suggests that the black hole is spinning rapidly. Using an AGN subsample from Mehdipour & Costantini (2019) that included III Zw 2, Garofalo (2019) find that high spinning black hole co-rotating with a radiatively efficient thin accretion disk could give rise to a strong wind and a weak jet. Therefore, we infer that a radiatively efficient thin accretion disc may be present in III Zw 2 at the current epoch.

Lohfink et al. (2013) suggest disc truncation and refilling as an explanation for the intermittent jet cycle in 3C120. Moreover, hard and soft spectral states of BHs are often associated with advection-dominated accretion flow (ADAF; Narayan & Yi, 1994) and thin accretion disk (Shakura & Sunyaev, 1973) respectively (Esin et al., 1997; Körding et al., 2006). These could indicate that intermittent jet production is associated with changing spectral states of the accretion disk in microquasars and quasars (Nipoti et al., 2005). A soft spectral state is characterized by a standard thin accretion disk that extends to the last stable orbit. Since the magnetic flux accumulation close to the BH in thin accretion disks is not efficient enough for launching of powerful radio jets (Sikora & Begelman, 2013), this state would produce suppressed radio jets and strong winds. Eventually, the inner regions of the standard accretion disk gets destroyed by some instability and replaced by a radiatively inefficient accretion flow. This hard state is accompanied by efficient magnetic flux accumulation close to the BH, giving rise to powerful and steady radio jets. After this the disk refills again with radiatively efficient accretion flow and the spectral state changes to soft state accompanied by quenched radio jets and this cycle repeats.

Overall, as and when the inner accretion disk region destabilizes, a jet component is ejected. The inner disk region in regular RL and RQ AGN destabilizes on much longer timescales, typically of the order of $10^{6}-10^{8}$ yrs (Alexander & Leahy, 1987; Tucker & David, 1997; Enßlin & Gopal-Krishna, 2001; McNamara et al., 2005; Shabala et al., 2008), than the AGN showing intermittent/sputtering jet activity on a much shorter, decadal timescales. This may suggest that the RQ AGN are the ones in the soft state, hosting strong winds and suppressed jets, and RL AGN are the ones in the hard state, launching powerful jets, at the time of observations.

The outflow in III Zw 2 appears to be a combination of an AGN jet embedded inside a broader wind. In the framework of changing spectral states of the accretion disk, the MHD wind may be launched from the outer regions of the accretion disk whereas the jet may be launched from the inner regions of the disk. Sputtering/intermittent jet activity, typically on decadal timescales (Brunthaler et al., 2005; Li et al., 2010), may suggest that the time scales of transition between the hard and soft states is so rapid that one may always find a combination of an AGN jet of moderate radio luminosity and an accretion disk wind.

The interaction of the jet with the wind plasma could possibly disrupt the jet or, much of the mass and energy may be carried away by the wind, rather than the jet. This could explain why the jets in RI AGN are low-powered and small-scaled, as compared to the RL AGN (see also Chamani et al., 2021). Moreover, the indication of the restarted jet activity and the associated bow-shock-like morphology in the lobes of III Zw 2 (discussed in Section 4.1) may arise because of the launching of the jet during the current epoch of the hard state into a region of thermalized wind plasma released from the previous epoch of the soft state. The wind component may also be the outer layers of a broadened jet (like a jet sheath) from a previous epoch, or an accretion disc wind or a combination of both. Overall, the close interplay between the jet and the wind may in turn explain the RI nature of III Zw 2. Moreover, our results are consistent with the idea that radio-loudness is a function of the epoch at which the source is observed (e.g., Nipoti et al., 2005; Kunert-Bajraszewska et al., 2020; Nyland et al., 2020; Das et al., 2021).