Gaia GraL: Gaia DR2 Gravitational Lens Systems. VIII. A radio census of lensed systems

Observations were carried out using the 6A and 6B array configurations (maximum baseline 6 km) with two 2048 MHz bands centered on 5.5 and 9 GHz respectively. This resulted in a synthesised beam with a minor axis of 1.5" and a major axis that is declination-dependent and typically 3" (based on -30 deg declination) at 9 GHz. We used the ATCA primary calibrator, PKS 1934–638, to calibrate the bandpass and flux density scale, while appropriate gain calibrators for each source were selected from the ATCA Calibrator Database²²2https://www.narrabri.atnf.csiro.au/calibrators/calibrator_database.html.

Data were calibrated using standard Miriad routines (Sault et al., 1995), with automatic flagging carried out using pgflag. Further manual flagging was carried out using blflag where necessary (e.g. particularly bad channels that were not completely flagged with pgflag). Observations of the lenses were then imaged using clean, with robust weighting ($R=0.5$) and a stopping threshold corresponding to approximately 10 times the thermal noise as calculated by invert (typically $\lesssim 20\,\mu$Jy).

Observations were carried out in the A configuration using the C-band receiver with one 4096 MHz band centered on 6 GHz, corresponding to a typical angular resolution of $0.3\,\arcsec$. To calibrate the bandpass and flux density scale we used observations of 3C286 for targets with right ascensions between 10h and 23h, and 3C147 for the remaining targets. Gain calibrators for each source were selected from the list of VLA Calibrators³³3https://science.nrao.edu/facilities/vla/observing/callist. Observations were calibrated using the standard VLA pipeline and then imaged in CASA using tclean with a $50\,\mu$Jy/beam threshold for all sources and robust weighting ($R=0.5$). The resulting images were then corrected for the primary beam using pbcor. We used Aegean (Hancock et al., 2012, 2018) with default settings to find the radio sources in each image. We then used CASA imfit to fit a Gaussian to all sources in the vicinity of the lens, constraining the centre to within a 10 pixel radius of the Aegean source position.

This quadruply imaged broad absorption line (BAL) quasar at $z=2.424$ was independently discovered by Delchambre et al. (2019) and Lemon et al. (2019), with spectra first published in Stern et al. (2021). Only three images are present in Gaia DR3 but four images are clearly visible in archival PanSTARRS data and in the Hubble Space Telescope (HST) images published by Shajib et al. (2019). Our VLA data reveals five components, among which four are roughly compatible with the optical positions, and the fifth one is located between the closest pair of images (labelled $A$ and $B$ in Figure 1). The small mismatch ($\sim 0.07\arcsec$) between the optical and radio emission of $A$ and $B$, and the evidence of a bright fifth component between those images suggest that the background radio source is partly extended and lies very close to the caustic. This extended structure requires higher resolution VLBI observations to be confirmed but is not uncommon among BALs (e.g. Bruni et al., 2013; Kunert-Bajraszewska et al., 2015).

This sub-arcsecond separation system has been proposed to be a doubly imaged lensed quasar candidate ($z=2.355$) by Krone-Martins et al. (2019) but as nearly identical quasars (i.e. a system without sufficiently good spectra to be classified as a binary quasar or a lensed system) by Lemon et al. (2023). The Gaia DR2 positions and flux ratios are consistent with those observed in the radio. While the lensing galaxy has not yet been detected, the consistency between the radio and optical data (in both astrometry and flux scale) support the lensing hypothesis.

This source was identified as a lensed quasar candidate based on Gaia DR2 where three point-source components were identified. Stern et al. (2021) presented the spectra of two lensed images, demonstrating that the source is a quasar at redshift $z_{s}=2.345$. Lemon et al. (2023) spectroscopically confirmed the quad nature of that system. The intervening absorption lines detected at $z>2.1$ are at a too large redshift to be due to the main deflector, which is not detected in existing imaging data. The triplet of images detected by Gaia suggested a fold configuration, as confirmed by the present radio data where four components are clearly detected. A small mismatch of $\sim 0.055\arcsec$ is observed between the radio and Gaia positions and each of the non merging images (i.e. $C$ and $D$), while image $A$ agrees with Gaia position to within $0.02\arcsec$. This is formally larger than the astrometric uncertainty, and could be an indication for an offset between the optical and radio structures such as a core-shift (e.g. Kovalev et al., 2017; Pashchenko et al., 2020). Another possible explanation is the presence of a small separation dual AGN (e.g. Chen et al., 2023). VLBI imaging across multiple frequencies may help to disentangle the two scenarios.

This source is a wide separation quad ($\Delta\theta=5.2\arcsec$) at redshift $z_{s}=3.083$ lensed by an early-type galaxy at $z_{l}=0.766$ (Stern et al., 2021; Lemon et al., 2023). While only three lensed image positions were listed in Gaia DR2, four components are clearly apparent in PanSTARRS images. Four components are also detected with the VLA, with positions in agreement with those from Gaia, where available. Despite the relatively low radio luminosity (which classifies this quasar as “radio-quiet”) the X-ray luminosity is extremely high (Connor et al., 2022).

The faintest radio image (image $D$ in Fig. 1) is part of the "cusp triplet". While it should approximately be about half as bright as the brightest lensed image (image $A$) based on optical imaging, it is instead a quarter of its brightness. The optical-NIR HST frames published by Schmidt et al. (2023) also show that image $D$ is the faintest image at any wavelength, but the flux ratio with image $A$ is never as extreme as in the radio. The lens model proposed by Schmidt et al. (2023) includes a small galaxy visible in the vicinity of this image, and predicts flux ratios in agreement with our radio data. However, the chromatic difference between radio and optical wavelengths may not be explained by the macro model, and suggests the presence of microlensing and/or differential extinction in the optical range.

Only the three brightest components of this large separation ($\delta\theta=10.1\arcsec$) quadruply imaged system, identified in Stern et al. (2021) as a quasar at $z_{s}=1.451$ lensed by a galaxy at $z_{l}=0.591$, are detected in the VLA data. The undetected fourth image is also the faintest one at optical wavelengths. The optical and radio positions do not match perfectly, with offsets ranging between $0.04\arcsec$ and $0.1\arcsec$. This is substantially larger than the astrometric uncertainty, even if considering relative image separation instead of absolute ones. Similar to GRAL J060841.42+422936.87, the origin of this offset could be a core-shift or a dual AGN (e.g. Kovalev et al., 2017; Pashchenko et al., 2020; Chen et al., 2023), but high resolution VLBI imaging would be required to confirm either.

This source was presented as a candidate quadruply imaged quasar by Delchambre et al. (2019), and later confirmed to be a quasar at $z=3.074$ lensed by an edge-on spiral galaxy (Lemon et al., 2019; Stern et al., 2021). While only three lensed images are detected in Gaia data, a fourth faint image is visible in ground-based and HST data (Rusu & Lemon, 2018; Lemon et al., 2019; Schmidt et al., 2023). This fourth image appears to be strongly reddened at optical wavelengths due to the lensing galaxy, but the optical and radio astrometry agree.

The observed radio flux ratios disagree with the HST-based lens model of Schmidt et al. (2023). This may be caused by the presence of a disc-like component in the lensing galaxy which is not included in the model of Schmidt et al. (2023). Such a disc-like component, already proposed for this system in (Rusu & Lemon, 2018), is known to produce flux ratio perturbations in other lensed systems (Möller et al., 2003; Hsueh et al., 2018). The presence of a disc in the lens is therefore a good candidate for explaining the radio fluxes. Alternatively, the background source may be intrinsically variable at radio wavelengths. In addition, Connor et al. (2022) note that the X-ray emission is fainter than expected. Further observations are required to distinguish between these competing hypotheses.

This system is a compact ($\Delta\theta\sim 0.7\arcsec$) quadruply imaged BAL quasar at $z_{s}=1.724$ (Stern et al., 2021). Only three of the quasar images were detected in Gaia DR2. The radio data identifies also three components, with one of them (labelled $C$ in Figure 1) slightly extended. A qualitative comparison of the optical and radio data suggests that the extended radio feature could be a blend of two components, i.e. the two images completing the quad in a fold-configuration system. While a similar configuration had been inferred from the modelling of the optical data in Stern et al. (2021), the fourth image position, predicted by the lens model, does not match the interpretation of the $C$ image as two blended components. While this is not unexpected (Luhtaru et al., 2021), the absence of a Gaia detection is slightly surprising, as the "missing optical image" corresponds to the second brightest radio-component and not the faintest one⁴⁴4We note that the Gaia completeness decreases for small image separations, so it is possible that the missing image is not necessarily the faintest.. While intrinsic variability could explain the lack of detection in the optical wavelength range, it could also be due to a radio structure that does not coincide with the accretion disc. If present, this radio structure is likely due to the BAL nature of this quasar – an extended structure is inferred in another BAL in this sample (GRAL J024848742+191330571). Such extended radio emission is often detected in high resolution radio data of BALs (e.g. Bruni et al., 2013; Kunert-Bajraszewska et al., 2015). The small apparent mismatch between the absolute optical and radio position may further support this interpretation, although tentative, owing to the potentially larger error on the absolute astrometry.

This system is a doubly-imaged lensed quasar at $z_{s}=1.604$ discovered by Krone-Martins et al. (2019). While the lens has not been directly detected, narrow absorption features at $z_{l}=0.855$ provide a plausible lens redshift. The radio and optical positions are in excellent agreement, while the flux ratios differ by more than a factor of two. An optical flux ratio of 0.778 is measured, while a ratio of 0.306 is measured in the radio. This might be explained by intrinsic variability of the source, by microlensing affecting the optical wavelengths, or by dust extinction. The difference of color of the two images in the optical data suggests that microlensing and/or dust reddening may play a role.

Several systems show either detection of only one lensed image, or very faint detections. We briefly discuss those systems hereafter:

• –

  GRAL J022958.17+032032.1: This system is a doubly imaged lensed quasar candidate with two components separated by $2.14\arcsec$. The astrometry of the two radio images is compatible with that of Gaia DR2, but remains tentative as the radio flux is only marginally above the noise.

• –

  GRAL J125955.5+124152.4: The lensed nature of this wide separation ($\Delta\theta=3.5\arcsec$) system (Hennawi et al., 2006; Foreman et al., 2009; Krone-Martins et al., 2019) composed of two quasar images at $z_{s}=2.196$ is not certain. Radio emission compatible with the optical positions is tentatively detected, but at low significance.

• –

  GRAL J210328980-085049486: Only one radio component is identified for this compact ($\Delta\theta<1.0\arcsec$) quadruply imaged quasar at $z_{s}=2.446$. The radio emission matches the brightest of the triplet of point-source emission detected in Gaia DR2 (Ducourant et al., 2018; Stern et al., 2021). The non detection of the other two components is compatible with their expected fainter flux based on the observed optical flux ratios. The strong elongation of the radio beam does not enable us to test whether the radio component results from the merging of two point-sources, and/or is due exclusively to point-like radio emission.

The VLA images positions can be reproduced by a SIE+shear model centroid compatible with the optical position (within 0.02″  astrometric uncertainty). The best model is found for an unrealistically large ellipticity ($e=0.54$) and external shear ($\gamma_{\rm ext}=0.29$). An exploration of the 7-dimensional parameter space reveals that two other local minima characterised by a slightly higher reduced $\chi^{2}$ (but lower than 1) exist. They both correspond to rounder mass distribution and lower shear. If we set a prior on the alignment between the lens and light mass such that $|\phi_{\rm light}-\phi_{\rm mass}|<20\deg$ (as suggested by numerous lensing studies, e.g. Sluse et al., 2012; Bruderer et al., 2016), we find a slightly rounder model yet characterised by a very large shear amplitude. Our lens model parameters are compatible with those obtained by Schmidt et al. (2023) using HST imaging. The small axis ratio $q=b/a\sim 0.5$ of the mass distribution may be physical as the lens light also displays a low value of $q=0.44$. The large inferred shear may be an artefact caused by choice of a purely elliptical mass distribution. A discy (and/or boxy) component is likely to be present in the galaxy, but its absence from the model may be captured by the shear term (Van de Vyvere et al., 2022a; Etherington et al., 2023). Other types of angular structures can also potentially play a role (Van de Vyvere et al., 2022b). Nevertheless, we do not exclude that a fraction of this large shear is effectively related to the lens environment, and produced by the bright galaxies visible in optical images within $<1\arcmin$ from the lens.

For this system we have simplified our model due to the large uncertainty on the position of the fourth radio lensed image. We have modelled the image positions using an SIE (i.e. no added shear), leaving the position of the lensing galaxy as a free parameter. The new lens model is substantially different from the one inferred from the triplet of Gaia images by Stern et al. (2021). The main difference occurs for the predicted position of the lensing galaxy which differs by 0.3″  compared to the previous model. Such a change is not unexpected as a triplet of images does not fully constrain a simple SIE (Luhtaru et al., 2021). The predicted flux ratio between the pair of fold images is close to 1:1 as expected asymptotically for fold images. The radio measurement suggests a 4:3 ratio, but is also consistent with the expected 1:1 ratio within uncertainties.

Our simple SIE+shear model successfully fits the radio positions within the uncertainties. The derived shape of the mass distribution of the lensing galaxy however deviates substantially from the shape of the luminosity profile ($e,\phi_{L}$) = (0.95, -27.8$\,\deg$) derived by Schmidt et al. (2023). Our model requires a flatter profile ($e=0.45$), with a major axis position angle offset by about $30\,\deg$ compared to the light. The likely reason for this discrepancy is the object detected on HST images in the vicinity of image $D$, and which is not included in our simple model. The model proposed by Schmidt et al. (2023), which is constrained by extended images of the quasar host and accounts for the companion, yields a better agreement between the shapes of the lens light and mass profiles.

We find that a simple SIE+shear model of this system is able to reproduce the VLA image positions within the uncertainties. The modelled position of the lens centroid is offset from the known optical position by $\sim 0.1\arcsec$ (Schmidt et al., 2023). This small discrepancy is not unexpected given the uncertainties on the relative astrometry, the presence of dust in the optical regime, and the possibility that the radio and optical emission do not share a common origin.

The derived mass distribution is very flattened ($q=0.42$) but rounder than the light ($q_{L}=0.26$). The position angle of the light and mass also seem to agree reasonably well, within about 10 degrees. While such a mass distribution is not at odds with the spiral nature of the lens, we note that the inferred shear is large and almost aligned with the position angle of the mass ($\Delta\phi=6.5$ deg). As in the case of GRAL J024848.7+191330, this may be caused by “discyness”, “boxyness”, or ellipticity gradients in the true mass distribution but not accounted for by the model (Van de Vyvere et al., 2022a; Van de Vyvere et al., 2022b; Etherington et al., 2023).

Our modelling yields an Einstein radius consistent with that measured by Schmidt et al. (2023), but the amplitude and orientation of the ellipticity and shear differ between the two models. The origin of this discrepancy is not clear, but may be related to the use of a prior on the relative orientation of the major axis of the light and mass distribution of the lens by Schmidt et al. (2023).