Giant Metrewave Radio Telescope detections of two high-opacity Hi 21 cm absorbers at $z\approx 1.2$

For radio-loud active galactic nuclei (AGNs), studies of “associated” Hi 21 cm absorption by neutral atomic hydrogen (Hi) in the AGN environment provide an interesting probe of physical conditions in the vicinity of AGNs (see Morganti & Oosterloo, 2018, for a recent review). For example, the strength of the Hi 21 cm absorption yields information on the Hi column density $\rm{N_{H{\small\textsc{i}}}}$ and the gas spin temperature ${\rm T_{s}}$, and how these are affected by proximity to the AGN. The Hi 21 cm absorption kinematics, relative to the AGN, can be used to determine whether the Hi is infalling (redshifted) or outflowing (blueshifted), and thus, to infer the importance of AGN fuelling and feedback at different cosmological epochs (e.g. van Gorkom et al., 1989; Vermeulen et al., 2003; Morganti et al., 2003, 2016). Very long baseline interferometric Hi 21 cm absorption studies can be used to track the influence of the AGN jets on the gas, and the driving of high-velocity outflows (e.g. Oosterloo et al., 2000; Morganti et al., 2013). The detection rates of Hi 21 cm absorption in different AGN types can provide information on the evolutionary history of AGNs, and on AGN unification schemes (e.g. Vermeulen et al., 2003). High-opacity Hi 21 cm absorbers are also good candidates for searches for radio molecular absorption (e.g. Wiklind & Combes, 1994; Kanekar et al., 2005; Allison et al., 2019), which can be used to probe fundamental constant evolution on cosmological timescales (e.g. Kanekar, 2011; Kanekar et al., 2018).

The detection of associated Hi 21 cm absorption at high redshifts, $z>1$, would allow one to extend the above studies to the early Universe and track the cosmological evolution of Hi in AGN environments. Unfortunately, despite a number of searches (e.g. Curran et al., 2008; Aditya & Kanekar, 2018a, b; Grasha et al., 2019), this endeavour remains limited by the paucity of known high-$z$ associated Hi 21 cm absorbers. While more than a hundred such absorbers have been found at $z<1$ (e.g. Vermeulen et al., 2003; Geréb et al., 2015; Maccagni et al., 2017; Aditya, 2019), there are only 8 confirmed detections at $z>1$ (Uson et al., 1991; Moore et al., 1999; Ishwara-Chandra et al., 2003; Curran et al., 2013; Aditya et al., 2017; Aditya & Kanekar, 2018a; Dutta et al., 2020). Curran et al. (2008) argued that this apparent redshift evolution in the strength of the associated Hi 21 cm absorption might arise because higher-redshift AGNs searched for Hi 21 cm absorption typically have high ultraviolet (UV) and radio luminosities; however, their results were based on a heterogenous AGN sample. This issue was further examined in an Hi 21 cm absorption survey of a homogenous sample of flat-spectrum AGNs by Aditya & Kanekar (2018b), who found that associated Hi 21 cm absorption is significantly weaker in AGNs at high redshifts, but also in AGNs with high rest-frame radio or ultraviolet (UV) luminosities. Unfortunately, the luminosity bias in their target sample, with high-luminosity AGNs at high redshifts, meant that it was not possible to break the above degeneracy between redshift evolution and AGN luminosity. Searches for associated Hi 21 cm absorption in low-luminosity AGNs at high redshifts offer the best route to break this degeneracy.

Next, searches for associated Hi 21 cm absorption have so far mostly been carried out in AGNs with known redshifts. This introduces a bias against AGN environments with high Hi column densities, as such systems are likely to have high dust columns, which would obscure the AGN in the optical and UV wavebands, making it difficult to measure its redshift. Hi 21 cm or mm-wave absorption surveys that target all radio AGNs independent of redshift information ensure no bias against dusty sightlines (e.g. Wiklind & Combes, 1996; Kanekar et al., 2014a; Allison et al., 2015).

We have used the upgraded Giant Metrewave Radio Telescope (GMRT; Gupta et al., 2017) to carry out such an unbiased search for Hi 21 cm absorption, covering the redshift range $z\approx 0.73-1.53$ against a large sample of radio sources. In this Letter, we report the discovery of two remarkable high-opacity Hi 21 cm absorbers at $z\approx 1.2$, both identified against reddened AGNs, with low rest-frame radio and UV luminosities.

The two new Hi 21 cm absorbers were discovered in our $550-850$ MHz survey of the DEEP2 Survey fields, using the Band-4 receivers of the upgraded GMRT. The DEEP2 Survey used the DEIMOS spectrograph on the Keck II Telescope to accurately measure the spectroscopic redshifts of 38,000 galaxies at $z\approx 0.70-1.45$, in four regions of the sky (Newman et al., 2013). Our GMRT survey used five pointings to cover two of the four DEEP2 field, with the main aim of measuring the average Hi mass of star-forming galaxies at $0.74\leq z\leq 1.45$, by stacking the Hi 21 cm emission spectra of galaxies with DEEP2 spectroscopic redshifts (Chowdhury et al., 2020). However, in addition to the primary science goal, the high spectral resolution of our survey ($\approx 18-25$ km s^-1) allowed us to carry out an unbiased search for Hi 21 cm absorption against all the ($\approx 250$) radio-continuum sources with flux densities $\geq 1$ mJy (without correcting for the shape of the primary beam) in the five target fields, including sources without a known redshift. Our search was sensitive to both associated and intervening Hi 21 cm absorption at $z\approx 0.73-1.53$. This search yielded the initial discovery of two high-opacity absorbers, at $z=1.2166$ towards J0229+0044, and at $z=1.1630$ towards J0229+0053, the focus of this Letter.

We followed up the two absorbers with GMRT Director’s Discretionary Time (DDT) in December 2019 (J0229+0044; proposal DDTC101) and February 2020 (J0229+0053; proposal DDTC121) to confirm the Hi 21 cm absorption, and accurately measure the Hi 21 cm optical depth. The total observing times were 10 hours (J0229+0044) and 12 hours (J0229+0053), using the GMRT Software Backend (GSB) and the GMRT Wideband Backend (GWB), respectively, as the primary correlator to cover the redshifted Hi 21 cm line. The choice of correlator and the correlator settings were based on the aim of simultaneously covering the redshifted Lambda-doubled OH 18 cm lines; however, local oscillator tuning issues in the observations of J0229+0044 led us to a different correlator setup for the observations of J0229+0053. For J0229+0044, the OH 18 cm line frequencies were covered using the GWB as the secondary correlator, with a bandwidth of 200 MHz, centered at 700 MHz and sub-divided into 16,384 channels. Conversely, for J0229+0053, a single GWB 200 MHz band with 16,384 channels was used to cover both the Hi 21 cm line and the OH 18 cm lines. The observational details (related to the Hi 21 cm line) for the two sources are provided in Table 1.

The data were analyzed in the Common Astronomy Software Package (casa version 5; McMullin et al., 2007), following standard procedures. The aoflagger package (Offringa et al., 2012) was additionally used for excision of data affected by radio frequency interference (RFI). After the initial data editing, and gain and bandpass calibration using observations of the compact source 0204+152, a standard self-calibration procedure was carried out for each target source, using line-free channels. The imaging was done with the task tclean, with the w-projection algorithm (Cornwell et al., 2008). In the case of J0229+0053, where the observing bandwidth was large, we additionally used the MT-MFS algorithm (second order expansion; Rau & Cornwell, 2011) to image the target field. The synthesized beams of the final continuum images have full widths at half maximum (FWHMs) of $6.7^{\prime\prime}\times 5.1^{\prime\prime}$ (J0229+0044) and $5.6^{\prime\prime}\times 4.4^{\prime\prime}$ (J0229+0053). Both sources were found to be unresolved in our continuum images, with flux densities of $(3.43\pm 0.38)$ mJy (J0229+0044, at $\approx 640$ MHz) and $(31.08\pm 0.41)$ mJy (J0229+0053, at $\approx 657$ MHz). The quoted errors include the error measured from a 2-dimensional Gaussian fit to a small region around each source and the typical error in the GMRT flux density scale, $\approx 10$% at these frequencies.

The task uvsub was then used to subtract out all detected continuum emission from the self-calibrated visibilities of each field. For each source, a spectral cube was made from these continuum-subtracted visibilities in the barycentric frame, using natural weighting. The Hi 21 cm spectrum was extracted by taking a cut through the cube at the AGN location. The final spectrum for each source was then obtained by fitting a second-order polynomial to the line-free regions of each spectrum, and subtracting this out. We further smoothed the spectra by 7 channels (J0229+0044) and 2 channels (J0229+0053), to velocity resolutions of $\approx 26.6$ km s^-1 and $\approx 11.2$ km s^-1, respectively. The final optical depth RMS noise values are $\approx 0.12$ per 26.6 km s^-1 (J0229+0044) and $0.0068$ per 11.2 km s^-1 (J0229+0053).

Our final Hi 21 cm spectra towards J0229+0044 and J0229+0053 are shown in Figs. 1[A] and 1[B], respectively. The velocity-integrated Hi 21 cm optical depths are $(74.2\pm 7.8)$ km s^-1 (J0229+0044) and $(78.41\pm 0.81)$ km s^-1 (J0229+0053). Assuming ${\rm T_{s}}=100$ K and a covering factor, $f=1$, these integrated Hi 21 cm optical depths imply Hi column densities of $\rm{N_{H{\small\textsc{i}}}}=(1.35\pm 0.14)\times(T_{s}/100)\times 10^{22}$ cm^-2 (J0229+0044) and $\rm{N_{H{\small\textsc{i}}}}=(1.429\pm 0.015)\times(T_{s}/100)\times 10^{22}$ cm^-2 (J0229+0053). Note that these are conservative assumptions: a lower covering factor or a higher spin temperature would imply even higher Hi column densities. The observational results are summarized in Table 1.

Our correlator setup allowed us to carry out a simultaneous search for the redshifted OH 1665 MHz and 1667 MHz lines towards both sources (e.g. Kanekar & Chengalur, 2002; Kanekar et al., 2005). The spectral RMS noise values at the redshifted OH line frequencies were 0.9 mJy/Bm per 9.7 km s^-1 channel (J0229+0044) and 0.36 mJy/Bm per 9.5 km s^-1 channel (J0229+0053); this yields $3\sigma$ upper limits to the velocity-integrated OH 1667 MHz optical depth of $\approx 9.5$ km s^-1 (J0229+0044) and $\approx 0.25$ km s^-1 (J0229+0053), assuming a Gaussian line profile with a line FWHM of 10 km s^-1.

The two new Hi 21 cm absorbers presented here are remarkable due to both the faintness of the background sources, and the high velocity-integrated Hi 21 cm optical depths, larger than that of any known redshifted Hi 21 cm absorber at $z>0.2$ and nearly an order of magnitude higher than those at $z>1$ (e.g. Curran et al., 2013; Aditya & Kanekar, 2018a). Indeed, this is the first time that Hi 21 cm absorption has ever been detected against a source with an integrated flux density lower than $\approx 25$ mJy at the redshifted Hi 21 cm line frequency at any redshift (e.g. Geréb et al., 2015; Maccagni et al., 2017; Aditya & Kanekar, 2018b); J0229+0044 has a flux density of just $\approx 3.4$ mJy! In passing, we note that VLBI Hi 21 cm spectroscopy of low-$z$ AGNs has yielded detections of Hi 21 cm absorption against resolved source components with individual flux densities $\approx 3-10$ mJy (e.g. Józsa et al., 2009; Srianand et al., 2015), as well as similarly high Hi 21 cm optical depths again against individual source components ($\int\tau_{\rm HI}d{\rm V}\approx 50-120$ km s^-1; e.g. Morganti et al., 2004; Srianand et al., 2015).

Both AGNs show evidence for significant reddening, with g$-$K$=3.53$ (J0229+0044) and g$-$K$=4.12$ (J0229+0053). For comparison, only $0.2$% of the 1,697 optically-selected QSOs at $z=1.1-1.3$ in the SDSS-UKIDSS matched catalog of Peth et al. (2011) have g$-$K$>2$, with none having g$-$K$>2.5$. The inferred Hi column density along the two sightlines is extremely high, $\rm{N_{H{\small\textsc{i}}}}\approx 1.4\times(T_{s}/100)\times 10^{22}$ cm^-2. This indicates a high dust column along the sightline, which is likely to cause high obscuration, and hence, reddening of the AGN (e.g. Webster et al., 1995). Searches for associated Hi 21 cm absorption towards red quasars have shown higher detection rates, consistent with the hypothesis that the reddening is due to dust obscuration in the AGN environment (e.g. Carilli et al., 1998; Yan et al., 2016). However, it is possible that the apparent reddening might also arise due to the colour of the AGN host galaxy (e.g. Benn et al., 1998). Aditya & Kanekar (2018b) found no significant evidence of a dependence of the strength of Hi 21 cm absorption on the AGN color, also suggesting that AGN reddening might arise from other causes.

We emphasize that the absorber towards J0229+0053 would not have been detected if we had limited our search to radio sources with known spectroscopic redshifts. Indeed, the detection of strong Hi 21 cm absorption against J0229+0053, which is highly reddened and hence faint at optical and UV wavelengths, demonstrates the power of such unbiased searches in identifying high Hi 21 cm opacity sightlines that are excellent candidates for molecular absorption.

Aditya & Kanekar (2018a, b) used Hi 21 cm absorption surveys of GHz-peaked-spectrum (GPS) AGNs and flat-spectrum AGNs (the latter drawn from the Caltech-Jodrell Flat-spectrum (CJF) sample; e.g. Taylor et al., 1996), respectively, to find that associated Hi 21 cm absorption is significantly weaker in AGNs at both high redshifts, and high rest-frame radio or UV luminosities. These associated Hi 21 cm absorption studies are unique due to the homogeneity of the target AGN samples. Our two AGNs have flat or inverted low-frequency spectra, with radio spectral indices consistent with the spectral-index selection criterion ($\alpha\geq-0.5$) of the CJF sample⁴⁴4We note that the spectral index of J0229+0053 is uncertain due to the difference between the two measurements of its 4.8 GHz flux density; however, the two 4.8 GHz measurements yield spectral indices ($\alpha=-0.45,+0.18$) consistent with the CJF selection criterion (see Section 3.2).. Fig. 2[A] shows the velocity-integrated Hi 21 cm optical depths of the 122 Hi 21 cm absorbers of the CJF and GPS samples of Aditya & Kanekar (2018a, b) plotted as a function of redshift (with detections shown in orange and non-detections in blue). The low Hi 21 cm absorption detection rate and the paucity of high-opacity absorbers at high redshifts are clear from the figure: there are only three confirmed Hi 21 cm absorbers at $z>1$, all with integrated Hi 21 cm opacities $<10$ km s^-1. Figs. 2[B] and [C] illustrate the luminosity bias in the samples, with higher-redshift AGNs having higher radio and UV luminosities. The lower strength of associated Hi 21 cm absorption in high-redshift AGNs of the GPS and CJF samples can hence be explained by (1) redshift evolution of the Hi column density or gas spin temperature in AGN environments, (2) excitation of the upper hyperfine Hi level due to the high AGN 1.4 GHz luminosity, resulting in a high spin temperature, or (3) ionization of the Hi due to the high AGN UV luminosity, resulting in a low Hi column density.

Our two new Hi 21 cm absorbers are shown as red stars in each of the panels of Fig. 2. It is clear that they have far higher Hi 21 cm optical depths, by nearly an order of magnitude, than any Hi 21 cm absorber from the CJF or GPS samples at $z>1$. Further, unlike all the CJF and GPS AGNs at $z>1$, the two AGNs have relatively low luminosities in both the rest-frame 1.4 GHz and the UV wavebands, L${}_{\rm 1.4\;GHz}=10^{25.1}-10^{26.08}$ W Hz^-1 and L${}_{\rm{UV}}\leq 10^{21.7}$ W Hz^-1. Our discovery of two high-$z$ Hi 21 cm absorbers with high velocity-integrated optical depths, and low rest-frame radio and UV AGN luminosities suggests that ionization and/or excitation effects play an important role in determining the strength of associated Hi 21 cm absorption. It thus appears likely that the current paucity of Hi 21 cm absorbers at $z>1$ may be due to the luminosity bias in AGN samples that have so far been searched for associated Hi 21 cm absorption.

Of course, our results do not rule out the possibility of redshift evolution in AGN environments, in either the Hi column density or the spin temperature. However, it is clear that tests of putative redshift evolution must be carried out on AGN samples with similar distributions of UV and radio luminosities, across a range of redshifts.

Finally, we note that the rest-frame 1215 Å luminosities of both J0229+0044 and J0029+0053 are significantly lower than the threshold of L${}_{\rm{UV}}=10^{23}$ W Hz^-1 suggested by Curran et al. (2008), above which UV photons from the AGN may ionize most of the Hi. Indeed, all the AGNs that show confirmed detections of associated Hi 21 cm absorption at $z>1$ have low rest-frame UV luminosities, L${}_{\rm{UV}}\lesssim 2\times 10^{23}$ W Hz^-1 (Curran et al., 2008; Aditya et al., 2017; Aditya & Kanekar, 2018a; Dutta et al., 2020), but show a wide dispersion in their rest-frame 1.4 GHz radio luminosities, L${}_{\rm 1.4\;GHz}\approx 10^{25.1}-10^{28.5}$ W Hz^-1. This suggests that the rarity of associated Hi 21 cm absorbers at $z>1$ may arise mainly because of a decrease in the Hi column density in the AGN environment due to ionization of the Hi by UV photons, rather than an increase in the Hi spin temperature due to proximity to the bright radio AGN.

In summary, we have used an unbiased GMRT Hi 21 cm absorption survey at $z\approx 0.73-1.53$ to discover two new Hi 21 cm absorbers at $z\approx 1.2$, both of which are likely to arise from Hi in the AGN environment. The two absorbers have the highest integrated Hi 21 cm optical depths of all known redshifted Hi 21 cm absorbers, with an inferred Hi column density of $\approx 1.4\times({\rm T_{s}}/100)\times 10^{22}$ cm^-2. Both AGNs are significantly reddened (g$-$K $=3.53,4.12$), consistent with dust obscuration at the above high Hi column densities. The two AGNs have lower rest-frame 1215 Å UV and 1.4 GHz radio luminosities than most high-$z$ AGNs that have hitherto been searched for Hi 21 cm absorption. Our results suggest that ionization effects at high AGN UV luminosities are likely to play an important role in determining the strength of associated Hi 21 cm absorption, and that the current dearth of Hi 21 cm absorbers at $z>1$ may be due to the luminosity bias in current high-$z$ AGN samples with searches for Hi 21 cm absorption. Unbiased Hi 21 cm absorption surveys should allow us to significantly increase the number of Hi 21 cm absorbers at high redshifts in the near future.