Evolution of Cold Gas at 2 < z < 5: A Blind Search for H i and OH Absorption Lines toward Mid-infrared Color-selected Radio-loud AGN

The H i 21 cm absorption lines in the spectra of radio sources can provide valuable insights into the cold atomic gas (T ∼ 100 K) associated with active galactic nuclei (AGN) and intervening galaxies along the line of sight. In the former, generally detected within a few thousand kilometers per second of the AGN redshift, the matter may be associated with the AGN, its host galaxy, a nearby companion galaxy, outflows driven by its feedback, or infalling material. In the latter, the absorbing gas corresponds to the interstellar or circumgalactic medium of an intervening galaxy or intragroup medium. The strength of the absorption signal does not depend on the distance to the observer. The H i 21 cm absorption line's strength depends on both the H i column density and the spin temperature of the gas. The line could thus be an important probe of the properties of cold gas in distant galaxies and investigating the role played by cold gas in fueling the cosmic evolution of star formation rate (SFR) density and that of the luminosity density of AGN, both of which peak at z ≃ 2.

For a long time, radio telescopes have had receivers capable of observing the H i 21 cm line up to arbitrarily high redshifts. Indeed, H i 21 cm absorption has been searched in AGN as distant as z ∼ 5.2 (e.g., Brown & Roberts 1973; Carilli et al. 2007, 1998). But technical limitations imposed by narrow bandwidths, hostile radio frequency environments, and a limited number of known bright radio AGN at high z have prevented large unbiased radio absorption line surveys. Consequently, to date, the majority of H i 21 cm absorption line observations and detections have been based on optically selected samples of AGN.

For associated absorption, AGN with known redshifts and, preferably, compact radio morphology have been observed to study the circumnuclear gas that may be fueling the radio activity or impacted by the AGN feedback (e.g., Vermeulen et al. 2003; Gupta et al. 2006; Darling et al. 2011; Curran et al. 2013; Allison et al. 2014; Geréb et al. 2015; Aditya et al. 2016; Dutta et al. 2019; Grasha et al. 2019). Although more than 500 AGN have been searched for H i 21 cm absorption, the vast majority of observations are at z < 2, and most of the detections are at z < 1 (see Morganti & Oosterloo 2018, for a review). Only three detections at z > 2 are known, the highest redshift being 3.53 (Aditya et al. 2021). Overall, the bulk of the detections are toward compact radio sources (detection rate ∼30%–50%) associated with galaxies having mid-infrared (MIR) colors suggesting gas- and dust-rich environments (Glowacki et al. 2017; Chandola et al. 2020). Among detections associated with more powerful AGN (radio luminosity, log(L_{1.4 GHz}/(W Hz⁻¹) > 24), the H i absorption profiles often show signatures of radio jet–interstellar medium (ISM) interaction in the form of blueshifted components representing outflowing gas (Maccagni et al. 2017).

For intervening H i 21 cm absorption line studies, the targets have been sight lines toward quasars, the most powerful AGN, selected from large optical spectroscopic surveys such as the Sloan Digital Sky Survey (SDSS; York 2000). Generally, sight lines with indications of large H i column densities (N(H i)) along the sight line suggested by the presence of a damped Lyα system (DLA; N(H i) > 2 × 10²⁰ cm⁻²; e.g., Srianand et al. 2012; Kanekar et al. 2014), strong Mg ii absorption (rest equivalent width W_r > 1 Å; e.g., Gupta et al. 2012; Dutta et al. 2017a), or a galaxy at a small impact parameter (typically <30 kpc; e.g., Carilli & van Gorkom 1992; Borthakur et al. 2010; Gupta et al. 2010; Reeves et al. 2016; Dutta et al. 2017c) are selected. The vast majority of the observations are sensitive to detecting the cold neutral medium (CNM; T ∼ 100 K) in N(H I) > 5 × 10¹⁹ cm⁻². The detection rates are typically 10%–50%, depending crucially on the sample selection criteria (see, for example, Dutta et al. 2017b). Although the highest redshift detection is at z ∼ 3.38 (Kanekar et al. 2007), the bulk of the reported H i 21 cm detections are associated with gas-rich galaxies at z < 2. These studies also suggest that the gas traced by DLAs at z > 2 is predominantly warm (T > 1000 K).

It is reasonable to expect optically selected samples of AGN to be affected by dust bias. Since cold gas is accompanied by dust, the bias is particularly relevant for H i 21 cm absorption line searches. In the case of associated absorption, the dust intrinsic to AGN may remove objects with a certain orientation (Type II) or going through the very early stages of evolution. In the case of intervening gas, it can substantially affect our ability to use optically selected samples of DLAs to detect translucent and dense phases of the ISM (Krogager et al. 2016; Geier et al. 2019) and influence the measurements of H i and metal mass densities (Krogager et al. 2019).

The limitations due to dust obscuration can be overcome by selecting AGN without resorting to any optical color selection scheme or carrying out blind searches of H i 21 cm absorption. The latter is becoming possible with various precursor and pathfinder telescopes of the Square Kilometre Array (SKA) equipped with wideband receivers. Specifically, the upcoming large H i 21 cm absorption line surveys such as the MeerKAT Absorption Line Survey (MALS; Gupta et al. 2017) and First Large Absorption Survey in H i (Allison et al. 2017) will characterize the evolution of cold gas without possible selection effects due to dust bias or from the choice of different methods used to select sight lines in different redshift ranges (see also Grasha et al. 2020). These will also simultaneously search the OH 18 cm main lines, providing additional constraints on the evolution of diffuse molecular gas in the ISM (Gupta et al. 2018a; Balashev et al. 2021).

In this paper, we present a spectroscopically blind search of H i 21 cm absorption at z > 2 based on a sample of AGN selected using the MIR colors from the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010; Cutri et al. 2014) and having spectroscopically confirmed redshifts using the Southern African Large Telescope (SALT; 180 hr) and the Nordic Optical Telescope (NOT; 6 nights; Krogager et al. 2018). Note that, similar to the radio wave band, the infrared wavelengths are also unaffected by dust obscuration. These AGN are being observed as part of MALS, which is a large project at the MeerKAT array in South Africa, to search H i 21 cm and OH 18 cm lines at z < 2. The upgraded Giant Metrewave Radio Telescope (uGMRT) survey presented here covers 2 < z < 5.1.

The paper is laid out as follows. In Section 2, we present the sample definition and its properties in the context of previous radio-selected samples to search for DLAs. The details of uGMRT observations and data analysis to obtain the radio spectra and spectral line catalog are presented in Section 3. We provide the details of the H i 21 cm absorber detected from the survey in Section 4. In Sections 5 and 6, we compute the incidences of intervening and associated H i 21 cm absorption lines, respectively. In Section 5, we apply the same formalism to also derive the incidence of intervening OH absorption. The availability of SALT-NOT spectra allows us to examine the properties of gas along the sight line using Lyα and various metal absorption lines. In particular, for a subset of uGMRT targets (z_e > 2.7) through deeper SALT observations, we have discovered six DLAs and one candidate proximate DLA (PDLA; i.e., a DLA within 3000 km s⁻¹ of z_q ). In Section 5, we also present the properties of these Lyα and metal line absorbers and discuss the nature of multiphase ISM in the context of the uGMRT survey results. The results and future prospects are summarized in Section 7.

Throughout this paper, we use the ΛCDM cosmology with Ω_m = 0.315, Ω_Λ = 0.685, and H_o = 67.4 km s⁻¹ Mpc⁻¹ (Planck Collaboration et al. 2020).

2.1. Definition and Properties

The targets for the uGMRT survey are drawn from the SALT-NOT sample of 303 AGN constructed for MALS. The SALT-NOT sample is selected on the basis of MIR colors from WISE. We defined the following color wedge based on the first three bands of WISE, i.e., W1 (3.4 μm), W2 (4.6 μm), and W3 (12 μm):

$$\begin{eqnarray}&&\begin{array}{l}{W}_{1}-{W}_{2}\lt 1.3\times ({W}_{2}-{W}_{3})-3.04;\\ {W}_{1}-{W}_{2}\gt 0.6.\end{array}\end{eqnarray} \tag{ 1 }$$

As shown in Figure 1 of Krogager et al. (2018), the MIR wedge defined above is optimized toward identifying the most powerful AGN (i.e., quasars) at z > 1.4.

Zoom In Zoom Out Reset image size

The details of the SALT-NOT target selection process will be presented in a future paper. In short, we cross-correlated the AllWISE catalog (Cutri et al. 2014) and radio sources brighter than 200 mJy in the NRAO VLA Sky Survey (NVSS; Condon et al. 1998) to identify 2011 high-probability quasar candidates satisfying the MIR wedge (Equation (1)). We restricted the sample to decl. <+20° to ensure reasonable observability with the MeerKAT telescope. A search radius of 10^'' for WISE-NVSS cross-matching was used, but all of the coincidences were verified using higher spatial resolution quick-look radio images at 3 GHz from the Very Large Array Sky Survey (VLASS; Lacy et al. 2020). These quick-look images have a spatial resolution of ∼25, and the positional accuracy is limited to ∼05. Consequently, our sample selects preferentially compact core-dominated AGN. We observed 303 candidates using SALT and NOT to measure redshifts and confirm the AGN nature. This optical spectroscopic campaign has led to a sample of AGN that can be split into the following three categories: (i) with emission lines in the optical spectrum (250 objects with confirmed redshifts at 0.1 < z < 5.1), (ii) with no emission lines in the optical spectrum (26), and (iii) empty fields; i.e., the radio continuum peak coincides with the MIR source, but neither an emission line nor a continuum source are detected in the optical spectra and images.

The uGMRT Band 3 covers 250–500 MHz, which is suitable to search for H i 21 cm absorption over 1.9 < z < 4.7. It nicely complements the MALS coverage of z < 1.4. For the uGMRT survey presented here, we selected 98 objects at z > 2 from the SALT-NOT sample. In the allocated observing time, we observed 88 of these, which are listed in Table 1. The redshift (median ∼2.5) and 1.4 GHz flux density (median ∼288 mJy) distributions are presented in Figure 1. The 1.4 GHz spectral luminosities are in the range of L_{1.4 GHz} ≃ 10^27−29.3 W Hz⁻¹. The lower end of the luminosity is well above the radio cutoff that separates FR I and FR II radio sources, and the upper end corresponds to the most luminous radio-loud AGN at z > 5 discovered from the SALT-NOT survey. All except one are spectroscopically confirmed quasars. The details of radio galaxy M1540–1453 are presented by Shukla et al. (2021).

Table 1. Sample of z > 2 MIR-selected Radio Sources (88) Observed with uGMRT

Source Name	F1.4 GHz	zem	Obs. Run	Beam	Fp,420 MHz	F420 MHz	Fp,420 MHz/F420 MHz	${\alpha }_{0.4}^{1.4}$	αinband	ΔF	Ncand
	(mJy)				(mJy b−1)	(mJy)				(mJy b−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
M004243.06+124657.6	635.0	2.150	16SEP	70 × 63, −120	1628.8	1882.5	0.87	−0.87	−0.86	1.9	⋯
M005315.65−070233.4	248.2	2.130	16SEP	73 × 65, +170	546.1	536.5	1.02	−0.62	−0.61	2.8	⋯
M013047.38−172505.6	250.3	2.528	14SEP	99 × 69, +400	532.4	560.2	0.95	−0.64	−0.66	1.5	⋯
M021231.86−382256.6	244.5	2.260	14SEP	175 × 68, +330	495.7	605.4	0.82	−0.72	−0.71	1.6	1
M022613.72+093726.3	374.6	2.605	14SEP	89 × 70, +870	435.1	436.6	1.0	−0.12	−0.02	2.1	⋯
M022639.92+194110.1	209.8	2.190	14SEP	104 × 67, −830	390.5	381.7	1.02	−0.48	−0.41	1.9	⋯
M024939.93+044028.9	420.5	2.008	17SEP	91 × 76, +890	927.7	992.7	0.93	−0.69	−0.67	1.6	⋯
M025035.54−262743.1	389.2	2.918	17SEP	110 × 67, +250	389.6	419.0	0.93	−0.06	+0.01	1.7	1
M032808.59−015220.2	221.9	2.679	17SEP	120 × 80, +720	370.2	527.3	0.70	−0.69	−0.63	1.3	⋯
M041620.54−333931.3	264.1	3.045	08SEP	175 × 90, −420	130.0	117.8	1.1	+0.64	+0.51	1.7	⋯
M042248.53−203456.6	224.3	2.582	08SEP	118 × 84, −580	187.3	192.1	0.98	+0.12	+0.16	1.7	1
M044849.48−093531.3	240.9	2.079	14SEP	93 × 72, +460	152.5	147.3	1.04	+0.39	+0.39	1.6	⋯
M050725.04−362442.9	212.4	2.930	08SEP	159 × 96, −350	494.5	474.5	1.04	−0.64	−0.51	1.6	⋯
M051240.99+151723.8	966.5	2.568	07SEP	68 × 65, −790	560.1	595.9	0.94	+0.39	+0.27	1.8	⋯
M051340.03+010023.6	447.0	2.673	07SEP	77 × 66, +640	342.5	349.7	0.98	+0.20	−0.04	2.1	4
M051511.18−012002.4	288.8	2.287	07SEP	89 × 64, +590	412.6	641.5	0.64	−0.64	−0.68	1.6	1
M051656.35+073252.7	231.7	2.594	07SEP	106 × 69, +900	44.2	44.1	1.0	+1.32	+1.10	1.9	⋯
M052318.55−261409.6	1354.9	3.110	08SEP	120 × 91, −520	477.7	451.8	1.06	+0.88	+1.08	2.2	⋯
M061038.80−230145.6	360.2	2.829	14SEP	110 × 70, +310	130.5	129.0	1.01	+0.82	+0.89	1.7	⋯
M061856.02−315835.2	346.1	2.134	09SEP	108 × 63, +00	877.1	828.6	1.06	−0.70	−0.35	2.0	1
M063602.28−311312.5	262.1	2.654	09SEP	192 × 111, +330	178.7	162.4	1.1	+0.38	+0.42	3.1	⋯
M063613.53−310646.3	208.0	2.757	09SEP	116 × 69, +170	436.2	474.2	0.92	−0.66	−0.73	2.1	⋯
M065254.73−323022.6 a	322.1	2.239	08SEP	114 × 101, +50	475.9	611.1	0.78	−0.85	−0.95	2.3	⋯
					279.0	328.2	0.85			2.3	⋯
M070249.30−330205.0	314.6	2.410	08SEP	123 × 98, +330	574.9	599.8	0.96	−0.52	−0.52	2.8	1
M073159.01+143336.3	316.5	2.632	16SEP	88 × 81, −490	180.0	185.4	0.97	+0.43	+0.40	7.8	⋯
M073714.60−382841.9	219.3	2.107	14SEP	145 × 66, +180	498.9	515.3	0.97	−0.68	−0.67	2.4	⋯
M080804.34+005708.2	317.0	3.133	08SEP	143 × 71, −730	434.8	450.7	0.96	−0.28	−0.26	2.1	⋯
M081936.62−063047.9	280.0	2.507	08SEP	119 × 67, −730	339.0	354.9	0.96	−0.19	−0.08	2.3	⋯
M085826.92−260721.0	404.5	2.036	09SEP	173 × 85, −190	561.0	523.9	1.07	−0.21	−0.28	2.6	⋯
M090910.66−163753.8	340.1	2.475	09SEP	90 × 73, −110	776.2	909.7	0.85	−0.79	−0.94	1.7	⋯
M091051.01−052626.8 a	337.9	2.395	16SEP	96 × 72, −440	151.4	166.3	0.91	+0.20	+0.29	1.6	⋯
					80.0	97.3	0.82			1.6	⋯
M095231.66−245349.1	209.5	2.626	14SEP	104 × 70, +50	224.2	210.3	1.07	−0.00	−0.17	1.8	⋯
M100715.18−124746.7	381.1	2.113	16SEP	77 × 62, −240	476.6	470.0	1.01	−0.17	−0.21	2.1	⋯
M101313.10−254654.7	248.8	2.965	14SEP	106 × 69, +100	216.7	218.6	0.99	+0.10	−0.0	2.1	⋯
M102548.76−042933.0	363.5	2.292	16SEP	69 × 67, +890	356.3	391.6	0.91	−0.06	−0.03	2.5	⋯
M104314.53−232317.5	212.1	2.881	07SEP	89 × 65, −100	460.6	505.8	0.91	−0.69	−0.79	2.2	⋯
M111820.61−305459.0	233.2	2.352	08SEP	185 × 84, −510	64.8	70.0	0.93	+0.96	+0.10	3.0	⋯
M111917.36−052707.9	1174.4	2.651	07SEP	71 × 63, −480	1696.6	1793.2	0.95	−0.34	−0.47	3.6	⋯
M112402.56−150159.1	261.9	2.551	07SEP	90 × 69, −230	194.7	196.0	0.99	+0.23	+0.01	2.6	⋯
M114226.58−263313.7	294.7	3.237	08SEP	155 × 83, −560	293.6	340.7	0.86	−0.35	−0.31	4.0	⋯
M115222.04−270126.3	238.2	2.703	08SEP	133 × 86, −600	410.4	441.4	0.93	−0.49	−0.64	2.5	⋯
M115306.72−044254.5 a	684.8	2.591	16SEP	70 × 65, +860	929.3	899.1	1.03	−0.78	−0.89	2.7	⋯
					980.1	912.2	1.07			2.7	⋯
M120632.23−071452.6	698.8	2.263	16SEP	80 × 62, −90	1402.8	1267.9	1.10	−0.48	−0.60	2.7	⋯
M121514.42−062803.5	360.4	3.218	16SEP	92 × 66, −80	511.1	461.5	1.11	−0.20	−0.31	2.8	⋯
M123150.30−123637.5	276.0	2.106	07SEP	72 × 65, −170	159.7	205.1	0.78	+0.24	+0.36	1.8	⋯
M123410.08−332638.5	297.9	2.820	08SEP	158 × 97, −570	665.4	710.4	0.94	−0.69	−0.72	3.4	⋯
M124448.99−044610.2	384.9	3.104	16SEP	91 × 77, +280	677.2	616.1	1.1	−0.38	−0.40	2.3	1
M125442.98−383356.4	219.2	2.776	17SEP	147 × 70, +70	297.4	300.9	0.99	−0.25	−0.25	2.3	⋯
M125611.49−214411.7	260.7	2.178	16SEP	103 × 65, +130	446.8	455.4	0.98	−0.45	−0.15	1.7	⋯
M131207.86−202652.4	778.1	5.064	16SEP	105 × 67, +210	1727.9	1721.9	1.0	−0.63	−0.62	2.0	1
			21APR b	158 × 138, −380 b	-	2577.0 b	-	-	-	-	⋯
M132657.20−280831.4	404.5	2.238	16SEP	150 × 70, +200	671.3	874.8	0.77	−0.62	−0.83	2.7	⋯
M135131.98−101932.9	726.1	2.999	16SEP	92 × 70, +470	1338.3	1886.5	0.71	−0.76	−0.79	2.0	⋯
M141327.20−342235.1	274.7	2.812	09SEP	209 × 67, −420	78.2	76.4	1.02	+1.02	+0.81	3.9	⋯
M143709.04−294718.5	273.8	2.331	09SEP	146 × 70, −440	333.0	456.1	0.73	−0.41	−0.49	2.5	⋯
M144851.10−112215.6	455.5	2.630	07SEP	73 × 60, −480	967.0	896.8	1.08	−0.54	−0.71	3.1	1
M145342.95−132735.2	254.5	2.370	07SEP	76 × 62, −330	477.3	634.7	0.75	−0.73	−0.83	2.2	⋯
M145502.84−170014.2	294.7	2.291	07SEP	83 × 64, −130	345.8	352.2	0.98	−0.14	−0.23	2.3	⋯
M145625.83+045645.2	287.9	2.134	09SEP	71 × 68, +870	800.8	813.2	0.98	−0.83	−0.82	2.6	⋯
M145908.92−164542.3	378.9	2.006	07SEP	85 × 66, −90	853.1	909.6	0.94	−0.70	−0.85	2.1	⋯
M150425.30+081858.6	210.8	2.035	09SEP	84 × 76, −320	122.1	138.5	0.88	+0.34	+0.27	4.8	⋯
M151129.01−072255.3	326.3	2.582	17SEP	78 × 68, −850	672.7	624.9	1.08	−0.52	−0.71	2.7	⋯
M151304.72−252439.7 a	217.6	3.132	09SEP	91 × 68, +10	855.7	819.4	1.04	−1.26	−1.26	3.6	⋯
					268.5	242.1	1.11			3.6	⋯
M151944.77−115144.6	441.0	2.014	17SEP	92 × 76, +840	425.0	561.5	0.76	−0.19	−0.01	3.7	⋯
M154015.23−145341.5	203.3	2.098	17SEP	98 × 75, +610	642.4	595.0	1.08	−0.86	−0.98	2.6	1
M155825.35−215511.1	206.9	2.760	09SEP	137 × 67, −420	209.5	235.8	0.89	−0.10	+0.09	3.0	⋯
M161907.44−093952.5	340.3	2.891	17SEP	89 × 68, +660	757.0	695.5	1.09	−0.57	−0.43	2.2	⋯
M162047.94+003653.2	317.8	2.438	17SEP	107 × 76, +20	226.9	234.3	0.97	+0.24	+0.16	3.3	⋯
M164950.51+062653.3	389.2	2.144	17SEP	136 × 93, −110	154.3	204.3	0.76	+0.51	+0.36	2.6	⋯
M165038.03−124854.5	275.5	2.527	09SEP	73 × 62, −200	729.6	675.1	1.08	−0.72	−0.52	5.0	⋯
M165435.38+001719.2	255.3	2.363	07SEP	92 × 73, −140	405.7	381.5	1.06	−0.32	−0.44	4.2	⋯
M194110.28−300720.9	315.0	2.059	17SEP	236 × 83, +510	164.0	227.2	0.72	+0.26	+0.40	3.6	⋯
M200209.37−145531.8	620.3	2.192	17SEP	86 × 71, −170	942.0	896.1	1.05	−0.29	−0.31	2.9	⋯
M201708.96−293354.7	327.2	2.617	17SEP	205 × 73, +500	1044.7	1201.1	0.87	−1.04	−1.02	2.7	1
M203425.65−052332.2	419.7	2.070	17SEP	80 × 73, +850	366.2	407.2	0.9	+0.02	+0.10	4.0	⋯
M204737.67−184141.2	241.7	2.994	16SEP	138 × 67, −490	273.0	284.3	0.96	−0.13	−0.17	2.2	⋯
M205245.03−223410.6	330.9	2.072	16SEP	123 × 60, −410	631.8	608.3	1.04	−0.49	−0.60	3.5	⋯
M210143.29−174759.2	959.5	2.803	16SEP	95 × 58, −390	2554.1	2477.5	1.03	−0.76	−0.91	3.6	⋯
M212821.83−150453.2	245.5	2.547	17SEP	92 × 69, +60	443.4	460.8	0.96	−0.50	−0.49	2.0	⋯
M220127.50+031215.6	300.5	2.181	17SEP	183 × 86, +770	180.5	191.6	0.94	+0.36	+0.33	1.7	⋯
M222332.81−310117.3	231.7	3.206	17SEP	250 × 92, +400	230.6	330.2	0.7	−0.28	−0.40	2.6	⋯
M223816.27−124036.4	213.6	2.623	16SEP	123 × 83, −500	464.3	435.5	1.07	−0.57	−0.65	3.7	⋯
M224111.48−244239.0	211.4	2.242	16SEP	109 × 61, −310	253.0	254.8	0.99	−0.15	−0.17	2.2	⋯
M224705.52+121151.4	223.7	2.185	14SEP	121 × 75, +830	474.3	489.1	0.97	−0.62	−0.60	1.9	⋯
M224950.57−263459.6	228.8	2.174	16SEP	98 × 57, −230	568.3	542.1	1.05	−0.69	−0.59	2.4	⋯
M230036.41+194002.9	210.4	2.160	17SEP	277 × 82, +790	475.8	556.0	0.86	−0.78	−0.65	11.2	⋯
M231634.61+042940.2	214.0	2.180	16SEP	69 × 66, +470	103.3	97.9	1.06	+0.70	+0.50	3.0	⋯
M234910.12−043803.2	206.1	2.240	16SEP	77 × 68, −620	168.3	185.6	0.91	+0.08	−0.04	5.1	⋯
M235722.47−073134.3	235.5	2.764	16SEP	70 × 60, −190	372.4	398.2	0.94	−0.42	−0.40	2.4	⋯

Notes. Column (1): source name based on R.A. and decl. (J2000) from NVSS. Column (2): 20 cm flux density from NVSS. Column (3): emission line redshift measured from SALT-NOT survey. Column (4): observing run (see Table 2). Note that only M1312–2026 was also observed at Band 2. Columns (5)–(7): synthesized beam, peak of the prominent Gaussian component, and total flux densities, respectively, from the continuum image based on a 390–450 MHz range (average 420 MHz). Column (8): ratio of columns (6) and (7). Column (9): spectral index derived using NVSS and 420 MHz flux densities. In a few cases, which are all sources with a single component fit, this ratio marginally (∼5%) exceeds 1, suggesting that the radio emission may be partially resolved. Column (10): in-band spectral index. Column (11): observed spectral rms at 420 MHz. Column (12): number of absorption candidates.

^a The radio source is double-lobed in the 420 MHz image (see Section 3.1 for details). ^b Corresponds to Band 2.

Download table as:  ASCIITypeset images: 1 2 3

With the right sample, it is possible to determine the evolution of cold gas in a dust-unbiased way. Therefore, we next examine the efficacy of our sample selection strategy by comparing it with samples of DLAs from radio-selected quasars.

2.2. Comparison with Radio-selected DLA Samples

The three notable DLA samples based on radio-selected quasars are (i) the Complete Optical and Radio Absorption Line System (CORALS) survey of 66 QSOs (z_em > 2.2) by Ellison et al. (2001), (ii) the University of California San Diego (UCSD) survey of 53 QSOs (z_em > 2.0) for DLAs by Jorgenson et al. (2006), and (iii) the survey of 45 QSOs (z_em > 2.4) selected from the Texas radio survey (Ellison et al. 2008). These surveys revealed 19, 7, and 9 DLAs over a redshift path, Δz, of 57.16, 41.15, and 38.79, respectively. The number of DLAs per unit redshift, n_DLA, are estimated to be ${0.31}_{-0.08}^{+0.09}$, ${0.17}_{-0.07}^{+0.08}$, and ${0.23}_{-0.07}^{+0.11}$, respectively. The CORALS survey found a slightly higher incidence of DLAs and suggested that optically selected DLA samples may be affected by dust bias. But overall, none of the surveys uncovered a population of dusty DLAs.

Targets for these three surveys have been selected at different radio frequencies and to different radio flux limits. While such differences might be subtle, they may still affect the optical properties of the quasars and hence the resulting statistics of DLAs. The CORALS survey has been selected at 2.7 GHz down to a flux density limit of 250 mJy, the UCSD sample has been selected at 4.9 GHz to a flux density limit of 350 mJy, and, lastly, the survey by Ellison et al. (2008) has been selected at 356 MHz down to a flux density limit of 400 mJy.

In order to compare the effects of the radio selection, we generate a mock distribution of i-band magnitudes for the three samples, as well as for the SALT-NOT sample presented in this work. The intrinsic ultraviolet luminosity function is assumed to be the same in all cases and is taken from the work by Manti et al. (2017). We assume a fixed distribution of the optical-to-radio flux ratio, R_i , following Baloković et al. (2012), as well as a fixed radio slope of α_ν = −0.8, in order to scale the various survey limits to the same frequency (1.4 GHz) as used in our survey and by Baloković et al. (2012). Since all of the surveys impose roughly the same optical follow-up strategy in order to detect DLAs in low-resolution spectra, we impose a final cut on B < 22 mag. For this cut, we use an average color correction for high-redshift QSOs: B = i + 0.3 with a scatter of 0.1 mag (see color relations by Krogager et al. 2019). The resulting mock magnitude distribution is shown in Figure 2 (blue curve) compared to the respective survey data (in black). While all surveys span a wide range of magnitudes, our survey more closely samples the underlying luminosity function and hence introduces a minimal bias in the optical properties of the sample. This is a direct consequence of the fact that the SALT-NOT survey has targeted optically fainter quasars (refer to the median i-band magnitudes in Figure 2). An analysis of dust bias in the sample using optical-infrared colors will be presented in a future paper.

Zoom In Zoom Out Reset image size

3.1. Observations

3.2. Data Analysis

All of the data were edited, calibrated, and imaged using the Automated Radio Telescope Imaging Pipeline (ARTIP) following the steps described in Gupta et al. (2021). After flagging and calibration, the spectral line processing of wideband Band 3 data was sped up by partitioning the 200 MHz bandwidth into four 50 or 60 MHz wide spectral windows with an overlap of 10 MHz between the adjacent windows. These spectral windows covered 300–360, 350–410, 400–460, and 450–500 MHz. The calibrated visibilities for each spectral window were processed separately (independently) for radio frequency interference (RFI) flagging, continuum imaging and self-calibration, and continuum subtraction. The continuum-subtracted visibilities were imaged to obtain RR and LL spectral line cubes. For this, a "common" synthesized beam corresponding to the lowest-frequency spectral window was used.

The narrowband data sets from the Band 2 survey and Band 3 follow-up observations were processed as a single 4.17 or 6.25 MHz wide spectral window, respectively.

3.2.1. Continuum Analysis

For Band 3, the spectral window covering 390–450 MHz, hereafter identified through the central reference frequency of 420 MHz, is least affected by RFI, resulting in the best possible continuum images from the data. We used CASA task IMFIT to model the radio continuum emission in these images as multiple Gaussian components. The nine cases requiring more than one Gaussian component are shown in Figure 3. Only in four cases, i.e., M0652–3230, M0910–0526, M1153–0442, and M1513–2524, does the second component contain more than 20% of the total flux density.

Zoom In Zoom Out Reset image size

In columns (5) and (6) of Table 1, we list the synthesized beams and peak flux densities of the most prominent Gaussian component. For the abovementioned four sources, the second component is also listed. In the remaining cases, the additional components are too faint (<50 mJy) to be useful for the objectives of this paper; hence, we do not list their individual properties. The total flux density as estimated from the single- or multiple-component fit is provided in column (7).

In Figure 3, we also show optical images from Pan-STARRS1 (PS1; Chambers et al. 2016). Note that M0652–3230 is too far south to be covered in PS1. The locations of MIR sources from WISE are also shown in the images. Owing to the MIR-selection wedge described in Section 2, all but one radio source in our sample are quasars (see Section 2.1 for details). Indeed, the median spectral index, ¹⁰ ${\alpha }_{0.4}^{1.4}$, derived using the NVSS 1420 MHz and uGMRT 420 MHz total flux densities is −0.38 (see column (9) of Table 1 and Figure 4). As expected, this is flatter than the overall radio source population, which has α ∼ −0.8. Thus, for our sample, when radio emission is dominated by a single component, we expect AGN to be located close to the peak of the radio emission. In case two prominent radio components are present, i.e., a compact symmetric object (Conway 2002) morphology, the AGN is expected to be located between them. In all but two cases (M0910–0526 and M1513–2524; details provided below), the optical/MIR counterpart is at the location of the AGN expected from the radio morphology. As previously mentioned, we have also verified these coincidences using higher spatial resolution 3 GHz VLASS images.

Zoom In Zoom Out Reset image size

In the case of M0910–0526, the northern component could be an unrelated radio source (Figure 3). We will exclude this component from the absorption line statistics. The only radio galaxy in the sample presented here, M1513–2524, is among the optically faintest (r > 23 mag) in our survey. Among the two radio components, one of them is closer to the MIR source (see Figure 3). We have tentatively detected faint radio emission, i.e., a radio core at the location of the MIR source. For details, see the higher spatial resolution radio images presented in Shukla et al. (2021). We will consider the eastern and western radio components as the two radio lobes of this radio galaxy.

For M1312–2026, the only target also observed in Band 2, the properties at 234 MHz are also provided in Table 1. The associated radio emission is compact, with a deconvolved size of 158 × 138 (position angle = −380). Based on the observed flux densities, M1312–2026, has a spectral luminosity of L(1.4 GHz) = 1.2 × 10²⁹ W Hz⁻¹, which is more than 3 orders of magnitude higher than the radio power cutoff that separates FR I and FR II radio sources and greater than the luminosity of any known radio-loud AGN at z > 5. The multifrequency Very Large Array and Very Long Baseline Array observations of this AGN have been obtained to investigate its radio morphology.

3.3. Spectral Line Analysis

For the spectral line analysis, RR and LL spectra in the heliocentric frame were extracted at the location of the radio continuum peaks from the spectral line cubes. The spectra show systematic oscillations or ripples due to residual bandpass calibration and numerous positive/ negative spikes (for an example, see Figure 5). The ripple is not identical in the two parallel hands and also varies from one target source to other. We removed its effect by iteratively fitting the underlying structure using Savitsky–Golay filtering with window length = 24 and polynomial order = 3. In each iteration, the pixels deviating beyond the threshold were flagged and excluded from the subsequent iterations. The continuum was interpolated across the masked pixels, and the process was repeated until no deviant pixels were found.

Zoom In Zoom Out Reset image size

The above determined continuum fit was subtracted from the RR and LL spectra, and an error spectrum was generated by calculating a rolling standard deviation (σ_rolling; window size = 48 channels). For Band 3, an additional step was to merge the spectra from adjacent spectral windows and unmask the spikes to obtain the final RR and LL spectra covering the entire 300–500 MHz. These were then averaged with appropriate statistical weights to obtain the final Stokes I spectra. The resultant Stokes I spectra have flat baselines but numerous positive and negative spikes (for example, see Figure 6).

Zoom In Zoom Out Reset image size

The Band 2 spectrum of M1312–2026 is presented in Figure 7. These data were severely affected by the RFI. The broadband RFI mostly affected the shorter baselines (<4 kλ), which were completely flagged. There were also narrowband impulse-like bursts of RFI, which affected all of the baselines. Overall, ∼55% of the data were flagged due to antenna/baseline-based problems and RFI. The spectral rms in the Stokes I spectrum presented here is 8.5 mJy beam⁻¹, which for the unsmoothed spectrum presented here corresponds to a 1σ optical depth sensitivity of 0.003. There are several statistically significant narrowband features with widths of one to two spectral channels detected in the spectrum. But all of these are coincident with the spikes present in the RFI spectrum and are most likely due to the low-level narrowband RFI, which could not be detected in the individual baselines.

Zoom In Zoom Out Reset image size

In general, no true emission is expected in our spectra, and only a tiny fraction of negative spikes are expected to represent true absorption. The majority of these spikes are RFI artifacts. The biggest issue in spectral line searching at low frequencies is to distinguish between true absorption and RFI artifacts. The rest of this section is concerned with an absorption line search in Band 3 spectra.

The worst RFI in the Band 3 spectra (e.g., at 360–380 MHz) was flagged by applying an initial mask prior to any calibration (see shaded regions in Figure 6). Further, to identify artifacts due to weaker but persistent RFI, we generated median Stokes I spectra for each observing run and the full survey. The median spectra from the survey and 08SEP run in which M0422–2034 was observed are shown in Figure 6. The pixels deviating by more than 5σ in the median spectra were taken to represent RFI artifacts. We rejected corresponding pixels in the individual source spectrum. In Figure 6, such pixels for M0422–2034 are plotted in red.

After this, we created a list of 550 absorption line candidates using a detection algorithm that requires (i) a flux density at a pixel j, F(j) < −5 × σ_rolling(j), and (ii) a heliocentric frequency at j, ν(j) ≥ ν_{21 cm}/(1 + (z_em + 0.01)), where ν_{21 cm} is the rest-frame 21 cm line frequency. The factor of 0.01 in the denominator, which approximately corresponds to a outflowing velocity of ∼3000 km s⁻¹, allows for the possibility of detecting redshifted absorption associated with AGN (see Figure 21 of Gupta et al. 2006).

Next, we created a false-detection spectrum by identifying pixels based on following two characteristics. First, we identified all positive spikes with F(j) > 5 × σ_rolling(j). These are unphysical and hence false detections because the H i emission lines are too weak to be detectable at z > 2 in our spectra. Second, we identified all negative spikes, i.e., F(j) < −5 × σ_rolling, but only at ν(j) < ν_{21 cm}/(1 + (z_em + 0.01)). These are unphysical because the absorbing gas must be in the front of the radio source. In Figure 6, we mark three candidates with stars. Two of these are clearly false detections, whereas the third one at 395.5 MHz (approximately +800 km s⁻¹ with respect to z_em) could be a true absorption associated with the AGN. The cumulative distributions of all of the false absorption (528) and emission (1359) detections from the survey are shown in Figure 8. These represent frequency ranges that may be affected by sporadic RFI. The majority of these are at the edges of frequency ranges masked in Figure 10. An updated RFI mask would get rid of them. This will certainly be the preferred strategy for defining the frequency ranges to be used for continuum imaging. Here we rejected all of the absorption candidates that are within one frequency channel of any of these false detections. This step reduced the number of absorption candidates by a factor of ∼10.

Zoom In Zoom Out Reset image size

We visually examined the RR and LL spectra of the remaining 48 absorption candidates for consistency. Specifically, we imposed the criteria that integrated optical depths estimated using the RR and LL spectra match within 3σ and, within the errors, the absorption profiles appear similar.

After all of the statistical filtering described above, we were left with a total of 15 candidates toward the sight lines that are identified in column (12) of Table 1. We extracted Stokes I spectra of the gain calibrators corresponding to these. For candidates M0212–3822 (z_abs = 2.1666), M0250–2627 (z_abs = 2.1665), M0422–2034 (z_abs = 2.5924), M0513+0100 (z_abs = 2.1612, 2.3183, 2.6434), M0515–0120 (z_abs = 2.1753), M0702–3302 (z_abs = 2.2769), M1448–1122 (z_abs = 2.2973), and M2017–2933 (z_abs = 2.0733), we find an "absorption" feature at the same redshifted frequency in the gain calibrator spectrum. The angular separation between a target source and its gain calibrator is typically 15°. Thus, it is unrealistic that at z > 2, a true absorption is present in both of them. Therefore, we rejected these 10 candidates.

Finally, we have five high-probability candidates. These are listed in Table 3. We also estimated integrated optical depths (∫τ dv) and velocity widths (ΔV₉₀) corresponding to the 5th and 95th percentiles of the apparent optical depth distribution. These are very similar to the values observed for the 21 cm absorption lines detected in various surveys (see, e.g., Gupta et al. 2009; Dutta et al. 2017c).

Table 3. High-probability 21 cm Absorption Candidates

Source Name	zem	zabs(21 cm)	∫τ dv(21 cm)	ΔV90
			(km s−1)	(km s−1)
(1)	(2)	(3)	(4)	(5)
M0513+0100	2.673	1.9526	3.59 ± 0.67	107
M0618−3158	2.134	1.9642	0.49 ± 0.07	15
M1244−0446	3.114	2.3871	1.61 ± 0.27	70
M1312−2026	5.064	3.0324	0.34 ± 0.08	20
M1540−1453	2.098	2.1139	9.14 ± 0.28	144

Note. Column (1): source name. Column (2): emission redshift. Columns (3)–(5): H i 21 cm absorption redshift, integrated 21 cm optical depth limit, and velocity width of the absorption profile based on the survey spectra presented in Figure 9, respectively. The spectra have been normalized using the corresponding peak flux densities.

Download table as:  ASCIITypeset image

We reobserved these high-probability candidates with uGMRT (see Section 3.1 and Table 2) using a bandwidth of 6.25 MHz centered at the H i 21 cm line frequency corresponding to z_abs(21 cm) given in column (3) of Table 3. These observations were carried out at night to reduce the effect of RFI. For better RFI mitigation, the frequency setup was chosen to provide a spectral resolution of ∼1 km s⁻¹. Recall that the survey observations had a spectral resolution of ∼18 km s⁻¹. In Figure 9, we present profiles from the survey and reobservation spectra. Clearly, only M1540–1453 is confirmed. The remaining four candidates are due to RFI.

Zoom In Zoom Out Reset image size

To summarize, based purely on the uGMRT survey spectra and blind 21 cm line search, we identified five absorption features (four intervening and one associated system). The follow-up observations confirmed only one of these, i.e., absorption associated with the radio source M1540–1453 at z_em = 2.098. The distribution of 5σ 21 cm optical depth limits at 420 MHz estimated assuming Δv = 25 km s⁻¹ is shown in Figure 4. The median 0.535 km s⁻¹ is well below the sensitivity (1.1 km s⁻¹) required to detect CNM (i.e., T ∼ 100 K) in DLAs (i.e N(H i) ≥ 10^20.3 cm⁻²).

The H i 21 cm absorption spectrum of M1540–1453 based on the follow-up observation is presented in Figure 10. It is normalized using the corresponding peak continuum flux density of 652 mJy beam⁻¹. The absorption is spread over ∼300 km s⁻¹. We measure a total optical depth of ∫τ dv = 11.30 ± 0.07 km s⁻¹, and about 90% of this is confined to ΔV₉₀ = 167 km s⁻¹. This translates to a H i column density of $(2.06\pm 0.01)\,\times {10}^{21}\left(\tfrac{{T}_{S}}{100}\right)\left(\tfrac{1}{{f}_{c}}\right)$ cm⁻². Here T_s is the spin temperature in kelvin and f_c is the covering factor of the absorbing gas. We note that measuring the covering factor of the absorbing gas would require milliarcsecond-scale spectroscopy, which is currently not feasible at z > 0.2 (Srianand et al. 2013; Gupta et al. 2018b). Therefore, unless explicitly stated otherwise, hereafter, we will assume f_c = 1. We also assume T_s = 100 K for the CNM (Heiles & Troland 2003).

Zoom In Zoom Out Reset image size

We obtained an optical spectrum of M1540–1453 using the Robert Stobie Spectrograph (Burgh et al. 2003; Kobulnicky et al. 2003) on SALT as part of our optical survey summarized in Section 2. The spectrum has a typical signal-to-noise ratio (S/N) of 7 pixel^–1. It shows C iv and [C iii] emission lines. The spectral S/N close to the [C iii] emission line is poor due to residuals from the subtraction of skylines. Hence, we focus on the C iv emission line. The peak of C iv emission corresponds to z_em ≃ 2.113, which is consistent with the 21 cm absorption peak (Figure 10). The emission line is superimposed with absorption lines possibly also of C iv but at a redshift slightly lower than the 21 cm absorption.

Our SALT spectrum does not cover Lyα absorption for M1540–1453. But it covers the rest wavelength range of 1436–2414 Å. Although the region is affected by skylines, in principle, we have access to the Fe ii lines associated with the 21 cm absorption. We detect an absorption feature exactly at the redshifted wavelength corresponding to the Fe ii λ2383 line. The redshift and rest equivalent width of the absorption are z_abs = 2.11404 and W_r = 1.05 ± 0.25 Å, respectively. We also find absorption dips at the expected positions of the Fe ii λ2344, Fe ii λ2374, and Si ii λ1526 lines. These coincidences are interesting because metal absorption line ratios can be a reasonable indicator of H i column density (Rao et al. 2006; Gupta et al. 2012; Dutta et al. 2017b). We note the rest-frame ultraviolet luminosity of the quasar at 912 Å, i.e., L₉₁₂ = 1.4 × 10²³ W Hz⁻¹ (see Section 6 for details). This is marginally above the ultraviolet cutoff of 10²³ W Hz⁻¹, above which H i 21 cm absorption is rarely detected (Curran & Whiting 2010). A better-quality optical spectrum covering Lyα and the abovementioned metal lines is needed to extract the physical conditions prevailing in the absorbing gas.

We follow the method described in Srianand et al. (2008) to constrain the visual extinction, A_V . Using our flux-calibrated SALT spectrum, along with the Small Magellanic Cloud–type extinction curve and the average QSO spectral energy distribution given in Selsing et al. (2016), we measure A_V = 0.13 ± 0.01. The moderate extinction observed toward M1540–1453 is consistent with the idea that cold atomic gas is accompanied by dust (see Figure 9 of Dutta et al. 2017b).

To date, only three associated H i 21 cm absorbers are known ¹¹ at z > 2. These are z_abs = 3.3968 toward B2 0902+345 (Uson et al. 1991; Briggs et al. 1993), z_abs = 2.6365 toward MG J0414+0534 (Moore et al. 1999), and z_abs = 3.52965 toward 8C 0604+728 (Aditya et al. 2021). Thus, the M1540–1453 absorber reported here is only the fourth detection at z > 2. The inferred column densities for the reasonably assumed values of spin temperature and covering factor (T_s = 100 K; f_c = 1) imply N(H i) ≫ 2 × 10²⁰ cm⁻², which is the formal DLA cutoff. So, in all of these cases, if the optical and radio sight lines coincide, then one expects to see a DLA at the 21 cm absorption redshift. We investigate this for B2 0902+345 and MG J0414+0534, the two sources for which optical spectra are available in the literature.

Object B2 0902+345 is associated with a radio galaxy that exhibits Lyα emission extended up to 50 kpc. The associated radio continuum emission (α = −0.94) has a highly distorted radio morphology over 6'' (∼45 kpc at z_abs) and a rotation measure in excess of 1000 rad m⁻² (Carilli et al. 1994). However, no signatures of Lyα absorption associated with the 21 cm absorber are seen. In fact, the 21 cm absorption is found to be redshifted with respect to the Lyα emission (shift ∼+300 km s⁻¹; see Adams et al. 2009, for details).

Object MG J0414+0534 is a highly reddened gravitationally lensed quasar (A_V ∼ 5; Lawrence et al. 1995b). The weakness of the Lyα emission prevents us from searching for a DLA. However, four associated strong Fe ii absorption components are detected in the redshift range 2.6317–2.6447 (Lawrence et al. 1995a). These are within the range over which CO emission is detected but do not exactly coincide with the 21 cm absorption redshift, which itself is shifted by ∼200 km s⁻¹ with respect to the peak of the CO emission line. The 21 cm absorption in this case may actually be toward a steep-spectrum radio jet component not spatially coinciding with the AGN (Moore et al. 1999). The same scenario may also apply to B2 0902+345.

In comparison, the radio emission associated with M1540–1453 is compact in the VLASS, and the radio continuum peak coincides well with the PS1/MIR counterparts. The deconvolved radio source size is 18 × 07 with a position angle of 155°. This corresponds to an upper limit on the size of ∼10 kpc at z_abs = 2.1139. Clearly, milliarcsecond-scale imaging is required to estimate f_c and understand the coincidence between 21 cm absorption and metal absorption lines observed toward this source. Interestingly, the H i 21 cm absorption is only slightly asymmetric at the base and does not show signatures of outflows, i.e., blueshifted absorption (Vermeulen et al. 2003; Gupta et al. 2006). In low-z samples, such absorption likely originates from the circumnuclear disk or gas clouds associated with the host galaxy (e.g., Geréb et al. 2015; Srianand et al. 2015). In Section 6, we note that the quasars in our sample are generally hosted in gas- and dust-poor galaxies. The CO emission line and millimeter continuum observations of M1540–1453 will reveal the nature of its host galaxy ISM and shed light on the origin of gas detected in 21 cm absorption.

In this section, we constrain the occurrence of intervening H i 21 cm absorbers at z > 2 using a blind spectroscopic search. We also use Lyα and metal absorption lines detected in our SALT spectra to interpret these statistics.

5.1. Blind H i 21 cm Absorption Line Search

To estimate the incidence of intervening absorbers, we first determine the sensitivity function, $g({ \mathcal T },z)$, as a function of integrated optical depth (${ \mathcal T }$) and redshift (z). For this, we follow the formalism provided in Gupta et al. (2021), which takes into account the varying optical depth sensitivity across the spectrum. The two crucial inputs required to determine $g({ \mathcal T },z)$ are spectral weight (W) and completeness fraction (C). The former accounts for the possibility that some of the targets in the sample may not have spectroscopic redshifts. Since all of the targets in our sample have spectroscopic redshifts, we assign W = 1.

The completeness fraction accounts for the detectability of absorption line features of different line shapes. To determine this, we consider the absorption profiles of all of the intervening absorbers detected from our surveys in the last 15 yr. We inject 200 single Gaussian components with widths consistent with the distribution of ΔV₉₀ of these absorbers (see Figure 8 of Gupta et al. 2021) at each pixel and apply the detection algorithm described in Section 3.3 to compute $C({{ \mathcal T }}_{j},{z}_{k})$ as

$$\begin{eqnarray}&&C({{ \mathcal T }}_{j},{z}_{k})=\displaystyle \frac{1}{{N}_{{\rm{inj}}}}\sum _{i=1}^{{N}_{{\rm{inj}}}}F({{ \mathcal T }}_{j},{z}_{k},{\rm{\Delta }}{V}_{i}),\end{eqnarray} \tag{ 2 }$$

where N_inj is the number of injected systems, and F = 1 if the injected system is detected and zero if not. The total completeness-corrected redshift path of the survey, $g({{ \mathcal T }}_{j}$), considering all sight lines, is plotted in Figure 11. The redshift path starts falling off rapidly below ${ \mathcal T }=1.5$ km s⁻¹.

Zoom In Zoom Out Reset image size

It is of particular interest to consider the detectability of H i 21 cm absorption in DLAs, i.e., ${ \mathcal T }=1.1$ km s⁻¹ (refer to the vertical dashed line in Figure 11). The sensitivity function providing the number of spectra in which it is possible to detect CNM in DLAs is shown in Figure 12. The total redshift and comoving path length are Δz = 38.3 and ΔX = 130.1, respectively. Then, the incidence or number of 21 cm absorbers per unit redshift and comoving path length are n₂₁ < 0.048 and ℓ₂₁ < 0.014, respectively. These 1σ upper limits are based on small number Poisson statistics (Gehrels 1986).

Zoom In Zoom Out Reset image size

The formalism to search for the H i 21 cm absorption line presented above can also be applied to OH main lines. For the stronger OH main line at 1667.359 MHz, ${ \mathcal T }=1.1$ km s⁻¹ and an excitation temperature of 3.5 K will correspond to N(OH) = 8.6 × 10¹⁴ cm⁻². The total redshift and comoving path length are Δz = 44.9 and ΔX = 167.7, respectively. The number of OH absorbers per unit redshift and comoving path length are n_OH = 0.041 and ℓ_OH < 0.011, respectively.

Besides H i 21 cm absorption, CNM at high z may also be searched using absorption lines of H₂ and C i (e.g., Srianand et al. 2012; Noterdaeme et al. 2018). Recently, Krogager & Noterdaeme (2020), using a canonical two-phase model of atomic gas and the observed statistics of H₂ and C i absorbers at high z, estimated the comoving path length of CNM, n_CNM = 0.012. The upper limit of n₂₁ < 0.048 obtained through our blind survey is consistent with this. This result is the first study comparing the CNM cross section of galaxies at z > 2 estimated using radio and optical/ultraviolet absorption lines. The detectability of H₂ and C i absorption at optical/ultraviolet and H i 21 cm absorption at radio wavelengths are affected by different systematic effects. Indeed, there does not exist a one-to-one correspondence between the presence of H₂ and H i 21 cm absorption, and the difference may be due to the small sizes of H₂-bearing clouds (see Srianand et al. 2012, for a discussion). Much larger radio surveys are needed to disentangle these possibilities.

5.2. Relationship with Lyα and Metal Lines

Our SALT spectra allow us to search for Fe ii λ λ λ2343, 2374, 2383 for z_abs ≤ 2.15 and DLAs at z > 2.65. For this search, we complement our SALT-NOT survey spectra with more sensitive long-slit SALT observations of 25 quasars z_em > 2.7 obtained to search for extended Lyα emission halos associated with powerful radio-loud quasars (see Shukla et al. 2021, for an example). In total, we identify seven DLAs and one strong Fe ii absorber in our sample. The implications of the lack of 21 cm absorption in these high H i column density absorbers are discussed below. In the redshift range 2.15 ≤ z_abs ≤2.65, we also identify 21 C iv absorbers for which neither Fe ii nor DLA can be searched in our spectrum. The redshifted H i 21 cm line frequencies corresponding to these C iv absorbers are unaffected by RFI, and no H i absorption is detected. Note that C iv can trace a wide range of ionization stages and so is not a good indicator of the presence of a DLA or 21 cm absorption. We will use this only as an indicator of the possible presence of multiphase gas along the sight line.

First, we focus on the subset of 23 quasars that have sufficient S/N in the optical continuum and, hence, are suitable to search for Lyα absorption. The absorption profiles of six DLAs detected from this subsample are shown in Figure 13. The measured absorption redshifts are consistent with five of these being intervening systems and the remaining one (z_abs = 2.7613 toward M2154–3826) being a PDLA.

Zoom In Zoom Out Reset image size

We also detect strong Lyα absorption at the systemic redshift of M0507−3624. While we see the evidence of the damping wings and a wide range of absorption from singly ionized species, we also see nonzero flux in the core of the Lyα absorption line. It is possible that this system is similar to the associated H₂-bearing DLAs studied by Noterdaeme et al. (2019), where the presence of flux in the absorption core is related to the partial coverage. Based on damping wings, we estimate N(H i) ∼ 10^20.2 cm⁻². However, with the present spectra, we are unable to confirm the origin of residual flux in the core. A higher spectral resolution data set covering both Lyα and Lyβ absorption is needed to get an accurate estimate of the N(H i) for this absorber. For the purpose of this paper, we consider this as a candidate PDLA.

For a total redshift path of 9.3, the detection of five intervening DLAs corresponds to a number of DLAs per unit redshift, n_DLA = 0.54 ± 0.24. This is slightly higher but, due to large uncertainties, consistent with the measurement of 0.24 ± 0.02 based on SDSS DLAs by Noterdaeme et al. (2012) and ${0.26}_{-0.05}^{+0.06}$ toward radio-loud quasars based on the combined CORALS and UCSD samples by Jorgenson et al. (2006). Since the quasars in our sample are fainter than previous surveys (see Figure 1), it is also possible that there is indeed a dependence between n_DLA and the faintness of quasars as noted by Ellison et al. (2001) for the CORALS sample.

As discussed in the next section, in comparison to associated H i 21 cm absorption detection rates in low-z AGN, the detection of just two PDLAs and one associated H i 21 cm absorber (M1540–1453) from our sample may seem surprisingly low. But actually, the detection of three PDLAs from our sample is in fact a factor of 3 larger than what we expect from the statistics of PDLAs observed at z ∼ 3 in SDSS (Prochaska et al. 2008a). Interestingly, from the statistics of damped H₂ absorption lines, Noterdaeme et al. (2019) suggested that the PDLA fraction in Prochaska et al. (2008b) may have been underestimated. A complete search of Lyα absorption toward all of the targets in our sample will confirm the abovementioned excesses of DLAs and PDLAs. Specifically, from the observations of all of the sources (Δz ∼ 60), we expect to detect another ∼30 DLAs.

Using SALT 1D spectra, we measure redshift, N(H i), and rest equivalent widths of metal absorption lines corresponding to these DLAs. These measurements are provided in Table 4. The quoted error in N(H i) also includes continuum placement uncertainties. The single-component Voigt profile fits to the DLAs are shown in Figure 13. The rest equivalent widths of the Si ii λ1526 lines are provided in column (4) of Table 4. The metallicities inferred using the W(Si ii) and metallicity correlation (see Equation (1) of Prochaska et al. 2008a) are given in column (5). We also estimated metallicity using weak transitions of Si ii or S ii. These are provided in column (6). For the z_abs = 2.7613 PDLA toward M2154−3826, we detect several weak transitions of Fe ii, Ni ii, and Si ii. We used a single-component curve of growth to measure metallicities in this case. Overall, the metallicities of these DLAs are typically less than a tenth of the solar metallicity. However, given the poor spectral resolution of our data, these estimates may suffer from hidden saturation effects.

Table 4. Properties of the DLAs Derived from Our Observations

QSO	zabs	log N(H i)	W(Si ii)	Z(P08)	Z	∫τ21 dv	Ts
			(Å)			(km s−1)	(K)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
M0416–3339	2.8561	20.50 ± 0.10	0.12 ± 0.03	−2.2	≤−2.3 a	RFI	⋯
M0610–2301	2.3987	20.85 ± 0.10	0.66 ± 0.02	−1.2	≤−1.1 b	≤0.64	≥603
M1013–2546	2.6834	20.35 ± 0.15	≤0.90	⋯	⋯	≤0.72	≥169
M1351–0129	2.7719	20.50 ± 0.10	0.19 ± 0.03	−1.9	≤−1.2 c	RFI	⋯
M1619–0939	2.7934	20.55 ± 0.10	1.03 ± 0.05	−0.9	≤−1.9 c	RFI	⋯
M2154–3826	2.7613	21.00 ± 0.10	0.53 ± 0.01	−1.3	−1.60 ± 0.11 d	⋯	⋯

Notes. Column (1): source name. Columns (2) and (3): DLA redshift and H i column density. Column (4): Si ii equivalent width. Column (5): metallicity inferred using W(Si ii) and the correlation from Prochaska et al. (2008a). Column (6): metallicity measured using weak Si ii or S i lines. Column (7): 5σ 21 cm optical depth limit, considering Δv = 25 km s⁻¹.

^a Based on Si ii λ1301 line. ^b Based on S ii λ1250 line. ^c Based on Si ii λ1808 line. ^d Using the curve-of-growth analysis.

Download table as:  ASCIITypeset image

Unfortunately, strong persistent RFI at 360–380 MHz prevents a H i 21 cm line search at z = 2.73–2.94 (see Figure 12). Thus, we could observe the 21 cm line only for the z_abs = 2.3987 DLA toward M0610–2301 and z_abs = 2.6834 toward M1013–2546. We do not detect 21 cm absorption from these systems. The 5σ integrated optical depths are provided in column (7) of Table 4. Object M0610–2301 is a compact radio source in the modest-quality VLASS quick-look image. The deconvolved source size is 13 × 01 with a position angle of 180° (i.e., size < 11 kpc at z_abs). Object M1013–2546 also appears to be a core-dominated source. Thus, we assume complete coverage, i.e., f_c = 1. This, together with the observed N(H i), translates to a lower limit on the spin temperature, T_S ≥ 603 K for M0610–2301 and ≥169 K for M1013–2546. These limiting values of the spin temperatures are higher than the measured median N(H i) weighted T_S of 70 K for the CNM in our galaxy (see Heiles & Troland 2003). We note that H i 21 cm observations of 23 DLAs from radio-selected samples of CORALS, UCSD, and Ellison et al. (2008) are available in the literature (Srianand et al. 2012; Kanekar et al. 2014). The overall 21 cm absorption detection rate of 3/25 (12${}_{-7}^{+11}$%) is consistent with the measurements from optically selected samples and the conclusion that DLAs at z > 2 are predominantly warm and tend to show high spin temperatures (see also Petitjean et al. 2000). Much larger radio-selected surveys (ΔX ≳ 10⁴) are needed to uncover the population of dusty DLAs.

Our SALT spectra also allow us to detect an Fe ii λ2383 line for z_abs ≤ 2.15. We detect a strong Fe ii absorber at z_abs = 2.1404 toward M0652–3230. This system also shows absorption lines from other singly ionized species (i.e., Si ii and Al ii). All of these suggest high N(H i) (Dutta et al. 2017b). But, as can be seen from Figure 3, the radio emission is dominated by the double-lobe structure. The separation between the two lobes is ∼27'' (250 kpc at z_abs). Therefore, the radio and optical sight lines are well separated. This explains the nondetection of 21 cm absorption at the redshift of the Fe ii absorbers.

We searched for H i 21 cm absorption within 3000 km s⁻¹ of the quasar emission line redshift. In 28/88 cases (32%), the redshifted frequency is affected by RFI. For the remaining 60 sources, the distributions of redshift and the 5σ 21 cm optical depth limit for a width of 25 km s⁻¹ are shown in Figure 14. The median redshift of this subsample, which includes M1312–2026 (z = 5.064), is 2.288. For sources with resolved morphology (Figure 3), we have searched for absorption toward multiple components but consider only the optical depth limit for the strongest component for statistics.

Zoom In Zoom Out Reset image size

The only 21 cm absorption detection from the survey is described in Section 4. For nondetections, the optical depth limits span a range of 0.103–3.6 km s⁻¹. The detection rate of ${1.6}_{-1.4}^{+3.8}$% without any sensitivity cut is among the lowest for a 21 cm absorption line survey. If we adopt the detection rate based on low-z surveys, we would expect to detect approximately 20 H i 21 cm absorbers from our sample. For example, Maccagni et al. (2017) reported a 27% ± 5% detection rate from the observations of 248 AGN at 0.02 < z < 0.25. The 3σ peak optical depth (resolution ∼16 km s⁻¹) limits are typically between 0.005 and 0.2 (median ∼0.04). For the 25 km s⁻¹ width adopted by us, the range corresponds to a 5σ integrated optical depth, ${ \mathcal T }$, of 0.2–7.1 km s⁻¹ (median ∼1.41). In our survey, we achieve a ${ \mathcal T }$ of better than 1.1 km s⁻¹ in 90% of the cases. Thus, the stark contrast in detection rates is not due to differences in the sensitivity.

At low-z, the vast majority of targets have a WISE color W1 − W2, which is an indicator of the AGN accretion rate, of less than 0.6. Among these, the higher fraction of detections are associated with larger W2 − W3 colors, which is an indicator of SFR (Glowacki et al. 2017; Chandola et al. 2020). The compact radio sources still embedded within such gas- and dust-rich environments are well placed to exhibit 21 cm detections. The powerful quasars identified by our MIR wedge (see Equation (1)) do not overlap with the abovementioned low-z AGN in color space defined by W1 − W2 and W2 − W3. Thus, we hypothesize that the low nondetection rate in our sample is a consequence of the gas-poor environment of the host galaxies. Since the detectability of H i 21 cm absorption also depends on the thermal state of the gas and the morphology of the radio emission, we validate this hypothesis through the Lyα and metal lines covered in our SALT-NOT spectra.

First, we consider the following four sources ¹² at 2.7 < z < 3.5: M0507–3624, M1013–2546, M1351–1019, and M2047–1841. For these, we have access to both Lyα and metal absorption lines, and the 21 cm frequency is not affected by RFI. In the cases of M1013–2546 and M2047–1841, we do not detect a PDLA or any associated C iv absorption system. This suggests the absence of neutral and metal-enriched ionized gas along the sight lines. In the case of M0507–3624, we identified a potential PDLA. For N(H i) ∼ 10^20.2 cm⁻² inferred from the damping wing of Lyα absorption and H i 21 cm nondetection, we place a lower limit of 216 K for the spin temperature (f_c = 1). In the case of M1351–1019, while Lyα absorption is not prominent, we do detect associated C iv absorption. In this case, we also detect extended Lyα emission. All of this suggests that these two sources are associated with a gas-rich environment. Therefore, the lack of H i 21 cm absorption here may just be due to a lower CNM filling factor.

In general, 21 quasars, i.e., ∼23%, in our sample show associated C iv absorption within 3000 km s⁻¹ of z_em. In 11 of these cases, the 21 cm absorption could not be searched due to strong RFI. Among the remaining 10, we also detect H i Lyα absorption in four cases, but none are DLAs. In the remaining six cases, the Lyα absorption is not covered in the spectra, but we do not detect corresponding absorption due to any low-ionization species such as Fe ii (in four cases) and Si ii (in six cases). Since C iv may come from a wide range of ionization stages, the absence of strong Lyα and low-ionization absorption lines indicate the lack of sufficient high column density neutral gas along the line of sight.

From the above, we conclude that the high fraction of quasars in our sample are indeed residing in gas- and dust-poor environments. An interesting counterpoint to the lack of H i 21 cm absorption in our sample is provided by the 100% detection rate of molecular gas through CO emission in a sample of eight hyperluminous WISE/SDSS (WISSH) quasars at z ∼ 2–4 (Bischetti et al. 2021). We note that a majority (6/8) of these would be selected by our MIR wedge (Equation (1)). But WISSH quasars are much brighter (1.5 Vega magnitude compared to our sample) in the W4 band (22 μm) of WISE. The deliberate selection of WISSH quasars as the most luminous infrared sources ensures that they are being observed through dust clouds (Weedman et al. 2012) and perhaps represent an early phase in the evolution of the quasar. Only ∼10% of the quasars in our sample have W4 magnitudes comparable to the abovementioned WISSH quasars with CO detections, and only in three cases is ${ \mathcal T }\lt 1.1$ km s⁻¹; i.e., the sensitivity to detect CNM in N(H i) > 10²⁰ cm⁻² is achieved. Although this may seem to conflict with our MIR-selected sample strategy, luminous quasars only spend a small fraction of their total lifetime (∼10⁷ yr; Martini & Weinberg 2001) in the dust-obscured phase. Therefore, the the representation of dust-obscured quasars in our unbiased sample will also be proportionately small. Considering only the AGN with the sensitivity to detect CNM in N(H i) > 10²⁰ cm⁻² gas, we estimate the CNM covering factor in the unobscured phase of quasars with radio luminosity L_{1.4 GHz} ≃ 10^27−29.3 W Hz⁻¹ to be 0.02.

Although the most straightforward explanation for the low detection rate in our sample is the gas- and dust-poor environment at host galaxy scales, the detectability of gas towards the AGN may also be influenced by additional factors, such as high intrinsic radio or ultraviolet luminosities of AGN (Curran & Whiting 2010; Aditya et al. 2016; Grasha et al. 2019) and the distribution of gas at nuclear scales (Gupta & Saikia 2006). Since high L_UV merely helps to select core-dominated AGN, these two are in fact interlinked in the context of the AGN unification scheme. We estimate the 912Å luminosities of the AGN searched for associated H i 21 cm absorption by interpolating the photometry from PS1. The distribution of L₉₁₂ is shown in the bottom panel of Figure 14. Clearly, the majority of objects have L₉₁₂ > 10²³ W Hz⁻¹, where the associated H i 21 cm absorption is rarely detected.

Unfortunately, none of the CO-detected quasars from Bischetti et al. (2021) are bright enough at radio wavelengths to search for H i 21 cm absorption. Interestingly, their available SDSS spectra do not show the presence of high column density neutral gas, i.e., a PDLA along the quasar sight line (although ionized gas outflows are present). Thus, although the molecular gas is distributed in rotating disks (extent 1.7–10 kpc; Bischetti et al. 2021), it is oriented such that the cold gas cross section toward the quasar sight line is minimal. A spatially extended few kiloparsec–sized radio source embedded in such an environment may have still shown H i 21 cm absorption (e.g., refer to the cases of B2 0902+345 and MG J0414+0534 in Section 4).

Finally, we note the nondetection toward the highest-redshift quasar (z_em = 5.062) in our sample: M1312–2026 (Carilli et al. 2007). This brightest radio-loud quasar at z > 5 has a radio-loudness parameter of R = f_{ν,5 GHz}/f_ν,4400Å = 1.4 × 10⁴. This R value is an order of magnitude greater than that of any other z > 5 AGN known to date (Momjian et al. 2018; Saxena et al. 2018). The host galaxies of quasars at such high redshifts can be associated with large amounts of dust and molecular gas (>10¹⁰ M_⊙) and high inferred SFRs (>100 M_⊙ yr⁻¹; Venemans et al. 2017; Decarli et al. 2018; Feruglio et al. 2018). Our H i 21 cm nondetection corresponds to a 5σ upper limit of N(H i) < 4 × 10¹⁹ cm⁻² (T = 100 K; f_c = 1 assumed) but can miss narrow absorption components due to the RFI. The current SALT spectrum also covers only Lyα emission. Further investigation of the nature of this very intriguing nondetection requires IR spectra and subarcsecond-scale radio imaging, which are in progress.

This paper described a spectroscopically blind search for H i 21 cm absorption lines in the wideband uGMRT spectra of 88 AGN at 2 < z < 5.1. We also applied the same formalism to constrain the occurrence of intervening OH 18 cm main lines. We show that compared to previous radio-selected samples of quasars to search for DLAs, our sample for the uGMRT survey has targeted fainter objects (median i = 19.5 mag; see Figure 2) and is a close representation of the underlying luminosity function of quasars. Thus, our dust-unbiased sample of AGN with a median radio spectral index, ${\alpha }_{0.4}^{1.4}=-0.38$; redshift, z = 2.5; and spectral luminosity, L_{1.4 GHz} = 10^27−29.3 W Hz⁻¹, is ideally suited to determine the occurrence of cold atomic gas (T ∼ 100 K) toward powerful quasars at z > 2.

Through a spectroscopically blind search of absorption lines in all of the uGMRT spectra, we detected one new associated H i absorption toward M1540–1453 at z_abs = 2.1139. No intervening H i 21 cm absorption line is detected. Our detection is only the fourth associated H i 21 cm absorber known at z > 2. It has a H i column density of $(2.06\pm 0.01)\times {10}^{21}\left(\tfrac{{T}_{S}}{100}\right)\left(\tfrac{1}{{f}_{c}}\right)$ cm⁻². In our SALT spectrum, the peak of C iv emission and the low-ionization metal absorption lines are coincident with that of the 21 cm absorption. The overall properties of the 21 cm absorption are consistent with it originating from a circumnuclear disk or gas clouds associated with the host galaxy. The CO emission line observations and optical spectra covering the Lyα absorption (i.e., λ ∼ 3785 Å), along with the subarcsecond-scale imaging, will allow us to understand the origin of cold gas detected in 21 cm absorption.

Our survey is sensitive to detecting CNM in DLAs corresponding to a total redshift and comoving path length of Δz = 38.3 and ΔX = 130.1, respectively. Using this, we constrain the incidence or number of 21 cm absorbers per unit redshift and comoving path length to be n₂₁ < 0.048 and ℓ₂₁ < 0.014, respectively. The same formalism applied to the OH main line at 1667.359 MHz corresponds to a total redshift and comoving path length of Δz = 44.9 and ΔX = 167.7, respectively. The number of OH absorbers per unit redshift and comoving path length are n_OH < 0.041 and ℓ_OH < 0.011, respectively. We note that the number of DLAs per unit redshift interval, i.e., n_DLA(z) at 2.3 ≤ z ≤ 2.9, is in the range of 0.21–0.29 (Noterdaeme et al. 2012). This implies that the covering factor of CNM gas in DLAs is ≤20%. These upper limits are also consistent with n_CNM = 0.012 estimated using H₂ and C i absorbers, also tracers of cold gas, at high-z (Krogager & Noterdaeme 2020). Our result shows that a moderately larger survey, such as MALS with ΔX ≳ 10⁴, is important to precisely characterize the CNM fraction and its redshift evolution at high-z.

The low-z AGN (z < 0.25) exhibit H i 21 cm absorption detection rates of ∼30% (e.g., Maccagni et al. 2017). Comparing to this the low associated H i 21 cm absorption detection rate (${1.6}_{-1.4}^{+3.8}$%) and the CNM filling factor of 0.2 from our survey is intriguing. We show that this is most likely due to the fact that the powerful quasars in our sample are residing in gas- and dust-poor environments, and that luminous quasars only spend a small fraction of their total lifetime in the dust-obscured phase. We use the spectral coverage of Lyα and various metal absorption lines in our optical spectra to confirm the absence of high column density atomic gas toward the quasar sight lines.

From our SALT spectra, we report detections of five intervening DLAs and two PDLAs in our sample. The measured number of DLAs per unit redshift, n_DLA = 0.54 ± 0.24, is slightly higher but, due to large uncertainties, consistent with the measurement based on SDSS DLAs (Noterdaeme et al. 2012) and the combined CORALS and UCSD sample of radio-selected quasars (Jorgenson et al. 2006). Interestingly, the PDLA detection fraction is also a factor of 3 larger. Since the quasars in our sample are fainter than in the previous surveys, there may indeed be a dependence between n_DLA and the optical faintness of the quasars (Ellison et al. 2001). These results also underline the need for larger surveys of dust-unbiased DLAs. Due to limited spectral coverage, we could search for Lyα in only 30% of our SALT-NOT sample presented here. A complete search of Lyα absorption toward all of the targets will allow us to examine the abovementioned excesses at a higher significance level.

Eventually, much larger radio-selected surveys (ΔX ≳ 10⁴) such as MALS are needed to uncover the population of dusty DLAs. The key science objectives are summarized in Gupta et al. (2017), and the survey is well underway. The first L- and UHF-band spectra based on the science verification data are presented in Gupta et al. (2021) and Combes et al. (2021), respectively. Each MALS pointing is centered at a bright (>200 mJy at 1 GHz) radio source. Through wideband spectra of the central bright radio source and the numerous off-axis radio sources, it will sample the column density distribution (N(H i) > 5 × 10¹⁹ cm⁻²; T_s = 100 K) relevant to characterize the cross section of cold atomic gas in and around normal galaxies and AGN at 0 < z < 1.4. Simultaneously, it will also be sensitive to detecting OH main-line absorption at z < 1.9 in gas with N(OH) > 2.4 × 10¹⁴ cm⁻² (excitation temperature = 3.5 K). Since the formation of OH is tightly coupled to H₂, the measurement of the OH cross section at z < 2 will be a crucial input through which to understand the redshift evolution of the CNM cross section (Balashev et al.2021).

The work presented here is the first to examine the detectability of high-N(H i) absorbers in an MIR-selected sample of quasars with searches based on radio/optically selected quasars. Although only one associated H i 21 cm absorber is detected, the slight excess of DLAs and PDLAs in our sample is encouraging and with larger surveys could potentially lead to estimates of the dusty AGN missed in optically selected surveys (the fraction could be anywhere between 10% and 50%; Richards et al. 2003; Glikman et al. 2012). Some of the ongoing work we are pursuing is to understand the impact of MIR selection on the colors of selected quasars. In the near future, we will also present the complete census of DLAs and PDLAs in our sample and results from the targeted deeper H i 21 cm line observations toward these. As such, this paper presents a crucial step toward a detailed understanding of the high-N(H i) searches at z > 2 to be eventually taken up with the low-frequency component of the SKA (Morganti et al. 2015).

We thank the anonymous referee for helpful suggestions. We thank the GMRT staff for their support during the observations. The GMRT is run by the National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research. This work is based on observations made with SALT and NOT. The uGMRT data were processed using the MALS data processing facility at IUCAA. The CASA package is developed by an international consortium of scientists based at the National Radio Astronomical Observatory (NRAO), the European Southern Observatory (ESO), the National Astronomical Observatory of Japan (NAOJ), the Academia Sinica Institute of Astronomy and Astrophysics (ASIAA), the CSIRO division for Astronomy and Space Science (CASS), and the Netherlands Institute for Radio Astronomy (ASTRON) under the guidance of NRAO. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.

Facilities: NOT - Nordic Optical Telescope, SALT, and uGMRT. -

Software: ARTIP (Gupta et al. 2021), Astropy (Astropy Collaboration et al. 2013, 2018), CASA (McMullin et al. 2007), and Matplotlib (Hunter 2007).