- Split View
-
Views
-
Cite
Cite
Arnab Chakraborty, Nirupam Roy, Detection of H i 21 cm emission from a strongly lensed galaxy at z ∼ 1.3, Monthly Notices of the Royal Astronomical Society, Volume 519, Issue 3, March 2023, Pages 4074–4081, https://doi.org/10.1093/mnras/stac3696
- Share Icon Share
ABSTRACT
We report the first 5σ detection of H i 21 cm emission from a star-forming galaxy at redshift z ∼ 1.3 (nearly 9 billion years ago) using upgraded Giant Metrewave Radio Telescope (uGMRT). This is the highest redshift H i detection in emission from an individual galaxy to date. The emission is strongly boosted by the gravitational lens, an early-type elliptical galaxy, at redshift z ∼ 0.13. The measured H i mass of the galaxy is |$M_{\rm H\, \small {\rm I}} = (0.90 \pm 0.14 \pm 0.05) \times 10^{10}\, \mathrm{M}_{\odot }$|, which is almost twice the inferred stellar mass of the galaxy, indicating an extended structure of the H i gas inside the galaxy. By fitting 2D Gaussian to the H i signal at the peak of the spectral line, we find the source to be marginally resolved with the position angle consistent with the emission being tangential to the critical curve of the lens mass distribution. This indicates that the solid angle of the approaching H i line flux comes very close to the inner lens caustic and results in very high magnification. These results, for the first time, demonstrate the feasibility of observing high-redshift H i in a lensed system with the modest amount of telescope time and open up exciting new possibilities for probing the cosmic evolution of neutral gas with existing and upcoming low-frequency radio telescopes in the near future.
1 INTRODUCTION
The reservoir of cold atomic neutral hydrogen (H i) gas provides the basic fuel for star formation in a galaxy. Understanding the evolution of galaxies over cosmic time requires knowledge of the cosmic evolution of this neutral gas. A detailed study of star-formation history, over the last decade, shows that the co-moving star-formation rate (SFR) density rises from z ∼ 8 to z ∼ 3 − 4, shows a peak, and remains flat in z ∼ 3 − 1, and then declines by an order of magnitude, from z ∼ 1 to the present epoch (e.g. Le Floc’h et al. 2005; Hopkins & Beacom 2006; Bouwens et al. 2009; Madau & Dickinson 2014). Also the nature of galaxies undergoing star formation evolves significantly from z ∼ 2 to the present epoch (Cowie et al. 1996). At the peak of star formation (z ∼ 1 − 3), the SFR density is dominated by massive galaxies with high SFRs, whereas in the local universe (z ∼ 0) it mostly arises in low mass systems with low SFRs (Le Floc’h et al. 2005). However, the neutral H i mass density (|$\Omega _{\mathrm{H\small {\rm I}}}$|) does not show any significant evolution over cosmic time (Chowdhury et al. 2020; CHIME Collaboration et al. 2022). The molecular hydrogen (H2) density also similar to SFR density shows a peak around z ∼ 1.5 and then declines by one order of magnitude to present day (Walter et al. 2020). On a contrary, the stellar mass density is increasing continuously with cosmic time and surpasses the total gas density (H i and H2) at redshift z ∼ 1.5 (Walter et al. 2020). This opposite nature of gas density and stellar mass density is puzzling and likely to be explained by the infall of ionized gas from intergalactic medium (IGM) or circumgalactic medium (CGM) to the H i reservoir and subsequent conversion of H i to H2 (Walter et al. 2020). Chowdhury et al. (2020) shows that accretion of gas onto galaxies at z ≤ 1 may have been insufficient to sustain high SFRs in star-forming galaxies and likely to be the cause of the decline in the cosmic SFR density at redshifts below one. However, the evolution of cold neutral gas during z ∼ 0 − 3 still needs to be constrained with more sensitive observation to understand the global flow of gas onto galaxies and to probe the history of star formation in the Universe. Hence the knowledge of H i mass of different types of galaxies and the relation between the atomic and molecular gas mass, stellar mass, and the SFR, is critical to study galaxy evolution.
The best way to probe the neutral atomic gas content in a galaxy is via the H i 21 cm spectral line emission. However, the probability for spontaneous emission of H i 21 cm radiation, due to the spin-flip transition between hyperfine states in the ground state of neutral hydrogen, is extremely low. Due to this, it is challenging to detect the H i line emission from individual galaxies beyond a redshift of about 0.4, with modern telescopes. The highest redshift detection of H i line emission to date from an individual galaxy was made at z = 0.376 (Fernández et al. 2016). The average properties of H i content of galaxies can be obtained by ‘stacking’ the H i line emission signals of a large number of galaxies with known spectroscopic redshifts (Chengalur, Braun & Wieringa 2001; Zwaan 2000). This method has been used to measure the average H i mass and cosmological H i mass density at higher redshifts, z ≳ 1 (Chowdhury et al. 2020, 2021), but it is not possible to measure the properties of individual sources via ‘stacking’. The measurement of H i masses of individual galaxies at z ≳ 1 would require a long integration time with today’s radio telescope or the large collecting area of the Square Kilometer Array (SKA).
The strong gravitational lens, nature’s gift, magnifies the weak emission signal coming from distant objects, enabling us to peer through the high-redshift universe. Strong gravitational lensing phenomenon can significantly amplify the faint signal, enabling us to detect the H i signal from galaxies at higher redshifts in a reasonable observation time. At small wavelengths (mm-wave), the amplification of the faint signal from a distant galaxy through gravitational lensing has been used to observe the high-redshift universe (e.g. Brown & Vanden Bout 1991; Vieira et al. 2013). Hunt, Pisano & Edel (2016) made the first attempt to detect the H i 21 cm signal from a lensed galaxy at z ∼ 0.4. They tried to detect the H i signal from three lensed galaxies, two at z = 0.398 and one at z = 0.487, using the Green Bank Telescope (GBT). The background galaxies are lensed by the foreground cluster Abell 773. However, they did not detect the signal and have reported a 3σ upper limit on the H i mass of the galaxies (Hunt et al. 2016). Blecher et al. (2019) have also tried to detect the H i signal from a galaxy at z ∼ 0.4, using gravitational lensing. They also did not detect the signal with a high signal-to-noise ratio. They estimate the H i mass, using Bayesian formalism, from the integrated H i spectrum. There is no strongly lensed H i detection in emission with high statistical significance to date (Blecher et al. 2019; Hunt et al. 2016).
In this paper, we report the first detection of H i 21 cm line emission from a galaxy, a galaxy–galaxy strong lens candidate at z ∼ 1.3, which is detected in Sloan Lens ACS (SLACS) Survey for the Masses (S4TM) Survey (Shu et al. 2017), using Giant Metrewave Radio Telescope (GMRT) archival data. This is the highest redshift H i detection (lookback time ∼ 9 Gyr) from an individual galaxy to date. The paper is organized as follows: we mention the details of the target galaxy in Section 2, we describe our data analysis and estimated H i spectrum in Section 3, the estimation of H i mass, atomic-to-stellar mass ratio and the extension of the H i emission are mentioned in Section 4, finally, we conclude in Section 7. Throughout this work, we use the Planck 2015 cosmological parameters (Planck Collaboration et al. 2016).
2 DETAILS OF THE TARGET GALAXY
The target source was selected from the catalogue of the galaxy–galaxy strong-lens candidates, detected in the SLACS S4TM Survey (Shu et al. 2017). The S4TM survey was designed to identify low to intermediate-mass Early-type galaxies (ETG), which act as a strong lens system. The S4TM survey detected 118 nearly strong lens candidates, selected spectroscopically from the galaxy spectrum data base of the seventh and final data release of the Sloan Digital Sky Survey (SDSS). The basic method to select such lens candidates is to search for multiple nebular emission lines in the spectrum coming from a common redshift, which is significantly higher than the redshift of the foreground lensing candidate. This indicates that there are two objects within the same lightcone of diameter 3 arcsec (diameter of the optical fiber) and a lensing event happened (Bolton et al. 2004). The candidates were further observed with Hubble Space Telescope (HST) in the F814W-band. Shu et al. (2017) modelled the foreground light of the lens galaxy with an elliptical radial B-spline model and subtracted it from the image. The residual image was then inspected for lens morphology, multiplicity, and lens grade. Any candidate was classified as a grade-A strong lens, if definite multiple lensed images were detected (Shu et al. 2015). There were 40 such grade-A strong lens candidates discovered for the first time (Shu et al. 2017). These grade-A lens candidates were modelled as singular isothermal ellipsoid (SIE) profile (Kormann, Schneider & Bartelmann 1994) and the background source light distribution was modelled as multiple elliptical Sersic components. Using the magnification factor, redshift, and the luminosity distance, we found out the value of the quantity |$(\tfrac{D_{L}^{2}}{\mu _{\rm H\, \small {\rm I}}(1+z)})$| for each of the candidate source. The velocity integrated H i flux is inversely proportional to this quantity for a fixed H i mass (see equation 6). Hence, the H i signal from a source galaxy is more likely to be detected if this quantity is less for that source. Then we ranked these candidates based on this value, that is, a source is more likely to be detected with a moderate telescope time if the value of this quantity is smaller for a fixed H i mass. Then we searched the archival data of GMRT for the first five sources from this list and found that the top-ranked candidate was observed with uGMRT in cycle-34 (Proposal code- 34_066). For this target system, the lensing galaxy (SDSSJ0826+5630) shows an average optical magnification of about 105 and is situated at zL = 0.1318, whereas the background source galaxy was at zs = 1.2907. This galaxy–galaxy lens system shows a nearly full Einstein ring with a radius of about 1.02″ (Shu et al. 2017). This is a very promising source because (i) the target source galaxy shows multiple nebular line emissions, which signifies that it is a star-forming galaxy. Also, the redshift of the source galaxy (zS ∼ 1.3) is close to the peak of the SFR density. Hence, the galaxy is expected to have a significant cold neutral gas reservoir, (ii) an extremely strong optical magnification suggests that the H i magnification will also be large and there will be a strong boost of the H i flux coming from this distant background source galaxy. Hence, it may be possible to detect the H i 21 cm emission from the background galaxy with moderate observing time, although the source galaxy resides at a large cosmological distance from us, (iii) the mass distribution of the lens is already modelled precisely using deep HST optical data set. Hence, we can use those model parameters for the analysis of the H i magnification. Details of different model parameters are mentioned in Table 1.
Source redshift (zS) | 1.2907 (Shu et al. 2017) |
Lens redshift (zL) | 0.1318 (Shu et al. 2017) |
Optical magnification | 105 (Shu et al. 2017) |
Einstein radius | 1.01″ (Shu et al. 2017) |
Position angle of lens | 82° (Shu et al. 2017) |
H i magnification (|$\mu _{\rm H\, \small {\rm I}}$|) | 29.37 ± 6 |
H i mass (|$M_{\rm H\, \small {\rm I}}$|) | |$(0.90 \pm 0.14 \pm 0.05) \times 10^{10}\, \mathrm{M}_{\odot }$| |
Inferred stellar mass (M*) | |$(0.38 \pm 0.11) \times 10^{10}\, \mathrm{M}_{\odot }$| |
|$M_{\rm H\, \small {\rm I}}/M_{*}$| | 2.37 ± 0.14 |
Source redshift (zS) | 1.2907 (Shu et al. 2017) |
Lens redshift (zL) | 0.1318 (Shu et al. 2017) |
Optical magnification | 105 (Shu et al. 2017) |
Einstein radius | 1.01″ (Shu et al. 2017) |
Position angle of lens | 82° (Shu et al. 2017) |
H i magnification (|$\mu _{\rm H\, \small {\rm I}}$|) | 29.37 ± 6 |
H i mass (|$M_{\rm H\, \small {\rm I}}$|) | |$(0.90 \pm 0.14 \pm 0.05) \times 10^{10}\, \mathrm{M}_{\odot }$| |
Inferred stellar mass (M*) | |$(0.38 \pm 0.11) \times 10^{10}\, \mathrm{M}_{\odot }$| |
|$M_{\rm H\, \small {\rm I}}/M_{*}$| | 2.37 ± 0.14 |
Source redshift (zS) | 1.2907 (Shu et al. 2017) |
Lens redshift (zL) | 0.1318 (Shu et al. 2017) |
Optical magnification | 105 (Shu et al. 2017) |
Einstein radius | 1.01″ (Shu et al. 2017) |
Position angle of lens | 82° (Shu et al. 2017) |
H i magnification (|$\mu _{\rm H\, \small {\rm I}}$|) | 29.37 ± 6 |
H i mass (|$M_{\rm H\, \small {\rm I}}$|) | |$(0.90 \pm 0.14 \pm 0.05) \times 10^{10}\, \mathrm{M}_{\odot }$| |
Inferred stellar mass (M*) | |$(0.38 \pm 0.11) \times 10^{10}\, \mathrm{M}_{\odot }$| |
|$M_{\rm H\, \small {\rm I}}/M_{*}$| | 2.37 ± 0.14 |
Source redshift (zS) | 1.2907 (Shu et al. 2017) |
Lens redshift (zL) | 0.1318 (Shu et al. 2017) |
Optical magnification | 105 (Shu et al. 2017) |
Einstein radius | 1.01″ (Shu et al. 2017) |
Position angle of lens | 82° (Shu et al. 2017) |
H i magnification (|$\mu _{\rm H\, \small {\rm I}}$|) | 29.37 ± 6 |
H i mass (|$M_{\rm H\, \small {\rm I}}$|) | |$(0.90 \pm 0.14 \pm 0.05) \times 10^{10}\, \mathrm{M}_{\odot }$| |
Inferred stellar mass (M*) | |$(0.38 \pm 0.11) \times 10^{10}\, \mathrm{M}_{\odot }$| |
|$M_{\rm H\, \small {\rm I}}/M_{*}$| | 2.37 ± 0.14 |
3 OBSERVATION, DATA ANALYSIS, AND RESULTS
The galaxy, with pointing centre at |$\rm RA =08^{h}26^{m}39.858^{s},\rm DEC =56^{\circ }30^{\prime }35.97^{\prime \prime }$|, was observed with uGMRT Band-4 receivers for a total of 18 h on-source time. A bandwidth of 100 MHz, sub-divided into 2 048 channels, was used for the observation with GMRT wideband backend (GWB) as the correlator. The frequency coverage was 550–650 MHz, with a velocity resolution of ∼24 km s–1. The integration time per visibility point was 5 s. The standard calibrators, 3C147 and 3C286, were observed to calibrate the flux density scale, while regular observations of the nearby compact source 0834+555 were used to calibrate the complex antenna gains.
The data were first inspected using aoflagger1 package for the detection and excision of radio frequency interference (RFI) (Offringa, van de Gronde & Roerdink 2012). We used a casa-based flagging and calibration pipeline to solve the complex gains and to remove any remaining bad data, following the standard procedure. We used the automatic algorithms, tfcrop and rflag, within casa’s flagdata task to identify and remove the RFI. The flux density of the primary calibrators, 3C147 and 3C286, were set using the Perley–Butler model (Perley & Butler 2017). The delay and bandpass corrections were derived from the observations of the primary calibrators. The time variable complex gains for each antenna were derived from the observation of secondary calibrator 0834+555, which was frequently observed for 2 min for every 15 min scan of the target. Following this, the calibration solutions were applied to the target field and we split the target for imaging and self-calibration. We did not average the data across frequency and time and retain the maximum resolution. This helped us to identify and flag the bad data during self-calibration and imaging loops.
We used wsclean (Offringa et al. 2014) to make the continuum image of the target field. The multiscale wide-band deconvolution along with automasking (Offringa & Smirnov 2017) was performed to capture the variation of sky brightness across this large bandwidth over different spatial scales. We made a large image of size 8192 × 8192 pixels, covering a total field of view of 2.27° × 2.27°, with a pixel size of 1.0″. This large image is required to deconvolve and model the bright confusing sources far away from the first null of the primary beam. We made the first image down to 6σ using the automasking routine of wsclean. The deconvolution was terminated after 50k iterations. Then we created a mask using the first image down to 10σ to remove any spurious features. Then we run another constrained deconvolution using that mask file in order to generate an artifact-free model for self-calibration purposes.
wsclean inverts the frequency-dependent skymodel derived from the deconvolution process into model visibilities at the end of the imaging process, which we used for the self-calibration. We performed several rounds of phase-only self-calibration, with an improved mask at each iteration, until no further improvements were seen in the continuum image. We used Briggs weighting with a robust parameter of -1 during self-calibration and the final continuum image was made using the robust parameter of 0.0 (Briggs 1995). The continuum image was shown in Fig. 1. The off-source RMS noise near the field centre, away from the bright source, is about 8 μJy beam−1 with a synthesized beam width of about 4.5″.
We then subtracted the continuum emission from the calibrated multichannel visibilities using uvsub routine in casa. Then, any residual continuum emission was subtracted via a second order polynomial fit to each visibility spectrum and the residual visibilities were then shifted to barycentric frame, using the mstransform routine in casa. We excluded 50 channels on each side of the central line frequency channel ν = 620.0683 MHz, corresponding to z = 1.2907, during the polynomial fitting to the visibility spectrum. We have also checked the data visually in the time–frequency plane for any residual RFI close to the line-frequency, using rfigui routine of aoflagger, but did not find any such contaminants. The fraction of data lost due to RFI mitigation as a function of frequency is shown in Fig. 2. The vertical black lines show the region, which is being used for the final line–cube analysis. We made a spectral image cube using tclean routine in casa, with natural weighting and w-projection algorithm (Cornwell, Golap & Bhatnagar 2008). We used the baselines |$\lt 18 \rm K \lambda$| and a Gaussian UV-taper at |$12 \rm K \lambda$| during imaging. This gave us the optimal spatial resolution of about 13″ to extract the H i signal with the highest signal-to-noise ratio. This spatial resolution corresponds to the physical size of about 112 kpc at the redshift of the background galaxy. The resulting spectral image cube has a frequency resolution of 48.83 kHz, corresponding to a velocity resolution of 24 km s–1.
We take a cut along the velocity axis at the central peak position of this spectral cube and the resultant spectrum is shown in the left panel of Fig. 3. The black line is the 1σ uncertainty on the spectrum. We performed three different tests to estimate the 1σ RMS noise on the spectrum. First, we take the spectrum corresponding to the line-free channels, that is, by excluding the three channels around the central peak channel as seen in Fig. 3. Then we perform the Anderson–Darling test on this line-free spectrum to check for Gaussianity. The null hypothesis is that the line-free spectrum corresponds to the Gaussian distribution. The estimated p value is 0.16, which shows that the line-free spectrum corresponds to the Gaussian noise distribution and we quote the RMS of this line-free spectrum as 1σ uncertainty here. The 1σ RMS noise is ∼154 μJy beam−1, which is being shown by black line in Fig. 3. Next, we take an off-source region of size 8 times the synthesized beam close to the phase centre and estimate the RMS for this region along the frequency axis. We found that the estimated RMS is consistent with our previous finding. In addition to this, we also take spectrum along 50 arbitrarily chosen line-of-sight through the spectral cube and found that the mean of those spectra is also consistent with our estimation of 1σ uncertainty on the spectrum. It is clear from the left panel of Fig. 3 that the H i 21 cm peak flux corresponding to the central channel (ν = 620.0683 MHz) from the background galaxy (zs ∼ 1.3) is detected at 4σ significance. The average map of the three channels, the central peak channel, and the two neighbouring channels, is shown in the right panel of Fig. 3. The contours are at [− 4, 4, 4.25, 4.5, and 5] × σ levels, where σ = 108 μJy beam−1 is the RMS noise of the channel average image. The synthesized beam of the channel average image, with major and minor axes 13.37″ × 12.32″ and position angle 37°, is shown in the bottom left. The channel average image also shows a clear detection of the H i 21 cm emission signal from the distant background galaxy at 5σ significance.
4 ESTIMATION OF H i MASS
The H i magnification factor is not the same as the optical magnification as reported in Shu et al. (2017). The H i mass and size are tightly correlated and the mass increases linearly with the size of the H i disc (Wang et al. 2016). In general, the distribution of H i in a galaxy is more extended than the stellar component, and as magnification is approximately equal to the lensed to the intrinsic angular size of a source, the H i magnification is typically lower than the optical magnification (Blecher et al. 2019). We performed simulations to estimate the H i magnification factor (|$\mu _{\rm H\, \small {\rm I}}$|) to infer the H i mass from the measurement of lensed velocity integrated H i flux. We followed the methodology presented in Blecher et al. (2019) for the simulation and only briefly mentioned it here.
In our simulation, we sampled |$\log _{\rm 10}(R^{\rm c}_{\rm mol})$| from a normal distribution with [mean, SD] = [−0.1, 0.3], which is consistent with the range of molecular to atomic gas mass ratio, |$M_{\rm H_2}/M_{\rm H\, \small {\rm I}} \sim 0.12-0.32$|, for the stellar mass range |$\rm log_{10} M_{*} \sim 9.18-11.20$| at z = 0 (Catinella et al. 2018). The H i mass is sampled between |$\rm log_{10}M_{\rm H\, \small {\rm I}} \sim 6-12$|, which is consistent with the stellar mass range defined in Maddox et al. (2015). For a given |$M_{\rm H\, \small {\rm I}}$| and |$R^{\rm c}_{\rm mol}$|, we first find out the value of |$D_{\rm H\, \small {\rm I}}$| using equation 3 and then use this to solve for rdisc in equation 1. To incorporate the orientation effects, the simulated H i disc is rotated in a 3D cube to sample the position and inclination angle of the disc. The inclination angle (i) was sampled with probability density function (PDF) of sin(i) over the range [0, π/2], and the position was sampled randomly between [0, π].
The general relativistic ray tracing was performed using the glafic package (Oguri 2010). The position of the centroid of the source galaxy with respect to the lens, called the impact factor, was not known a priori (Shu et al. 2017). We varied the impact factors between [0.0 and 0.5] to yield the published optical magnification.
We ran 104 Monte Carlo simulations, by varying the model parameters, and estimate the magnification factor, which yield the velocity integrated observed H i flux. We found that the H i magnification factor entirely depends on the H i mass and does not show any dependence on |$R^{\rm c}_{\rm mol}$|, inclination angle, and impact factor in our simulations. The magnification is approximately equal to the ratio of the lensed to the intrinsic angular size of the source. As shown in equation 3, the H i mass is a increasing function with the H i size, hence the magnification strongly depends on the H i mass. Note that, Blecher et al. (2019) also found a similar behaviour of magnification in their analysis. Hence, we marginalized over all other nuisance parameters and the 1D PDF of the H i magnification factor (|$\mu _{\rm H\, \small {\rm I}}$|) is shown in Fig. 4. The mean value of |$\mu _{\rm H\, \small {\rm I}}$| with 1σ error bar is 29.37 ± 6. In Fig. 5, we show the data (left panel), the simulated model (middle panel), and the residual (right panel). We see that the simulated model captures the H i emission of the source galaxy accurately.
The most precise measurement of the H i mass function (HIMF) to date using Arecibo Legacy Fast ALFA (ALFALFA) catalogue, at z ∼ 0, shows a power law-like increase towards the lower masses and a sharp exponential decline towards the higher mass end (Jones et al. 2018). This transition to the exponential decrease of the HIMF happened around a ‘knee’ mass, |$M_{\rm \rm knee} = 0.87 \times 10^{10}\, \mathrm{M}_{\odot }$| (Jones et al. 2018). This was found after fitting the Schechter function to the measured HIMF of local galaxies detected in the ALFALFA survey. Hence, the estimated H i mass of the lensed source galaxy in this analysis falls close to the ‘knee’ mass of the HIMF at z = 0.
5 ATOMIC TO STELLAR MASS RATIO
The H i to stellar mass ratio of our source galaxy is |$M_{\rm H\, \small {\rm I}}/M_{*} = 2.37 \pm 0.14$|, suggesting that the cold atomic gas is higher than the stellar component of the galaxy. This result is in agreement with the findings of average H i to stellar mass ratio of about 1.26 ± 0.28 at z ∼ 1 (Chowdhury et al. 2020) and 2.6 ± 0.5 at z ∼ 1.3 (Chowdhury et al. 2021) in star-forming galaxies via stacking. However, this is in clear disagreement with the findings for the local star-forming galaxies with similar stellar mass distribution, where the average H i mass is about 35 per cent of the average stellar mass of galaxies detected in extended GALEX Arecibo SDSS Survey (xGASS) between 0.01 < z < 0.05 (Catinella et al. 2018). There is a significant evolution of stellar mass function from z = 0 to z = 1; however, the predicted HIMF shows negligible evolution in this redshift period (Lagos et al. 2011). This suggests that at a given stellar mass the cold gas reservoir of galaxies will be larger at higher redshifts. The fact that we also found a larger H i mass in comparison with the stellar mass at redshift around 1.3 in the star-forming galaxy, suggests that an evolution of the H i to stellar mass ratio from high redshift to the present epoch in star-forming galaxies.
6 EXTENSION OF H i EMISSION
We fit a 2D Gaussian to the peak channel image, at ν = 620.0683 MHz (z = 1.2907), using imfit task in casa. We found that the fitted major and minor axes are 17.1″ × 12.9″ and the position angle (p.a) is 52° ± 6°. Fig. 6 shows the optical HST image in the F814W-band of the foreground galaxy at zL = 0.13. The orange lines are the [ − 3, 3, 3.5, and 4] × σ contour levels of the peak channel image (ν = 620.0683 MHz), where σ ∼ 154 μJy beam−1 is the RMS noise in the image. We show the central 1.5′ × 1.5′ region (∼770 kpc at z ∼ 1.3) of the peak channel image in the inset. The red-dashed circle around the central foreground galaxy shows the outer critical curve of the lens as estimated from the SIE model of lens mass distribution (Shu et al. 2017). The measured Einstein radius was 1.01″ and the major axis position angle of the SIE component with respect to the north is 82° (Shu et al. 2017). The ellipse in yellow shows the 2D fitted Gaussian to the central channel image, where the fitted centre is marked by a plus sign in yellow at |$\rm RA =08^{h}26^{m}39.90 \pm 0.25^{s},\rm DEC =56^{\circ }30^{\prime }36.88 \pm 1.4^{\prime \prime }$|. At the bottom left, we show the synthesized beam, with major and minor axes are 13.37″ × 12.32″ and position angle is 37°. We found that the fitted source is marginally resolved along the major axis and the size along that axis, deconvolved from the synthesized beam, is 6.8″ and the p.a of the major axis is 52.2° ± 9°. The elongation of the H i emission is tangential to the critical curve of the lens and in a different orientation compared to the orientation of the synthesized beam. We also tried to fit a 2D Gaussian to the averaged image (Fig. 3) and found a similar result as that of the central channel. However, the H i emission did not show any resolved structure when we fit the two nearest neighbour channels to the central channel, and the fitted size of the major and minor axes was the same as that of the synthesized beam.
This is probably due to the fact that the magnification of the central channel is higher than the neighbours. The reason behind this differential magnification is the position of the channel flux solid angle with respect to the caustic. If the H i emission for a narrow channel overlaps with the inner lens caustic, it creates a full Einstein ring and the magnification becomes higher. However, for the neighbouring channels, the H i emission is either approaching or going away from the lens caustic and there is no perfect alignment happened. Due to this, the H i magnification can be lower for those channels and we are unable to see the full H i emission region (see Deane, Obreschkow & Heywood 2015 for a detailed discussion). We also tried with a coarser 35 km s−1 channel resolution, but the final SNR for that was not better than the original 24 km s−1 channel resolution. This suggests that a narrow channel width is optimal for higher SNR detection of the H i line emission, simply due to the larger magnification in some channels (Deane et al. 2015). This indicates that one needs to be careful to select an optimal channel resolution that yields a higher probability of detection.
7 CONCLUSIONS
Strong gravitational lensing helps to study high-redshift galaxies, which can only be possible with next-generation telescopes in the absence of lensing. Here, for the first time, we report the discovery of the H i 21 cm emission signal from a star-forming galaxy at z ∼ 1.3 (nearly 9 billion years ago) using uGMRT via strong gravitational lensing. This opens up a new window to probe the cold neutral gas at high redshifts. We found that the H i lensing magnification is different than the optical magnification and it largely depends upon the H i mass of the source galaxy. Since the H i mass increases with the size of the H i disc, the H i magnification decreases with the H i mass (Blecher et al. 2019). However, the `knee’ of the HIMF might shift to lower masses (Lagos et al. 2011) at higher redshifts and due to which the intrinsic size of the H i disc (see equation 3) is expected to be smaller, resulting into higher magnification factor. Due to this magnification boost, we would expect to detect more lensed H i galaxies at high redshifts in the future.
In the absence of the high signal-to-noise ratio optical data, the stellar mass of the source galaxy can not be estimated. However, we use the relations between dynamical mass and colour and the ratio between dynamical mass and stellar mass of high-redshift galaxies presented in van de Sande et al. (2015) and infer the stellar mass of our source galaxy. We found that the atomic-to-stellar mass ratio is significantly higher than the local star-forming galaxies. Chowdhury et al. (2021), Chowdhury et al. (2020) also found higher atomic-to-stellar mass ratio at high redshifts (z ∼ 1) using stacking of many star-forming galaxies. This indicates that the atomic gas reservoir of high-redshift galaxies is large in these star-forming galaxies.
Deane et al. (2015) shows that the fraction of lensed galaxies out of all galaxies increases by 2–3 orders of magnitude from z ∼ 0.5 to z ∼ 2, for an integrated H i flux cut at about 1.0 mJy km s−1. The large instantaneous bandwidth of modern receivers in current and next-generation telescopes, such as uGMRT, VLA, MeerKAT, and SKA1-MID; will detect a large number (∼104) of lensed H i galaxies and significantly improve our understandings of the cold neutral gas reservoirs, the evolution of the HIMF and the star-to-gas mass ratio at high redshifts.
ACKNOWLEDGEMENTS
We thank the staff of GMRT for making this observation possible. GMRT is run by National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research. We thank Yiping Shu for providing help and suggestions. We thank the editor and the reviewers for their helpful comments and suggestions, which improved the manuscript.
DATA AVAILABILITY
The data are publicly available in GMRT archive, with proposal code - 34_066.