Gravitational lensing of 21 cm HI signal: detection prospects at z ~ 1 with uGMRT in galaxy cluster lenses

TL;DR

This study assesses the detectability of strongly lensed 21 cm HI emission from background galaxies behind galaxy cluster lenses using the uGMRT. By simulating a population of HI sources drawn from two HI mass functions and ray-tracing through 50 cluster lens models, it characterizes how lensing magnification and distortions affect HI line profiles and SNR. The results indicate that blind HI detections at $z\lesssim1.58$ require substantial observing time (up to ~900 hours per cluster for 5σ with uGMRT), while focusing on optically lensed systems (e.g., the Dragon Arc in Abell 370 or the HU in Abell 1703) can yield robust detections within tens of hours. The work also shows that HI magnification can occasionally exceed optical magnification near caustics, and it discusses strategies (e.g., matched filtering) to optimize signal extraction in forthcoming deep HI surveys.

Abstract

The atomic hydrogen HI content of galaxies is intimately related to star formation and galaxy evolution through the baryon cycle, which involves processes such as accretion, feedback, outflows, and gas recycling. While probing the HI gas over cosmic time has improved our understanding, direct HI detection with the redshifted 21 cm line is essentially limited to $z\lesssim 0.42$. Detections beyond this redshift are based on stacking to obtain average HI mass of galaxy populations. Gravitational lensing by the cluster lenses enhances the HI signal and can extend the redshift limit further. In this work, we describe simulations of HI lensing in cluster lenses. We explore the feasibility of detecting strongly lensed HI emission from background galaxies using known $50$ cluster lenses within the uGMRT sky coverage. We demonstrate that certain clusters offer a strong likelihood of HI detection. We also investigate how strong lensing distorts the HI spectral lines. The shape of the HI signal in these lensing models provides useful information and can be used in optimising signal extraction in blind and targeted HI surveys. We find that blind detection of HI signal from galaxies in the redshift range up to $1.58$ requires more than a few hundred hours of observations of individual clusters with the uGMRT. Detecting HI emission in galaxies with strong optical lensing seems promising, with a $5σ$ detection potential in less than 50 hours for Abell 370 and 75 hours for Abell 1703 using the uGMRT.

Figures (9)

• Figure 1: SNR dependence on the source inclination (without lensing). The blue and red contours are constant SNR curves for face-on and edge-on galaxies, respectively. We can see that the face-on Hi galaxies require less magnification compared to the edge-on galaxies, as the narrow linewidth has a relatively high peak flux density for the face-on galaxies.
• Figure 2: An example of lensed Hi source in Abell 370 and its SNR estimation at uGMRT resolution. Top-left panel shows the source plane with black and green curves showing the tangential and radial caustics, respectively. The stellar density profile is shown by a black disk shape, and the orange region around it shows the Hi surface density extent. The inset plot shows the Hi velocity profile with red/blue marking red/blue-shifted regions. Top-right panel shows the image plane with black and green curves representing the tangential and radial critical curves, respectively. Middle-left panel shows the Hi velocity distribution in the lensed Hi images. Middle-right panel shows the Hi line profile assuming the sky resolution is the same as the lens map resolution. The black dashed curve represents the Hi line profile of the unlensed source. The green, blue, and red curves show the lensed Hi line profile corresponding to the three lensed images enclosed in the same color contour in the top-right panel. Bottom-left panel shows the image plane map of integrated SNR as seen by the uGMRT. Note that the uGMRT pixel size is much larger ($\sim3"-5"$) compared to the lensing map pixel size ($0.05"$). Bottom-right panel shows the lensed Hi line profiles as seen by uGMRT on the coarser grid shown in the bottom-left panel.
• Figure 3: Optical and Hi magnification comparison of sources from $50 \times 200$ simulations (i.e., 200 realisations for each of the 50 clusters). The log–log scatter plot compares the optical and Hi magnifications for simulated sources. Each point is color-coded by its SNR, with higher-SNR points plotted on top to show the expected number and distribution of high-SNR sources. The corresponding conditional probability density distribution, $P(\mu_x|\mu_y)$, of optical and Hi magnification is shown in the histogram above. On the right side, we present four cut-outs to highlight the source plane for cases where the Hi magnification is larger than the corresponding optical magnification (see Sec. \ref{['ssec:opt_vs_HI']} for details).
• Figure 4: Hi mass vs. redshift distribution for simulated sources across 200 realisations per cluster (total simulations runs $=200\times50$). The top and bottom rows correspond to ALFALFA and combined HiMFs, respectively. In the left column, each point represents one lensed source and is colour-coded according to maximum pixel SNR. Each point is colour-coded by its SNR, with higher-SNR points plotted on top to show the expected number and distribution of high-SNR sources for a given redshift and mass. In the right column, the colour represents the total SNR calculated by co-adding all pixels with SNR $>1$.
• Figure 5: Average cumulative number of (un)lensed sources as a function of SNR. The left and right panels correspond to the ALFALFA and combined HiMFs, respectively. For each cluster, we simulate 200 realizations, and the thin blue and orange curves represent the number of lensed and unlensed sources per cluster in each realization, respectively. This gives us 200 thin curves. The thick blue and orange curves represent the average of these 200 thin curves for lensed and unlensed cases, respectively. The dashed vertical line mark the $\rm SNR>5.0$ threshold.