Constraining the Evolution of the H i Spin Temperature with Fast Radio Bursts H. Roxburgh, M. Glowacki, A. Bera, C. W. James Abstract Fast radio bursts (FRBs) emit broad band radio wave radiation that may, in rare cases, encode atomic hydrogen (H i ) absorption signals produced as they traverse the interstellar medium of their host galaxies. Combining such signals with high resolution H i emission maps offers a unique opportunity to probe the dynamics of neutral gas at cosmological distances through constraints of the H i excitation temperature T spin T_{\text{spin}} , which characterises the balance of neutral gas phases and the underlying thermal processes within these galactic environments. While no absorption signal has been recorded in an FRB to date, we demonstrate a proof of concept with the bright ( F = 35 F=35 Jy ms) and narrow (0.2 ms) FRB 20211127I detected by ASKAP. We find a 3 σ \sigma upper limit on the integrated optical depth in the pulse-averaged spectrum of 33 km s -1 , and, based on the H i emission observed in a 3 hr MeerKAT L L -band observation, subsequently find a lower limit on T spin T_{\text{spin}} of 26 K. While this test case provides little constraining power, we find that narrow, non-repeating FRBs with fluences greater than 20/70/150 Jy ms observed with all dishes with the current MeerKAT/ASKAP/DSA telescopes can probe integrated optical depths below 5 km s -1 . Furthermore, we highlight that utilising FAST’s incredible sensitivity to stack thousands of bursts from hyperactive repeaters also provides a plausible avenue through which H i absorption, and hence T spin T_{\text{spin}} , can be measured. Finally, we discuss how H i absorption can address several modern challenges in FRB science, providing a physical anchor for locating bursts within their host galaxies and helping to disentangle the host contribution to dispersion and scattering. H. Roxburgh]hugh.roxburgh@postgrad.curtin.edu.au \alsoaffiliation Institute for Astronomy, University of Edinburgh, Royal Observatory, Edinburgh, EH9 3HJ, United Kingdom \alsoaffiliation ASTRON, the Netherlands Institute for Radio Astronomy, Oude Hoogeveensedijk 4,7991 PD Dwingeloo, The Netherlands 1 Introduction Since its peak around redshift z\sim 2 , the observed star formation rate density (SFRD) in the Universe has declined roughly tenfold ( Hopkins & Beacom 2006 ; Madau & Dickinson 2014 ; Driver et al. 2018 ) . This evolution of the cosmic star formation history (SFH) is largely attributed to the interplay between the gravitational infalling of matter onto galaxies and the expulsion of matter from galaxies due to various feedback processes (e.g. Cox & Smith 1974 ; McKee & Ostriker 1977 ; Gatto et al. 2015 ) , though the driving factors surrounding this intricate cycle remain unclear. It is well established that the baryons lying in the interstellar medium (ISM) exist in various phases of matter ( Tielens 2005 ) . The bulk of this material exists as atomic hydrogen gas (H i ) with temperatures ranging from a few hundred to a few thousand Kelvin. Through various radiative cooling processes, this gas can contract into clouds of molecular hydrogen (H 2 ) which are dense enough to gravitationally collapse into stars ( Krumholz et al. 2009 ) . As such, a galaxy’s H i gas reservoir constitutes its star formation fuel, and thus its properties and dynamics are intrinsically tied to the rate of star formation. While observations of the 21-cm emission line – which arises from the spin flip transition of atomic hydrogen – broadly reveal the bulk distributions and kinematics of H i gas in galaxies, observations of our own Milky Way reveal that the reservoir can be further separated into distinct, coexisting phases of H i gas ( McKee & Ostriker 1977 ; Heiles & Troland 2003 ; Wolfire et al. 2003 ) . The cooler, denser material forms what is called the cold neutral medium (CNM) at temperatures around 100 K, whilst hotter material forms a diffuse warm neutral medium (WNM) around 10,000 K. These are stable phases of matter, though an unstable bridging phase also exists around 500 K, produced through shocks in the ISM ( Murray et al. 2018 ) . The interaction between these phases of matter plays an integral role in galaxy evolution ( McKee & Ostriker 1977 ) , and thus understanding how it varies over cosmic time and how it relates to particular galactic feedback processes – such as those from supernovae (SNe) and active galactic nuclei (AGN) winds – is a critical field of inquiry. Probing the nature of the physical conditions that differentiate these gas phases in galaxies is challenging. Studies focused on the absorption of background emission by foreground atomic hydrogen is considered one of the most effective methods through which to infer deeper information on the phase dynamics ( Allison 2021 ) . In particular, the equivalent width of the H i absorption line is directly connected to the population ratio between atoms in the two hyperfine spin states, through the effective excitation or ”spin” temperature, T_{\text{spin}} . This variable quantifies the population ratio, and in the cold dense gas of the CNM, it is considered a solid proxy for the gas’ bulk kinetic temperature, T_{\text{k}} , as atomic collisions are common enough for the weak spontaneous H i transition to occur ( Roy et al. 2013 ) . Measuring T_{\text{spin}} , however, presents its own challenges. Given T_{\text{spin}} quantifies a population ratio, one must know the total column density along the line of sight, N_{\text{HI}} , to explicitly measure it. This limits true constraints on T_{\text{spin}} to be made from only the rare cases where the bulk H i gas density along the line of sight is known, either through Ly- \alpha absorption of the background source (e.g. Curran et al. 2010 ; Kanekar et al. 2014 ) , or, in a few cases, through direct detection of H i emission from the foreground galaxy (e.g. Reeves et al. 2016 ; Gupta et al. 2018 ) . Both approaches depend on the presence of a background-quasar/foreground-galaxy pair, which occurs only rarely. However, the discovery of a bright extragalactic population of fast radio bursts (FRBs) over the past decade ( Lorimer et al. 2007 ; Thornton et al. 2013 ) may provide a far more powerful means of directly measuring extragalactic T_{\text{spin}} . While the nature of their progenitors remains elusive, FRB signals can theoretically encode H i absorption from the ISM of their host galaxies ( Fender & Oosterloo 2015 ; Margalit & Loeb 2016 ) ; combining this with highly resolved H i emission maps thus offers a direct measurement of T_{\text{spin}} in their hosts. Given their detection beyond z=2 ( Caleb et al. 2025 ) , and with hundreds of events occurring across the sky each day ( CHIME/FRB Collaboration et al. 2021 ) , FRBs could enable a substantially expanded, redshift-dependent census of T_{\text{spin}} over cosmic history. In this paper, we explore a proof of concept for constraining the evolution of the H i spin temperature over cosmic time using FRBs. In Section 2 , we cover the theory through which T_{\text{spin}} is measured from H i emission and absorption signals, and introduce our methodology for doing so with FRBs. In Section 3 , we explore a proof of concept with a real FRB whose host has been observed in H i emission. In Section 4 , we discuss the outlook for H i absorption and constraining T_{\text{spin}} using FRBs moving forward, and conclude in Section 5 . 2 Theory 2.1 Deriving T_{\text{spin}} In an H i gas reservoir, the relative population of atoms in the lowest singlet hyperfine state and the higher triplet hyperfine state ( F=0,1 respectively, where F is the total angular momentum quantum number) is determined using the Boltzmann distribution, and is expressed \frac{n_{2}}{n_{1}}=\frac{g_{2}}{g_{1}}\exp\Biggl(\frac{-\Delta E}{kT_{\text{spin}}}\Biggr)\ , (1) where g is the state degeneracy, \Delta E is the energy difference between the two states, and k is Boltzmann’s constant. This effective temperature directly impacts how opaque a medium is to the 21-cm signal; if most atoms are in the singlet state, the optical depth is much higher as absorption is more likely to occur. The optical depth of a medium can be measured from the absorption spectra of a source. If the continuum emission flux density is denoted S_{c} , then the frequency dependent optical depth of an absorption feature can be found using \tau(\nu)=-\ln\Biggl(1-\frac{\Delta S(\nu)}{f_{c}~S_{c}}\Biggr)\ , (2) where \Delta S is the difference between S_{c} and the minimum of the absorption line, and f_{c} is the covering fraction — the fraction of the background source covered by the foreground absorber on the sky. As FRBs are extremely physically compact sources ( r\sim 10 km, Farah et al. 2018 ) , f_{c} is always equal to unity, so we drop it henceforth. For optically thin emission ( \tau<<1 ), this becomes \tau(\nu)\simeq\frac{\Delta S(\nu)}{S_{c}}\ . (3) One can then calculate the total column density of H i atoms along the line of sight using (e.g. Wilson et al. 2013 ) : N_{\text{HI}}=1.823\times 10^{18}~T_{\text{spin}}\int\tau(\nu)~d\nu\ \ \text{cm}^{-2}. (4) This expression can thus easily be rearranged to calculate T_{\text{spin}} given a measurement of N HI from H i emission, and \tau(\nu) from H i absorption. 2.2 H i Absorption in FRBs The morphologies of FRB signals are highly varied (e.g. Pleunis et al. 2021 ) , though a good fraction of the population emit broad-band radiation. Many signals overlap with the frequency range of interest for studies of the H i transition, which has its rest frequency, \nu_{\text{HI}} , at \sim 1420 MHz. As such, it is plausible that FRBs may encode an absorption signal associated with the redshifted frequency of the H i transition in its host galaxy. Both Fender & Oosterloo 2015 ; Margalit & Loeb 2016 explored this concept in the early days of the field, when just a handful of FRBs had been detected. Their focuses were on using the redshift of a potential absorption line as a distance estimator, as no FRB had been localised to its host galaxy yet. Furthermore, the establishment of the repeating subclass of FRBs was very new ( Spitler et al. 2016 ) , and neither paper explored in detail the very interesting possibilities such objects offer through stacking methods, which we discuss later. It is therefore worth revisiting the theory of detecting absorption in the signals of FRBs. An FRB’s sensitivity to absorption detection is solely dependent on the signal-to-noise ratio (SNR) of its spectrum averaged over its temporal pulse width w_{\text{FRB}} at a spectral resolution of \Delta V in velocity space. Defining \sigma_{\nu} as the 1 \sigma uncertainty on the FRB flux density at each frequency channel, and \sigma_{\tau} as the corresponding 1 \sigma uncertainty on the derived optical depth, Eq. 3 can then be written as \sigma_{\tau}=\frac{\sigma_{\nu}}{S_{\text{FRB,HI}}}=\frac{1}{\text{SNR}_{\text{FRB,HI}}}\ . (5) Here, the subscript emphasises that this is the spectral SNR measured at the location of the H i line. This is not the same SNR as is reported in FRB discovery papers, which relates to the temporal pulse profile instead. Therefore, while we stress that SNR is the sole deciding factor in an FRB’s sensitivity to absorption, if we wish to estimate the sensitivities probed by the current population of detected FRBs, we must expand this expression in terms of their reported properties and those of the theoretical absorption features. For a resolved absorption line with a width W spanning N channels of velocity width \Delta V , the integrated optical depth is given by \int\tau=\sum_{i}^{N}\tau_{i}\ \Delta V\ \ \text{km s}^{-1}\ . (6) Assuming uncorrelated noise, the uncertainty adds in quadrature: \sigma_{\int{\tau}}^{2}=\text{Var}\Biggl(\sum_{i}^{N}\tau_{i}\ \Delta V\Biggr)=\sum_{i}^{N}\Delta V^{2}\ \text{Var}(\tau_{i}) =N\ \Delta V^{2}\ \sigma_{\tau}^{2}=W\Delta V\ \sigma_{\tau}^{2}\ . (7) The noise in the pulse-averaged spectrum can be derived from the radiometer equation: \sigma_{\nu}=\frac{\text{SEFD}}{\sqrt{N_{\text{pol}}\ w_{\text{FRB}}\ \Delta\nu}}\ \ \text{Jy} =\text{SEFD}\sqrt{\frac{c}{\nu_{\text{HI}}}}{\sqrt{\frac{1+z}{N_{\text{pol}}\ w_{\text{FRB}}\ \Delta V}}}\ \ \text{Jy}\ , (8) after transforming into the FRB’s rest frame velocity space using its redshift z . Using this expression and Eq. 7 , we arrive at the sensitivity to integrated optical depth in terms of a telescope’s SEFD, the line width, and an FRB’s properties: \sigma_{\int{\tau}}=\sigma_{\tau}\sqrt{W\Delta V}=\frac{\sigma_{\nu}\sqrt{W\Delta V}}{S_{\text{FRB,HI}}} =\frac{\text{SEFD}}{S_{\text{FRB,HI}}}\sqrt{\frac{c}{\nu_{\text{HI}}}}\sqrt{\frac{W~(1+z)}{N_{\text{pol}}\ w_{\text{FRB}}}}\ \ \text{km s}^{-1}\ . (9) The final simplification we can incorporate deals with the spectral behaviour of FRBs, which can be complex; some bursts occupy only part of the observing bandwidth ( Pleunis et al. 2021 ) , and others encode interstellar scintillation such that even bright bursts may show little flux at the H i line ( Masui et al. 2015 ) . However, while there is some evidence for a preference toward lower-frequency emission ( James et al. 2022 ; Shin et al. 2023 ; Shannon et al. 2024 ) , no strong systematic trend in flux density is observed across the narrow bandwidth relevant here. For simplicity, and following several previous studies ( CHIME/FRB Collaboration et al. 2021 ) , we therefore can assume an idealised FRB with a flat spectrum, such that S_{\text{FRB,HI}}(\nu)=\bar{S}_{\text{FRB,HI}}(\nu) . This can be expressed in terms of the FRB fluence, F_{\text{FRB}} , as \bar{S}_{\text{FRB,HI}}(\nu)=\frac{F_{\text{FRB}}}{w_{\text{FRB}}}\ , which then allows Equation 9 to be helpfully rewritten as a fluence-dependent metric. Using this form, a 3 \sigma limit is equal to L_{3\sigma}=\frac{3\times\text{SEFD}}{F_{\text{FRB}}}\sqrt{\frac{c}{\nu_{\text{HI}}}}\sqrt{\frac{W~w_{\text{FRB}}\ (1+z)}{N_{\text{pol}}}}\ \ \text{km s}^{-1}\ . (10) 2.3 Measuring T_{\text{spin}} & Beam Size It is important to briefly consider the scales being probed through both H i emission and absorption in this context, and what that means for the interpretation of the measured T_{\text{spin}} . Most direct T_{\text{spin}} measurements are made by observing the H i and Lyman- \alpha absorption of a foreground galaxy in the spectrum of a background QSO, though there are a few cases where the column density measurement has been derived from H i emission maps as we attempt here (e.g. Reeves et al. 2016 ; Gupta et al. 2018 ) . This method is largely required for probing low redshift ( z<1 ) gas; the rest wavelength of the Ly- \alpha line lies in the ultraviolet (UV), a regime which the Earth’s atmosphere is opaque to. These studies acknowledge that the scale over which the absorption occurs — i.e. the physical size of the absorbing region in the foreground galaxy — is much smaller than the angular resolution of the telescopes used to probe H i emission. However, they stress that the measured T_{\text{spin}} value is a column density weighted harmonic mean temperature, which describes the broader thermal state of the galaxy’s ISM in the region around the absorber, rather than any individual cloud intersected along the line of sight. Given FRBs are effectively point sources, they probe only the gas the pencil beam traverses, and thus any derived T_{\text{spin}} measurement is a harmonic mean of the surrounding medium. Particularly in face-on galaxies, where any line of sight through the disk intersects a relatively short path length and the emitting gas sampled by the beam arises from a similar physical region of the disk, the beam-averaged H i emission provides a reasonable proxy for the column density associated with the absorbing material. In this geometry, small-scale variations in cloud structure occur primarily side-by-side across the disk rather than being stacked along the line of sight. This prevents the signal from being distorted by multiple overlapping layers of gas, and therefore the large-scale emission map provides a much cleaner reflection of the specific environment the FRB passes through. 3 Analysis We now attempt to demonstrate a proof of concept with real FRBs and H i intensity maps of their host galaxies. Over 100 FRBs have been localised to their host galaxies, a good fraction at arcsecond scales ( Shannon et al. 2024 ; Sharma et al. 2024 ; CHIME/FRB Collaboration et al. 2025 ; Pastor-Marazuela et al. 2025 ) . Until recently, only a handful of these hosts had been studied individually for their H i content ( Michałowski 2021 ; Kaur et al. 2022 ; Glowacki et al. 2023 ; Lee-Waddell et al. 2023 ; Glowacki et al. 2024 ; Kaur et al. 2025 ) . However, Roxburgh et al. 2025 expanded the sample to 22 hosts, providing the first population-level analysis through new observations with MeerKAT ( Jonas & The MeerKAT Team 2016 ) , the Giant Metrewave Radio Telescope (GMRT; Swarup et al. 1991 ; Gupta et al. 2017 ) , and the Five-hundred-meter Aperture Spherical Telescope (FAST; Nan et al. 2011 ) . The majority of these detections were resolved, offering insights into the spatial distribution of gas in FRB host galaxies. Candidate FRBs for this proof of concept must satisfy five requirements: 1. They must have spectra captured at very high spectral resolution, at least around 10 km s -1 to unequivocally resolve any potential absorption line. 2. They must be localised to arcsecond or better precision to ensure the line of sight is well understood. 3. The redshift of their host galaxy must lie in the nearby Universe ( z<0.1 ) — this is due in part to the inherent weak intensity of the H i emission line, but mostly due to the dominant radio frequency interference (RFI) that occupies frequencies until z\sim 0.2 . 4. Their signals must have been detected in the frequency range of the redshifted H i transition. 5. The H i emission from their host galaxies should be bright and resolved in both spatial and spectral dimensions. A tentative sixth requirement may include the host galaxy exhibiting low inclination (i.e. face-on) to mitigate line-of-sight confusion and beam-averaging effects in the associated H i emission, thereby improving the interpretability of any derived T_{\text{spin}} measurement. However, higher inclination also increases the likelihood of absorption detection, so we do not consider this a strict requirement. Of the options presented in Roxburgh et al. 2025 , only one meets all these criteria: FRB 20211127I. This was detected by the Australian SKA Pathfinder (ASKAP; Hotan et al. 2021 ) telescope and analysed by Glowacki et al. 2023 ; Shannon et al. 2024 ; Scott et al. 2025 . In Figure 1 , we present its dedispersed dynamic spectrum, which has been processed offline through the CELEBI (CRAFT Effortless Localisation and Enhanced Burst Inspection; Scott et al. 2023 ) pipeline. Alongside this, we display VLT imaging of its host galaxy with its localisation and the host’s H i emission overlaid as measured by MeerKAT. Figure 1 : (Left) Dynamic spectrum of FRB 20211127I. The right panel displays the spectrum of the FRB averaged across its temporal pulse width. The location of the redshifted H i line is shown by the horizontal dashed line, and the band-averaged flux density is denoted by the vertical dashed line in the right panel. (Right) VLT i -band image of the host galaxy of FRB 20211127I overlaid with its H i emission as seen by MeerKAT (white contours) and the localisation region in green. Contours increase as [1,1.25,1.5,…] \times 5.34 \times 10^{20} cm -2 and the beam is shown in the bottom right. Unfortunately, as is evident in the “striped” appearance of the FRB’s signal in the dynamic spectrum, this FRB exhibits strong interstellar scintillation. Some FRBs show scintillation patterns with decorrelation bandwidths \nu_{\text{DC}} — i.e. the scale over which scintillation fringes oscillate — near or equal to the characteristic widths of absorption features. Scott et al. 2025 find a \nu_{\text{DC}} of 2.9 MHz for this FRB, which corresponds to around 650 km s -1 in the rest frame, and thus this is not a significant issue for this FRB. However, scintillation still makes discerning an absorption signal difficult; as shown by the horizontal line in the dynamic spectrum, the H i line lies near the boundary of a scintle and a trough. Regardless, we can extract a high resolution spectrum around this region and search for absorption signals. In Fig. 2 , we present the H i emission line spectrum arising from the pixel nearest the FRB’s localisation region. We also show the same frequency range in the FRB’s dedispersed signal, which has been flattened by fitting a local polynomial baseline, calculated while excluding all emission channels. Each spectrum is arbitrarily normalised here. We find that, as expected, no clear H i absorption signal is present in the spectrum of FRB 20211127I. Instead, we calculate an upper limit on the strength of any potential H i absorption, and use this to propagate through to a lower limit on T_{\text{spin}} . Figure 2 : (Top) Normalised H i emission measured by MeerKAT at the pixel closest to the localisation region. The raw data is overlaid with a Gaussian profile fit to the binned data. (Bottom) Flattened and normalised FRB spectrum at the location of the H i line. For visual purposes, the spectrum is overlaid with a binned spectrum at a velocity resolution of 11 km s -1 in the rest frame and the 2 \sigma deviation from unity is shown. Due to scintillation, the SNR of the spectrum increases towards the left. If we assume an H i line width W of 50 km s -1 — which is broadly a representative width of H i absorption features detected in spectra of extragalactic sources ( Prochaska et al. 2008 ) — we can rebin the absorption spectrum such that our individual channel width is equal to W and measure the resulting noise level. As the spectrum is normalised, this noise level is simply the inverse of the spectral SNR outlined earlier, which we find to be equal to 5.9. This SNR is too low to use the simplifying expression presented in equation 3; reincorporating the log term into equation 10 gives our 3 \sigma limit in the integrated optical depth of roughly 33 km s -1 . We can now place a lower limit on T_{\text{spin}} using equation 3. To most accurately compare emission and absorption, we need to calculate our column density over the same line width we assume for our absorption profile. Given the width of the emission line at the pixel nearest the FRB location is roughly 170 km s -1 , we can construct an equivalent boxcar emission spectrum containing equal total flux and then integrate over just 50 km s -1 . This results in an H i column density, N HI , of 1.43 \times 10^{21} cm -2 . Combining this with our upper limit for the integrated optical depth, we find a 3 \sigma lower limit for T_{\text{spin}} of roughly 26 K. 4 Discussion As expected for non-detections in absorption, the result for this test case is unremarkable. The derived T_{\text{spin}} values from direct emission-absorption systems are in the hundreds of K ( Reeves et al. 2016 ; Gupta et al. 2018 ) , where interesting conclusions can be drawn about the dynamics of the CNM and WNM. Milky Way absorption finds typical lower limits for CNM cloud temperatures around 80 K ( Murray et al. 2018 ) ; as such, our lower limit of 26 K provides little useful constraining power. However, there are a number of reasons to believe such a prescription will be viable moving forwards. Here, we comment on the specific pros and cons of this particular FRB, and the outlook of future FRB observation campaigns. 4.1 Strengths and Limitations of FRB 20211127I FRB 20211127I is in many ways an ideal burst for such an investigation. It is a bright and narrow burst, exhibiting little to no scattering, which allows for a high SNR in the pulse-averaged spectrum. Even the dominant interstellar scintillation pattern noted earlier does not present a major barrier in this case. Scintillation typically complicates absorption searches, and this FRB exhibits one of the strongest patterns in the CRAFT sample; Scott et al. 2025 report a spectral modulation index of m=0.74 (values near unity indicate strong scintillation). Even so, its impact on the effective sensitivity of this test case is minimal. As shown by the FRB spectrum in the left image of Fig. 1 , the flux density of the spectrum at the location of the H i line almost perfectly coincides with the band-averaged flux density of the FRB. As such, our 3 \sigma limit would not change significantly if the scintillation was weaker or even non-existant. In fact, scintillation of this manner could, in a different scenario, be considered an advantage rather than a disadvantage; for FRBs with \nu_{\text{DC}} values significantly larger than the typical width of absorption features, as is the case for FRB 20211127I, a coincidental alignment of a scintle with the H i line could provide a boost in SNR across the line that could significantly alter the likelihood of detection. To illustrate this, we consider the resulting limits on both integrated optical depth and T_{\text{spin}} were the H i line to coincide with the local spectral maxima near 1362 MHz, rather than at 1357 MHz. In this case, the SNR increases by roughly a factor of 2, yielding an integrated optical depth limit of approximately 14 km s -1 and a corresponding T_{\text{spin}} lower limit of 56 K. Such an integrated optical depth is still very much on the extreme boundary of plausible detection ( Braun 2012 ) , though the resulting T_{\text{spin}} limit would be within \sim 15 K of typical temperatures observed in the Milky Way’s CNM ( Murray et al. 2018 ) . 4.2 Detectability of HI Absorption in Non-repeating FRBs 4.2.1 Current Facility Capabilities As outlined in Section 2 , the detectability of H i absorption in FRB signals is governed by a combination of instrumental and source properties, including the system noise, the H i line width, and the FRB’s intensity, temporal duration, and redshift. To understand the broader likelihood of any given facility to detect absorption/constrain T_{\text{spin}} , we must also consider their fields of view (FoV) and frequency ranges, as well as the intrinsic luminosity distribution and spectral behaviour of the FRB population. Although absorption measurements are not limited by the distance–dependent sensitivity constraints that restrict H i emission studies to the local Universe, this method used to constrain T_{\text{spin}} relies on the availability of emission maps. For this reason, facilities which do not cover the 1.3–1.42 GHz band where local Universe H i is detected cannot meaningfully contribute to such analysis. We therefore exclude the Canadian Hydrogen Intensity Mapping Experiment (CHIME; Bandura et al. 2014 ) which observes between 400-800 MHz, but note that its successor, The Canadian Hydrogen Observatory and Radio-transient Detector (CHORD; Vanderlinde et al. 2019 ) , will operate with an ultra-wide band between 300-1500 MHz and thus is poised to become an important resource for future work. Henceforth, we consider only the current capabilities of the remaining major facilities in the field — ASKAP, the Deep Synoptic Array (DSA; Kocz et al. 2019 ) , FAST, and MeerKAT. Our aim is to estimate the likelihood that a non-repeating FRB detected by each of these facilities will be bright enough to probe realistic H i absorption. By calculating this probability, we can quantify the broader relative ability of these different observatories to conduct such studies. Eq. 10 provides us with the tools to make such estimates; all we require are the SEFDs of the telescopes. Since ASKAP, DSA, and MeerKAT each have the capability to process FRBs offline after detection, we can assume the full sensitivity achievable through coherent beamforming. At current maximum operating capacities (36, 63, and 64 antennas, respectively), their effective SEFDs are approximately 50 Jy, 100 Jy, and 12.5 Jy, while FAST achieves around 1.25 Jy at L -band. Once again setting the absorption line width W to 50 km s -1 , Figure 3 presents grids of 3 \sigma integrated optical depth limits as a function of burst width and fluence for each telescope. We overlay FRBs with reported widths and fluences to contextualise current samples — the recent sample expansions for DSA ( Sharma et al. 2024 ; Connor et al. 2025 ) and MeerKAT ( Pastor-Marazuela et al. 2025 ) do not currently include such information — though we note that many of these FRBs were observed with fewer than the maximum number of antennas; in these cases the true limits are higher than those shown here, meaning that these events appear artificially high relative to the displayed limits. The ASKAP FRBs, which are taken from Shannon et al. 2024 , are observed with smaller bandwidths than the other facilities, and thus we colour each by their central observing frequencies, noting that many lie outside our near-Universe band of interest. However, this figure indicates that integrated optical depths of a few km s -1 are certainly detectable, though only for very high-fluence bursts. One event stands out clearly in this figure: the MeerKAT FRB 20210405I. This burst was exceptionally bright and was localised to a nearby galaxy at z\sim 0.06 ( Driessen et al. 2023 ) . Follow-up observations by Roxburgh et al. 2025 revealed a face-on H i disk in the host galaxy, making this source an excellent candidate for probing H i absorption and constraining the host-galaxy T_{\text{spin}} . Unfortunately, the burst was detected prior to the implementation of MeerTRAP’s transient buffer system ( Rajwade et al. 2024 ) , which now enables the recording of full-voltage data for later processing. Consequently, only a coarse 1024-channel spectrum spanning \sim 800 MHz is available, providing insufficient spectral resolution for a meaningful absorption search. Nonetheless, this event demonstrates that such favourable FRBs do occur, and that similar sources will be accessible to future absorption studies with the voltage-capture capabilities now available on ASKAP, DSA, and MeerKAT. Figure 3 : 3 \sigma limits on the integrated H i optical depth detectable in the pulse-averaged spectra of non-repeating FRBs observed by various telescopes. The black dashed lines in each panel indicate a limit of 5 km s -1 . These limits assume maximal sensitivity (i.e. with full antenna configuration) and coherent beamforming, an H i linewidth of 50 km s -1 , and a flat FRB spectrum. Overlaid are samples of reported FRBs with known fluences and widths: ASKAP ( Shannon et al. 2024 ) , DSA ( Law et al. 2024 ) , FAST ( Zhu et al. 2020 ; Niu et al. 2021 ; Zhou et al. 2023 ) , and MeerKAT ( Rajwade et al. 2022 ; Jankowski et al. 2023 ; Driessen et al. 2023 ) . ASKAP is divided in colour based on the central frequency of the band used to observe the FRB. In all panels, points with red borders indicate FRBs with known localisations to galaxies with redshifts below 0.1. To calculate the fraction of detected FRBs capable of probing HI absorption, we must define an SNR threshold, \text{SNR}_{\text{thr}}(w) , required to surpass a target 3\sigma integrated optical depth limit, \tau_{\text{thr}} . We then can estimate the probability P(\text{SNR}\geq\text{SNR}_{\text{thr}}(w)\mid\text{SNR}\geq\text{SNR}_{\text{det}}) , where \text{SNR}_{\text{det}} is the detection threshold of the facility. This estimation relies on the SNR distributions observed by each telescope, which reflect the underlying FRB fluence distribution. If this distribution is scale-invariant above each telescope’s fluence threshold ( F_{\text{det}} ), all facilities observe SNR distributions of identical shape. Consequently, the fraction of absorption-sensitive bursts scales as (F_{\text{thr}}(w)/F_{\text{det}})^{\gamma+1} , where \gamma is the differential power-law index. Several studies have attempted to model the fluence distribution of the FRB population (e.g. James et al. 2019 ; Shin et al. 2023 ; Ryder et al. 2023 ; Arcus et al. 2025 ) , generally employing a Schechter function to determine the high end cutoff energy imposed by the expected physical limit to the FRB emission mechanism. This function is not scale invariant, but given the downturn must lie orders of magnitude above telescope thresholds ( Arcus et al. 2025 ) , we can consider the SNR distributions effectively equivalent across instruments. Possible deviations at low fluences — such as variations in the power-law slope reported for some repeaters ( Li et al. 2021 ; Kirsten et al. 2024 ) and for non-repeating FRBs detected by MeerKAT ( Jankowski et al. 2023 ) — remain inconclusive and are neglected here. Studies in the L -band generally find \gamma values near -2 ( James et al. 2022 ; Hoffmann et al. 2024 ; Arcus et al. 2025 ) , as expected for a cosmological population of sources. Thus, the fraction of interest reduces simply to F_{\text{det}}/F_{\text{thr}}(w) . As F_{\text{thr}} is a function of w , we must multiply this fraction by the probability distribution of pulse widths p(w) and then integrate over w . Ideally we would use each telescope’s observed width distribution; due to the current low number of reported widths in DSA, FAST, and MeerKAT, we fit a log normal function to the width distribution of ASKAP and assume this for all. Now only the fluence detection thresholds of each telescope and observing system are required ( Shannon et al. 2024 ; Wang et al. 2025 ; Connor et al. 2025 ; Niu et al. 2021 ; Jankowski et al. 2023 ) . As only very bright FRBs can probe low \tau_{\mathrm{thr}}~ , such events are detectable even with the incoherent summation modes used by ASKAP and MeerKAT; in these modes, F_{\text{det}} is a factor of \sqrt{N_{\text{ant}}} closer to F_{\text{thr}} , which doesn’t change because it is always defined with respect to the telescopes’ coherent sensitivity due to the fact that the array facilities can process offline after detection. Table 1 presents the resulting absorption-probing fractions of the four observatories in their different detection modes as percentages ( \tau~\% ), given a target \tau_{\mathrm{thr}} of 5 km s -1 . From this we can see that a sizeable fraction of FRBs detected with the incoherent modes of ASKAP and MeerKAT are expected to probe realistic H i absorption. Of course, there are the extra factors concerning the redshift distribution and observation band that further reduce the fraction of usable bursts; however, these numbers indicate that absorption detection in the spectra of FRBs is certainly not implausible. To compare the relative likelihood of detection between facilities, we must simply estimate the total number of FRBs with fluences greater than F_{\text{thr}} each facility will detect in an equal observing time. This figure of merit (FoM) is effectively the total number of FRBs a system detects multiplied by the absorption-probing fractions calculated above. The former is proportional simply to the FoV / F_{\text{det}} , and thus the FoM is given by \mathrm{FoM}=\int\frac{\mathrm{FoV}}{F_{\mathrm{thr}}(w)}\,p(w)\,\mathrm{d}w\ . (11) The resulting values are also found in Table 1 . This illustrates that ASKAP, observing in either incoherent or coherent mode, is by far the most likely telescope to probe H i absorption in the spectra of non-repeating FRBs. Table 1 : Comparison of facility-specific sensitivity to H i absorption. The fourth column indicates the percentage of detected FRBs with fluences great enough to probe an integrated optical depth threshold ( \tau_{\text{thr}} ) of 5 km s -1 . The Figure of Merit in the final column indicates the relative likelihood of detecting an FRB that probes the same \tau_{\text{thr}} in an equal observing time, calculated using Eq. 11 . \headrow Facility Detection Mode F_{\text{det}} (Jy ms) a \tau~\% FoV (deg 2 ) a FoM Incoherent 8.0 6.2 30 23.2 ASKAP Coherent 1.2 0.9 30 23.2 DSA \dagger Coherent 1.9 0.8 3.4 1.32 FAST Single dish 0.015 0.5 0.008 0.25 Incoherent 3.4 10.6 1.3 4.02 MeerKAT Coherent 0.7 2.0 0.4 1.23 a Values are taken from Shannon et al. 2024 ; Wang et al. 2025 ; Law et al. 2024 ; Connor et al. 2025 ; Niu et al. 2021 ; Jankowski et al. 2023 . \dagger DSA only observes in coherent detection mode. 4.2.2 Future Facility Capabilities Estimating the performance of future facilities like SKA-Mid or the full DSA is challenging, as it depends heavily on the specific configurations of their FRB detection backends. However, the limits presented in Figure 3 are independent of detection architecture; provided the coherent SEFD of a future facility can be estimated, similar sensitivity projections can be made. The predicted SEFDs for SKA-Mid 1 1 1 https://www.skao.int/en/resources/technical-documents and the 1650-dish DSA 2 2 2 https://www.deepsynoptic.org/ are approximately 2.5 Jy and 4 Jy, respectively. At these sensitivity levels, a 1 ms burst would require a fluence of 5 Jy ms (SKA-Mid) or 8 Jy ms (DSA) to probe a 50 km s -1 wide absorption line at a 3 \sigma integrated optical depth limit of 5 km s -1 . 4.3 Repeating FRBs The population of repeating FRBs poses an entirely different opportunity for probing H i absorption and T_{\text{spin}} . Dozens of these sources have now been detected ( CHIME/FRB Collaboration et al. 2023 ) , with some ”hyperactive” repeaters ( Konijn et al. 2024 ; Tian et al. 2025 ) exhibiting thousands of outbursts. As these bursts originate from the same environment and traverse identical structures within the host galaxy, stacking their signals increases the cumulative integration time for any imprinted absorption features, effectively boosting the sensitivity to the H i line. This clearly makes hyperactive repeaters — several of which have been localised to host galaxies with redshifts below 0.1 ( Day et al. 2021 ; Ravi et al. 2023 ) — the best testing grounds for H i absorption. With its incredible sensitivity, FAST’s true strengths are not in the serendipitous detection of apparently non-repeating FRBs, but in the regular monitoring of these repeating FRBs. If we assume that each burst has the same pulse width (for simplicity; width can vary by an order of magnitude, e.g. Li et al. 2021 ) , then the improvement in sensitivity to H i absorption scales with \sqrt{N_{\text{bursts}}} . As such, we can estimate how FAST’s fluence limit for detecting a given integrated optical depth changes as a function of the number of stacked FRBs, which we demonstrate in Fig. 4 . This shows that just 100 FRBs with a pulse width of 1 ms and an average fluence of 2 Jy ms can probe integrated optical depths of just 0.5 km s -1 . Of course, we reiterate that this estimate uses idealised, flat spectrum FRBs; however, the z=0.077 repeater FRB 20220912A has outbursted thousands of times with fluences around 1 Jy ms ( Zhang et al. 2023 ) , and several tens of times above 50 Jy ms ( Sheikh et al. 2024 ) , indicating that highly sensitive probes of absorption already exist. Furthermore, with CHIME’s recent second data release reporting dozens of new repeaters ( CHIME/FRB Collaboration et al. 2026 ) , the sample of hyperactive repeaters is bound to continue to expand, making absorption detection increasingly realistic. Figure 4 : Average burst fluence required for FAST to detect various integrated H i optical depths as a function of the number of bursts stacked from a repeating FRB. 4.4 FRB Distance Inside Host Galaxy Fender & Oosterloo 2015 ; Margalit & Loeb 2016 previously identified the use of potential H i absorption in FRBs as an independent distance estimator. The rapid improvements in localisation ability since their release have reduced the value for such a concept. In this paper, we have focused predominantly on using absorption as a tool to measure T_{\text{spin}} , but here we briefly comment on another use case that would be of high value to the FRB community. Currently a significant issue facing those who aim to use FRBs as cosmological probes is the unknown contribution of the host galaxy on the dispersion measure (DM) of FRBs ( James et al. 2022 ; Connor et al. 2025 ) . Disentangling this factor from the cosmological contribution, which is sensitive to several key cosmological parameters, is an ongoing thorn in the side of modellers. Part of this issue arises from the unknowable projected distance an FRB lies within its host; the dispersion of a burst’s signal may be significantly larger if it lies on the far side of a galaxy rather than the near side. If a resolved absorption feature is present in a nearby FRB, it may be possible to compare its location in the band with that of an H i emission line. If the absorption line is blueshifted relative to the emission line, this indicates that the FRB lies on the near side of the bulk of the host’s interstellar medium (ISM) or is associated with an intervening cloud. A redshifted feature relative to the emission line would provide strong evidence that the FRB originates on the far side of the galaxy and is seen through the receding half of the local gas column. Comparing this kinematic depth with the measured DM would provide stronger host galaxy constraints than is currently possible. This is particularly critical given the lack of correlation between excess DM and other FRB properties (Marnoch et al. in prep.; Scott et al. 2025 ; Mas-Ribas & James 2025 ) , as it would provide a physical rather than statistical basis for DM-host subtractions. Furthermore, measuring this kinematic depth would provide a direct method to quantify the impact of host galaxy scattering. By definitively placing the FRB relative to the host’s H i reservoir, one could determine whether observed pulse broadening is dominated by the dense, turbulent environment immediately surrounding the progenitor or the integrated column of the galactic disk. This distinction is vital for accurately modelling the scattering-DM relation, as it allows us to identify if the host’s interstellar medium acts as a significant scattering screen or if such effects are largely localised to the source region. 5 Conclusion We demonstrate a method of constraining the spin temperature of atomic hydrogen gas in galaxies through the comparison of potential H i absorption features in localised FRBs with high resolution H i emission maps. Using the signal of FRB 20211127I, and a 3 hr L -band observation with MeerKAT, we find a 3 \sigma upper limit in the integrated optical depth of 33 km s -1 , corresponding to a lower limit on T_{\text{spin}} of 26 K. While this non-detection unsurprisingly offers little constraining power, we discuss the possibility of detecting absorption in the signals of FRBs given the capabilities and strategies of modern observatories. We first calculate the fraction of FRBs detected by ASKAP, DSA, FAST, and MeerKAT capable of probing integrated optical depths below 5 km s -1 , finding that the incoherent detection modes of MeerKAT and ASKAP show the greatest proportions at 10 \% and 6 \% . By also factoring in FoV and post-detection processing capabilities, we find that ASKAP, observing at full 36-dish strength, currently offers the greatest chance of detecting H i absorption in a non-repeating FRB. This would still require a relatively narrow ( <3 ms) burst with a fluence greater than 100 Jy ms. However, we reiterate that such FRBs have been previously observed; the z\sim 0.06 MeerKAT FRB 20210405I would have probed deep absorption and T_{\text{spin}} limits had its voltages been saved to produce a higher resolution spectrum. We also consider the opportunity presented by the repeating FRB population, particularly the hyperactive sub-population which repeat thousands of times. Monitoring these objects offers a separate avenue that, through FAST’s current sensitivity, pushes absorption detection into plausibility through stacking. Finally, we highlight that combining H i absorption with high resolution H i emission maps may provide a way to disentangle the host galaxy contribution to an FRB’s dispersion measure from its cosmological contribution, aiding in providing constraining power to cosmological studies using FRBs.