MIGHTEE-HI: MeerKAT Neutral Hydrogen Survey

• MIGHTEE-HI is a deep, blind, medium-wide survey that uses the MeerKAT telescope to detect neutral hydrogen emission up to z≈0.6.
• The survey integrates direct detections, stacking analyses, and intensity mapping to reveal detailed H I scaling relations and kinematics across diverse environments.
• Its multifaceted methodology bridges local resolved H I studies and SKA-era research by uniting cataloguing efforts with advanced algorithmic and statistical approaches.

MIGHTEE-HI is the neutral-hydrogen emission component of the MeerKAT International GHz Tiered Extragalactic Exploration survey, carried out with the MeerKAT radio telescope. It is one of the first deep, blind, medium-wide interferometric surveys for neutral hydrogen ever undertaken, extending H I emission studies to $z \simeq 0.6$, with a 20 deg$^2$ main survey area centred on deep legacy fields and an effective H I survey area of 32 deg$^2$ at $z \approx 0$ when the background of the MeerKAT Fornax Survey is included (Maddox et al., 2020). In its published form, MIGHTEE-HI encompasses blind source catalogues, resolved H I kinematics, stacking analyses, galaxy-scaling relations, environmental studies, and H I intensity-mapping constraints, with imaging products for the COSMOS field released at full spectral resolution (Heywood et al., 2024).

1. Survey architecture and scientific remit

MIGHTEE-HI occupies the medium-deep, medium-wide niche within the MeerKAT H I survey landscape. Its core fields are COSMOS, XMM-LSS, ECDFS, and ELAIS-S1, all selected for exceptional ancillary data from optical through far-infrared wavelengths and for extensive spectroscopic coverage (Maddox et al., 2020). Within MeerKAT L-band, the survey covers the 21-cm line to $z \simeq 0.58$, while the same spectral products also encompass OH megamasers to $z \simeq 0.9$ (Heywood et al., 2024).

The survey design targets several linked problems in galaxy evolution. These include the evolution of the neutral gas content of galaxies over the past 5 billion years, the H I mass function and cosmic H I density, the dependence of H I content and morphology on environment, the low-mass galaxy population, and the acquisition of homogeneous resolved H I kinematics across a cosmologically useful volume (Maddox et al., 2020). Simulations associated with the survey design predicted nearly 3000 galaxies over $0$z=0.6$ (Maddox et al., 2020).

This remit has given MIGHTEE-HI a dual status. It is simultaneously a survey of galaxies in atomic gas and a platform for method development, because its science products range from direct detections of nearby dwarfs to stacking and intensity-mapping measurements at intermediate redshift. This suggests that MIGHTEE-HI was structured not only to populate catalogues, but also to connect traditionally separate H I methodologies within a single observing programme.

2. Observations, imaging products, and catalogues

The first full-resolution COSMOS release used a close-packed mosaic of 15 individual pointings with a total of 94.2 h on-target, covering two relatively interference-free spectral windows at 960–1150 and 1290–1520 MHz (Heywood et al., 2024). The highest-resolution imaging reaches 26 kHz spectral resolution, corresponding to 5.5 km s$^{-1}$ at $z=0$, with a median noise of 74 $^2$0Jy beam$^2$1 and a 12″ angular resolution for H I at $^2$2 (Heywood et al., 2024). In these data, the quoted 5$^2$3 H I mass limit is $^2$4 $^2$5 for a 300 km s$^2$6 line at $^2$7, and the mosaics cover more than 4 deg$^2$8 at multiple angular-resolution and sensitivity pairings (Heywood et al., 2024).

Earlier MIGHTEE-HI Early Science products used a coarser 209 kHz channel width, corresponding to about 44 km s$^2$9 at $^2$0, with median per-channel noise near 85 $^2$1Jy beam$^2$2 and beam sizes of order $^2$3–$^2$4 in COSMOS and XMM-LSS (Rajohnson et al., 2022). Those data underpinned the first survey-scale measurements of the H I size–mass relation, the baryonic Tully-Fisher relation, the spin–filament connection, and several environmental case studies (Rajohnson et al., 2022).

Mature catalogue products have now appeared. A COSMOS-field H I catalogue of 293 sources spans $^2$5 and includes H I masses, velocity widths, optical-through-near-infrared photometry, stellar masses, and star-formation rates (Maksymowicz-Maciata et al., 27 May 2026). That catalogue is coupled to a comparative study of source-finding strategies using simulated source injection, with tests of SoFiA, PyBDSF, ProFound, and LESHI, and with an analytic completeness function for the detected sample (Maksymowicz-Maciata et al., 27 May 2026). The combination of public imaging products and a characterised untargeted catalogue marks the transition of MIGHTEE-HI from Early Science to survey infrastructure.

3. Analysis framework

MIGHTEE-HI has been methodologically heterogeneous by design. Its foundational observable is the integrated 21-cm flux, converted to H I mass through the standard relation

$^2$6

with $^2$7 in $^2$8, $^2$9 in Mpc, and $z \approx 0$0 in Jy km s$z \approx 0$1 (Pan et al., 2022). Around this basic conversion, the survey deploys several distinct inference modes.

Blind source finding has evolved from visual cube inspection in Early Science work to comparative 3D algorithm studies in DR1-era cataloguing (Maksymowicz-Maciata et al., 27 May 2026). In parallel, targeted and statistical techniques are used when direct detection is inefficient. One major intermediate-redshift analysis stacked 9023 spectra of star-forming galaxies undetected in H I at $z \approx 0$2, extracted from MIGHTEE-HI Early Science datacubes, after binning by galaxy properties such as stellar mass, SFR, and sSFR (Sinigaglia et al., 2022). A complementary COSMOS analysis combined MIGHTEE-HI and CHILES spectra, rebinned onto a common velocity grid and stacked in stellar-mass bins to recover a robust $z \approx 0$3–$z \approx 0$4 relation at $z \approx 0$5 (Bianchetti et al., 31 Jan 2025).

For direct detections and resolved systems, 3D kinematic modelling is central. Rotation curves and H I surface-density profiles are extracted with 3DBarolo in recent dynamical work (Ponomareva et al., 17 Mar 2026), while TiRiFiC and FAT were used in the detailed tilted-ring modelling of NGC 895 (Namumba et al., 2023). Bayesian machinery is also prominent in unresolved scaling-relation work: MultiNest was used to fit the upper envelope of the $z \approx 0$6 relation directly in flux space, without binning and with explicit intrinsic scatter (Pan et al., 2022). At the largest scales, MIGHTEE-HI has been repurposed for H I intensity mapping, where delay-spectrum methods and incoherent averaging over a 15-pointing COSMOS mosaic yielded the first MIGHTEE upper limits on the H I power spectrum at $z \approx 0$7 (Mazumder et al., 29 Jan 2025).

4. Scaling relations and redshift evolution

A major result of MIGHTEE-HI is the recovery of classical H I scaling relations in a homogeneous interferometric dataset beyond the very local Universe. For 204 galaxies over $z \approx 0$8, the H I size–mass relation was measured as

$z \approx 0$9

with observed scatter 0.057 dex and intrinsic scatter $z \simeq 0.58$0 dex; no evidence for redshift evolution or morphology dependence was found over the last billion years (Rajohnson et al., 2022). In the same redshift interval, the H I-based baryonic Tully-Fisher relation for 67 galaxies yielded slopes $z \simeq 0.58$1 for $z \simeq 0.58$2 and $z \simeq 0.58$3 for $z \simeq 0.58$4, with intrinsic orthogonal scatter $z \simeq 0.58$5, again with no evidence for evolution over the last billion years (Ponomareva et al., 2021).

The H I mass–stellar mass relation has also been treated both as an observed upper envelope and as an inferred underlying population relation. Using 249 H I-selected galaxies at $z \simeq 0.58$6, a Bayesian comparison between a single power law and a double power law found decisive evidence for a turnover, with a preferred transition stellar mass of $z \simeq 0.58$7 and intrinsic scatter $z \simeq 0.58$8 dex for the upper envelope (Pan et al., 2022). Mock-catalogue analyses indicated that the turnover persists when detections and non-detections are modelled together, and that a purely linear underlying $z \simeq 0.58$9 relation is strongly disfavoured (Pan et al., 2022).

At higher redshift, MIGHTEE-HI has moved from resolved local calibrations to explicit evolutionary constraints. The stacking analysis of 9023 star-forming galaxies at $z \simeq 0.9$0 found moderate evolution from $z \simeq 0.9$1 to $z \simeq 0.9$2: low-$z \simeq 0.9$3 galaxies at $z \simeq 0.9$4 showed strong H I depletion of about 0.5 dex in $z \simeq 0.9$5, while massive galaxies at $z \simeq 0.9$6 kept their H I mass nearly unchanged (Sinigaglia et al., 2022). At fixed SFR or sSFR, highly star-forming galaxies evolved significantly in $z \simeq 0.9$7 and $z \simeq 0.9$8, whereas the lowest-SFR systems showed no evolution (Sinigaglia et al., 2022). A joint CHILES+MIGHTEE-HI stacking analysis at $z \simeq 0.9$9 obtained

$0

and concluded that galaxies at $0Bianchetti et al., 31 Jan 2025).

Direct intermediate-redshift detections now complement stacking. A targeted search in the COSMOS L1 cube detected 11 galaxies at $0Jarvis et al., 13 Jun 2025). These galaxies are markedly H I massive for their stellar mass, and the first H I-based baryonic Tully-Fisher relation at $0Jarvis et al., 13 Jun 2025). Taken together, these results suggest that MIGHTEE-HI is resolving a mass-dependent H I evolution in which low-mass systems deplete atomic gas more strongly than massive systems across $0

5. Resolved kinematics, mass models, and the baryon–dark matter connection

MIGHTEE-HI has extended resolved H I mass modelling beyond the very local Universe. In a sample of 20 galaxies at $0$z=0.6$0m imaging, and, for the first time in an H I-selected sample, spatially resolved stellar mass-to-light ratios from multi-wavelength SED fitting (Ponomareva et al., 17 Mar 2026). The analysis showed that adopting a fixed $z=0.6$1 yields a clear dependence of $z=0.6$2 on galaxy luminosity, while $z=0.6$3 largely suppresses that trend; resolved $z=0.6$4 preserves the luminosity dependence while modifying the stellar contribution on a galaxy-by-galaxy basis (Ponomareva et al., 17 Mar 2026).

Dark-matter haloes in that study were modelled with Navarro-Frenk-White profiles in a Bayesian MCMC framework, with halo masses compared to theoretical stellar-to-halo and baryonic-to-halo mass relations (Ponomareva et al., 17 Mar 2026). Fixed $z=0.6$5 assumptions systematically shifted galaxies relative to those $z=0.6$6CDM benchmarks, whereas spatially varying $z=0.6$7 yielded the closest agreement. The median offset from the abundance-matching stellar-to-halo mass relation was +0.39 dex for fixed $z=0.6$8, $z=0.6$9 dex for fixed $^{-1}$0, and +0.08 dex for resolved $^{-1}$1; for the baryonic-to-halo mass relation the corresponding offsets were +0.09, $^{-1}$2, and +0.04 dex (Ponomareva et al., 17 Mar 2026). The explicit conclusion was that accurate dark-matter inference from rotation curves requires $^{-1}$3 variations both across galaxies and within galaxies.

A related DR1 analysis used 19 H I-selected galaxies with resolved rotation curves and resolved stellar masses from 10-band SED fitting to measure the radial acceleration relation (Varasteanu et al., 29 Apr 2025). It found a tight relation with a low-acceleration power-law slope of about 0.5 and an intrinsic scatter of only $^{-1}$4 dex when a spatially varying mass-to-light ratio was adopted (Varasteanu et al., 29 Apr 2025). The fitted MOND-inspired acceleration scale for the MIGHTEE-HI sample alone was $^{-1}$5, and the study reported the first tentative evidence for redshift evolution in the acceleration scale, while stressing that more data are required for confirmation (Varasteanu et al., 29 Apr 2025). Within MIGHTEE-HI, these dynamical papers transform the survey from a gas census into a testbed for baryon–halo coupling.

6. Environment, interactions, and the cosmic web

Environmental science has been a prominent branch of MIGHTEE-HI from the outset. A serendipitously discovered galaxy group in the XMM-LSS field at $^{-1}$6 contained 20 H I-detected galaxies, had a dynamical mass $^{-1}$7, and a group-integrated H I-to-stellar mass ratio $^{-1}$8, which is 0.7 dex greater than expected (Ranchod et al., 2021). Its disturbed H I morphologies, non-Gaussian velocity distribution, and predominance of star-forming disk galaxies were interpreted as evidence for a gas-rich group in the early stages of assembly (Ranchod et al., 2021).

A more detailed nearby case study revised the environmental status of the spiral galaxy NGC 895. Deep MIGHTEE-HI observations showed extended spiral arms and two newly discovered H I companions, overturning the earlier view that NGC 895 was isolated (Namumba et al., 2023). The galaxy has $^{-1}$9, an H I diameter of 115 kpc at $z=0$0, stellar mass $z=0$1, and $z=0$2, placing it on the local star-forming main sequence despite warped and asymmetric H I morphology (Namumba et al., 2023). The absence of a clear stellar bridge and the presence of outer light excess in one companion indicate that H I can reveal environmental disturbance even when optical tidal debris is faint or absent (Namumba et al., 2023).

MIGHTEE-HI has also been used to test how gas-rich galaxies orient relative to the cosmic web. In a sample of 77 H I-selected galaxies with resolved 3D spin vectors, the alignment between the spin axis and the closest filament was stronger for galaxies within 5 Mpc of a filament, with $z=0$3, than for galaxies at $z=0$4 Mpc, for which $z=0$5 (Tudorache et al., 2022). The same study found that low H I-to-stellar mass ratio systems, defined by $z=0$6, are more aligned with filaments than high-gas-fraction systems, and reported no evidence for the predicted spin transition at $z=0$7 (Tudorache et al., 2022). A plausible implication is that recent gas-rich accretion or merger activity can reorient the H I spin axis away from the ambient filament direction.

7. Legacy, data products, and survey horizon

The mature form of MIGHTEE-HI is increasingly defined by the combination of public data products and cross-method analyses. The COSMOS spectral-line release provides calibrated imaging over both low- and high-redshift H I windows, and a public data release accompanies the imaging paper (Heywood et al., 2024). The untargeted 293-source COSMOS catalogue adds well-characterised completeness and detection-bias information derived from injection tests of multiple source finders (Maksymowicz-Maciata et al., 27 May 2026). At the same time, the survey has already demonstrated that targeted searches can directly recover 11 interferometric H I detections at $z=0$8, while stacking can recover population-mean H I scaling relations at $z=0$9 and $^2$00 (Jarvis et al., 13 Jun 2025).

MIGHTEE-HI has also entered the intensity-mapping regime. Using the COSMOS field over $^2$01, the first MIGHTEE H I power-spectrum analysis found no detection but obtained a best 1$^2$02 upper limit of 28.6 mK$^2$03 Mpc$^2$04 at $^2$05, consistent with the power spectrum detected in the MeerKAT DEEP2 field (Mazumder et al., 29 Jan 2025). That work concluded that combined analysis of the full MIGHTEE survey can potentially detect the H I power spectrum at $^2$06 over $^2$07 (Mazumder et al., 29 Jan 2025).

The survey’s long-term significance lies in the way its individual outputs cohere. Design studies projected nearly 3000 direct H I detections over $^2$08, with stacking and related statistical methods extending to $^2$09 (Maddox et al., 2020). Published work has already filled much of the intended methodological space: direct detections, scaling-relation evolution, resolved kinematics, dark-matter inference, cosmic-web alignment, group assembly, and intensity mapping. This suggests that MIGHTEE-HI functions not merely as a catalogue-generating survey, but as a bridge between local resolved H I astronomy and the SKA-era statistical study of atomic gas across cosmic time.