Neutral gas phase distribution from HI morphology: phase separation with scattering spectra and variational autoencoders Journal: ApJ astropy ( Astropy Collaboration et al. 2013 ) , scattering ( Cheng & Ménard 2021 ) , NumPy ( van der Walt et al. 2011 ) , healpy ( Górski et al. 2005 ) , PyTorch ( Pa
Minjie Lei, S. E. Clark, Rudy Morel, E. Allys, Iryna S. Butsky, Caleb Redshaw, Drummond B. Fielding · 2025
Read Neutral gas phase distribution from HI morphology: phase separation with scattering spectra and variational autoencoders Journal: ApJ astropy ( Astropy Collaboration et al. 2013 ) , scattering ( Cheng & Ménard 2021 ) , NumPy ( van der Walt et al. 2011 ) , healpy ( Górski et al. 2005 ) , PyTorch ( Pa by Minjie Lei, S. E. Clark, Rudy Morel, E. Allys, Iryna S. Butsky, Caleb Redshaw, Drummond B. Fielding (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-20407.txt.
Introduction
Minjie Lei and co-authors present a data driven way to sort the phases of neutral hydrogen gas in our galaxy, in a paper from 2025. The gas exists in warm and cold phases mixed together, and its split shapes how stars form. The split is usually read from line shapes in radio spectra, which leaves it poorly pinned down.
The new method reads phase structure from the shapes seen in sky maps alone. It joins a scattering based summary of texture with a learning model that groups patterns. Applied to maps from a large radio survey of high latitude sky, the method pulls apart phases in narrow speed slices. The result is a clearer map of the gas mix that other teams can check against the classic spectral route.
Catalog record
| CREATOR | Minjie Lei, S. E. Clark, Rudy Morel, E. Allys, Iryna S. Butsky, Caleb Redshaw, Drummond B. Fielding |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-20407
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Neutral gas phase distribution from HI morphology: phase separation with scattering spectra and variational autoencoders Journal: ApJ astropy ( Astropy Collaboration et al. 2013 ) , scattering ( Cheng & Ménard 2021 ) , NumPy ( van der Walt et al. 2011 ) , healpy ( Górski et al. 2005 ) , PyTorch ( Paszke et al. 2019 ) Minjie Lei, S. E. Clark, Rudy Morel, E. Allys, Iryna S. Butsky, Caleb Redshaw, Drummond B. Fielding Abstract Unraveling the multi-phase structure of the diffuse interstellar medium (ISM) as traced by neutral hydrogen (H I) is essential to understanding the lifecycle of the Milky Way. However, H I phase separation is a challenging and under-constrained problem. The neutral gas phase distribution is often inferred from the spectral line structure of H I emission. In this work, we develop a data-driven phase separation method that extracts H I phase structure solely from the spatial morphology of H I emission intensity structures. We combine scattering spectra (SS) statistics with a Gaussian-mixture variational autoencoder (VAE) model to: 1. derive an interpretable statistical model of different H I phases from their multi-scale morphological structures; 2. use this model to decompose the 2D channel maps of GALFA-H I emission in diffuse high latitude ( | b | > 30 |b|>30 °) regions over narrow velocity channels ( Δ v = 3 \Delta v=3 km/s) into cold neutral medium (CNM), warm neutral medium (WNM), and noise components. We integrate our CNM map over velocity channels to compare it to an existing map produced by a spectrum-based method, and find that the two maps are highly correlated, while ours recovers more spatially coherent structures at small scales. Our work illustrates and quantifies a clear physical connection between the H I morphology and H I phase structure, and unlocks a new avenue for improving future phase separation techniques by making use of both H I spectral and spatial information to decompose H I in 3D position-position-velocity (PPV) space. These results are consistent with a physical picture where processes that drive H I phase transitions also shape the morphology of H I gas, imprinting a sparse, filamentary CNM that forms out of a diffuse, extended WNM. I Introduction Neutral atomic hydrogen (H I), as traced by the 21cm hyperfine transition line, is a ubiquitous and critical component of the interstellar medium (ISM). The neutral ISM as traced by H I emission is composed of the cold neutral medium (CNM), the cold, progenitor material from which star-forming molecular clouds develops; the warm neutral medium, the diffuse gas that extends to the disk-halo interface of galaxies; and the unstable medium, the transitory state between cold and warm phase ( McClure-Griffiths et al. 2023 , see recent review: ) . Resolving the underlying phase structure H I emission observations, which contain contributions from all H I phases, is therefore essential to understanding the transfer of matter and energy in galaxies across a wide range of environments and scales. However, direct observational determinations of H I phases are limited by the availability of H I absorption measurements, which are needed to constrain H I optical depth and derive temperature and density properties ( Heiles & Troland 2003 ; Murray et al. 2018 ) . While H I emission lines are ubiquitous along every line of sight (LOS) ( Kalberla & Kerp 2009 ; Winkel et al. 2016 ) , H I absorption is limited by the availability of background continuum emission sources. This fundamentally limits our ability to make spatially-resolved, phase-separated H I maps. As a result, recent work has focused on the development of methods that can extract phase information from H I emission data alone. A common approach is decomposing the H I emission spectra into Gaussian components and determining the properties of each component based on its width and amplitude ( Takakubo & van Woerden 1966 ; Mebold 1972 ; Haud & Kalberla 2007 ; Kalberla & Haud 2018 ) . However, many sightlines have complex line profiles that cannot be uniquely fit by Gaussian components, further complicated by factors such as velocity blending and systematics. More sophisticated techniques have been developed to mitigate these issues including regularization and automated component selection ( Marchal et al. 2019 ; Riener et al. 2020 ) . Still, the limitations of Gaussian decomposition have prompted the active development of alternative approaches, such as Fourier transform thresholding ( Marchal et al. 2024 ) . Advances in realistic multi-phase ISM simulations have also enabled the application of supervised learning approaches, where neural network models are trained on simulated H I spectra, and then used to predict the mass fraction of CNM ( f_{\mathrm{CNM}} ) for a given observed H I emission sightline ( Murray et al. 2020 ; Nguyen et al. 2025 ) . However, differences between the synthetic and observed H I spectra can contribute to bias in the model predictions. In short, inferring H I phase structure from H I emission data is an important problem that has prompted active development of many innovative methods that furthered our understanding of how spectral line information connects to H I gas properties. However, in the absence of spatially-resolved absorption line measurements, particularly towards diffuse regions, H I phase separation remains a challenging, under-constrained problem. In this work, we propose a new data-driven, morphology-based component separation method that complements and advances existing approaches in two important ways: by utilizing spatial structure of H I emission, and learning phase separation directly from data without relying on simulation training. Previous phase-separation methods use only the line-of-sight velocity structure from PPV H I emission data, with some work making limited use of spatial information to ensure coherence of derived structures ( Marchal et al. 2019 ) . However, there is significant information about H I phase structure encoded in the spatial dimension of H I emission as well ( Lei & Clark 2023 ) . Large-scale, high spatial-resolution H I emission observations and simulations have both pointed to a general picture of a dense, sparse CNM component, compared to a diffuse, more extended WNM component ( Kim et al. 2014 ; Peek et al. 2018 ; Fielding et al. 2023 ; McClure-Griffiths et al. 2023 ) . Moreover, studies of H I emission intensity structures have revealed the presence of magnetically-aligned, filamentary H I fiber features that are preferentially associated with the CNM ( Clark et al. 2014 ; Clark et al. 2015 ; Kalberla & Kerp 2016 ; Kalberla & Haud 2018 ; Peek & Clark 2019 ; Clark et al. 2019 ) . In general, the magnetized turbulence that shapes the morphology of H I emission also affects the formation of the CNM out of the WNM ( Villagran & Gazol 2018 ; Das & Gronke 2024 ; Granda-Muñoz et al. 2025 ) . Quantifying the distinct morphology features encoded in each H I phase is an important step to better understand the physical processes that shape the multiphase structure of the ISM. Motivated by these results, in an earlier work we developed a flexible quantification of H I emission morphology using scattering transform (ST) statistics ( Allys et al. 2019 ; Cheng & Ménard 2021 )