Cold Neutral Medium (CNM) Overview

• CNM is a phase of the atomic interstellar medium defined by low temperatures (20–100 K), moderate to high hydrogen densities (10–100 cm⁻³), and significant 21 cm optical depth.
• Filamentary, magnetically aligned structures with log-normal column densities and narrow linewidths characterize the CNM, linking its morphology to molecular cloud precursors.
• Turbulence, thermal instabilities, and magnetic pressure shape CNM formation and evolution, affecting its mass fraction and the initiation of star formation in galaxies.

The cold neutral medium (CNM) is a key phase of the atomic interstellar medium (ISM), characterized by low temperatures (typically 20–100 K), modest to high hydrogen volume densities (10–100 cm⁻³), and significant 21 cm optical depth. Composed predominantly of neutral hydrogen (H I) with a minor contribution from helium and dust—including polycyclic aromatic hydrocarbons (PAHs)—the CNM exists in rough pressure equilibrium with the warm neutral medium (WNM, T ~ 10³–10⁴ K). The CNM structure is dominated by filamentary, magnetically aligned sheets, and it forms the direct atomic precursor to molecular clouds and star formation in galaxies.

1. Physical Properties, Phase Structure, and Distribution

The CNM is defined by thermodynamic and chemical criteria: kinetic temperatures T ≈ 20–100 K, hydrogen volume densities n_H ~ 20–100 cm⁻³, and thermal pressure P/k ~ 2×10³–5×10³ K cm⁻³ in the solar neighborhood and star-forming disks. All-sky H I spectral decompositions, such as those using HI4PI data, show that CNM line components possess typical FWHM ≈ 3.6 km/s (σ_v ≈ 1.5 km/s), corresponding to Doppler temperatures T_D ~ 280 K, though removal of turbulent broadening yields true kinetic temperatures in the 50–100 K range (Kalberla et al., 2018).

CNM material is not volume-filling. At high Galactic latitudes, it occupies only ≲1% of the sky in area but may cover a much larger fraction in projection due to elongated filamentary geometry. The global CNM fraction is modest: averaging f_CNM ~ 0.25 by mass and locally rising to ≳0.5 in CNM-dominated regions (Kalberla et al., 2018).

Column densities in individual filaments peak around N_HI ~ 10¹⁹ cm⁻², with log-normal distributions and typical volume densities bounded by magnetic and geometric constraints (n_HI ~ 14–47 cm⁻³ for d ~ 0.09–0.3 pc at 100 pc distances) (Kalberla et al., 2016). Observed vertical scale heights in the solar neighborhood are σ_z ~ 50–60 pc for optically thick (τ>0.1, N_HI>10¹⁹.⁵ cm⁻²) CNM, with a thicker, more diffuse component at σ_z ~ 120 pc (Rybarczyk et al., 2024). The thin CNM layer coincides with the molecular cloud disk (CO σ_z ~ 40–70 pc), suggesting a shared vertical distribution for H I and CO-bright gas (Rybarczyk et al., 2024).

2. Morphology: Filaments, Magnetic Fields, and Multiphase Environment

The CNM is dominantly organized in narrow, sheet-like filaments and caustics, often detected in both H I emission and FIR maps, and strongly aligned with the plane-of-sky magnetic field as traced by Planck dust polarization (Kalberla, 3 Feb 2025, Kalberla et al., 2016, Lei et al., 2022). High-resolution Hessian filtering and morphological metrics—such as the scattering transform—quantify the linearity, coherence, and alignment of these filaments, demonstrating that increased small-scale linearly-aligned structure is tightly correlated with higher local CNM fraction (Lei et al., 2022).

Unsharp masking, wavelet transforms, and statistical tools show that these filaments reveal log-normal distributions of Doppler temperatures and column densities, occupy only a small area fraction, and are nested within more diffuse warm (WNM) or unstable (LNM) envelopes. Anti-correlation between CNM filaments and the LNM supports a physical scenario where cold sheets condense inside larger unstable layers (Kalberla et al., 2018). At high latitude, every sightline with detectable H I absorption is associated with a filamentary caustic discerned in FIR or H I structure (Kalberla, 3 Feb 2025).

Magnetic fields play a dominant role in shaping and confining these filaments. For typical field strengths B ≈ 6 μG, the magnetic pressure P_B/k ≈ 10⁴ K cm⁻³ matches the internal gas pressure of the CNM, and filaments are seen as edge-on sheets with their orientation parallel to the local B-field (Kalberla et al., 2016). Anisotropy in CNM turbulence is evident down to ~0.2 mpc scales (40 au), with power spectra (P(k) ~ k^α, α ≈ −3.0 to −3.5) that remain break-free and highly anisotropic, further aligning with the ordered magnetic field (Vigoureux et al., 3 Feb 2026).

3. Formation, Evolution, and Interplay with Turbulence and Instabilities

The CNM emerges as the end-state of thermal and dynamical instabilities in the ISM. Classical two-phase models identify a narrow pressure window (P_min < P < P_max) supporting stable CNM–WNM coexistence, connected through an unstable neutral medium (UNM) (Iwasaki et al., 2012).

High-resolution hydrodynamic simulations of converging WNM flows illuminate two regimes: for small upstream density fluctuations (δ≤10%), shocks efficiently condense gas to CNM (cold mass fraction f_CNM ~ 0.7, post-shock turbulence σ_turb ~ 2–3 km/s); for δ>10% (realistic ISM), large post-shock turbulence (σ_turb > 3 km/s) limits CNM formation (f_CNM ~ 0.45) and results in a strongly bimodal, multiphase layer (Kobayashi et al., 2020). The critical resolution for converged macroscopic properties is set by the cooling length of the thermally unstable phase, λ_c ~ 0.05–0.2 pc.

At condensation fronts (e.g., WNM→CNM interfaces), the front thickness is set by the Field length (λ_F) for low mass flux, and by the cooling length (λ_c) for high-j, shock-compressed gas (Iwasaki et al., 2012). In pressurized/shocked settings (P_CNM ≫ P_max), direct dynamical condensation into CNM occurs, even bypassing stable WNM.

Recent studies have identified a previously unrecognized “UNM instability”: in the thermally unstable phase, the effective polytropic index n = d ln P/d ln ρ is negative, so the pressure force is attractive, causing self-contraction of UNM and setting a ceiling on the aspect ratio of CNM filaments (ℓ∥/ℓ⊥ ~ 5–20) unless arrested by conversion to CNM or magnetic tension (Ho et al., 2021). This is significantly less than the observed aspect ratios >60–200 in HI filaments, suggesting that many observed ultra-long filaments are either UNM-dominated or transient, non-equilibrium configurations.

Turbulence is highly anisotropic in the CNM, with power spectra consistent with subsonic MHD cascade (Iroshnikov–Kraichnan), and the turbulent energy cascade extends uninterrupted through all observed scales down to at least 40 au, with no evidence for a dissipation cutoff within the JWST range (Vigoureux et al., 3 Feb 2026). Turbulent velocities are measured at δv_turb ~ 2.5–3.9 km/s, with Mach numbers M_t ~ 3–4, confirming highly supersonic (for CNM) but subsonic with respect to thermal sound speed (Kalberla, 3 Feb 2025, Kalberla et al., 2018).

4. CNM Fraction, Metallicity Dependence, and Star Formation Links

Galaxy-scale surveys and targeted absorption studies show that the CNM mass fraction is environment-dependent but typically in the range 0.2–0.4 in the Milky Way and similarly in metal-rich Local Group spirals. In low-metallicity dwarfs (e.g., IC 10, NGC 6822, SMC), CNM is detected locally with lower spin temperatures (T_s ~ 30–55 K for IC 10, 32 ± 6 K for NGC 6822), with localized cold-gas mass fractions of 0.12–0.37 (Pingel et al., 2024, Stelea et al., 13 Nov 2025). These localized detections reveal intimate spatial and kinematic association of CNM with molecular gas (CO, HCO⁺, HCN) and star-forming regions, confirming the role of the CNM as a self-shielding precursor and coolant before molecular condensation.

In simulated spiral galaxies, the CNM fraction is nearly constant at f_CNM ~ 0.2 across the star-forming disk (Σ_CNM rising with midplane pressure P_T), with most of the CNM confined to clumpy structures—covering fraction ~5.5%—that closely trace spiral arm and H₂ locations (Smith et al., 2023). The observed scaling between CNM surface density and star formation rate surface density (Σ_SFR ∝ Σ_CNM^1.56) closely parallels the classical Kennicutt-Schmidt relation for total gas, supporting the CNM as a key determinant for the onset of self-gravitating, molecule-rich star-forming clouds (Smith et al., 2023).

5. Observational Methodologies and Phase Diagnostics

Direct measurement of CNM properties historically relies on 21 cm H I emission–absorption pairing. Under the isothermal approximation, spin temperature per velocity channel is

$T_s(v) = \frac{T_B(v)}{1-e^{-\tau(v)}}$

and column density is derived as

$$N_\mathrm{HI} = 1.823 \times 10^{18}\, \mathrm{cm}^{-2} \int T_s(v)\, \tau(v)\, dv$$

(Stelea et al., 13 Nov 2025, Pingel et al., 2024). Gaussian decomposition is essential to isolate multiple H I components.

Phase fractions (f_CNM, f_LNM, f_WNM) are mapped by detailed spectral Gaussian decomposition or, more recently, by deep-learning methods (CNNs) trained on synthetic spectra that assimilate realistic turbulence, line blending, and radiative transfer (Hensley et al., 2021, Murray et al., 2020). The CNN-inferred f_CNM maps correlate strongly with direct absorption-based measures and are especially suited for high-latitude, low-complexity profiles (Murray et al., 2020).

Statistical and morphological classification—e.g., via wavelet scattering transforms—demonstrates that spatial filamentarity, as quantified by second-order coefficients, is a strong predictor of f_CNM when validated against absorption-based benchmarks (Lei et al., 2022). Morphological analysis shows that CNM-rich regions have both enhanced small-scale FIR/HI ratios and characteristic linear structures, reinforcing the spatial–phase link.

6. CNM and Dust, PAHs, and Foreground Emission

Cross-correlation studies reveal that regions with high CNM content are also enriched in PAHs, as shown by increased mid-IR (12 μm / FIR) dust surface brightness; PAHs are more efficiently preserved and less destroyed in the denser, shielded CNM compared to the more diffuse WNM (Hensley et al., 2021). However, the anomalous microwave emission (AME) per unit dust, presumed to be spinning-dust emission from PAHs, shows no direct increase with f_CNM, suggesting environmental dependencies in PAH properties and emission efficiencies.

The optically thin assumption underpredicts total dust extinction (E(B–V)) in CNM-rich regions by up to ~20% due to optical-depth and phase-structure effects, implying the necessity of correcting for the CNM when constructing all-sky reddening and CMB foreground models (Murray et al., 2020).

7. Scaling Laws, Power Spectra, and Structural Statistics

Absorption-based studies combined with high-resolution interferometry recover the real-space power spectrum slope α_HI of CNM column density fluctuations,

$P_\tau(U) \propto U^{-(2+\alpha)}$

with α_HI typically 0.5–1 at pc–sub-pc scales, corresponding to power-law indices observed in emission and absorption (e.g., −2.5 to −3.0) (Vishwakarma et al., 2019). This self-similar scaling, extending across orders of magnitude in scale, points to a scale-free, turbulent cascade that shapes CNM structure down to the microphysical dissipation limits.

Summary Table: Benchmark Physical and Observational Properties

Parameter	Typical Value/Range	Reference
Kinetic temperature, T	20–100 K	(Kalberla et al., 2018, Kalberla, 3 Feb 2025)
Volume density, n_H	10–100 cm⁻³	(Kalberla, 3 Feb 2025, Kalberla et al., 2016)
Column density, N_HI	(0.1–3)×10¹⁹ cm⁻² (filaments)	(Kalberla et al., 2016, Kalberla, 3 Feb 2025)
Spin temperature, T_s	20–200 K (MW), ~30 K (low-Z)	(Kalberla, 3 Feb 2025, Pingel et al., 2024, Stelea et al., 13 Nov 2025)
CNM fraction, f_CNM	0.1–0.4 (environment-dependent)	(Smith et al., 2023, Stelea et al., 13 Nov 2025)
Scale height, σ_z	50–60 pc (dense), 120 pc (diffuse)	(Rybarczyk et al., 2024)
Filament aspect ratio	5–20 (theory), ≳60–200 (obs.)	(Ho et al., 2021)
Turbulent Mach number	~3–4 (highly supersonic)	(Kalberla, 3 Feb 2025, Kalberla et al., 2018)
Magnetic field, B	~6 μG (solar neighborhood)	(Kalberla et al., 2016)

• (Kalberla et al., 2016) Cold Milky Way Hi gas in filaments
• (Kalberla et al., 2018) Properties of cold and warm HI gas phases derived from a Gaussian decomposition of HI4PI data
• (Vishwakarma et al., 2019) HI column density statistics of the cold neutral medium from absorption studies
• (Murray et al., 2020) Extracting the cold neutral medium from HI emission with deep learning: Implications for Galactic foregrounds at high latitude
• (Ho et al., 2021) How the existence of unstable neutral media restricts the aspect ratio of cold neutral media?
• (Hensley et al., 2021) Polycyclic Aromatic Hydrocarbons, the Anomalous Microwave Emission, and Their Connection to the Cold Neutral Medium
• (Lei et al., 2022) Probing the cold neutral medium through HI emission morphology with the scattering transform
• (Smith et al., 2023) On the distribution of the Cold Neutral Medium in galaxy discs
• (Pingel et al., 2024) The Local Group L-Band Survey: The First Measurements of Localized Cold Neutral Medium Properties in the Low-Metallicity Dwarf Galaxy NGC 6822
• (Rybarczyk et al., 2024) Revisiting the Vertical Distribution of HI Absorbing Clouds in the Solar Neighborhood. II. Constraints from a Large Catalog of 21 cm Absorption Observations at High Galactic Latitudes
• (Kalberla, 3 Feb 2025) The cold neutral medium in filaments at high Galactic latitudes
• (Stelea et al., 13 Nov 2025) The Local Group L-band Survey: Probing Cold Atomic Gas in IC10 with Neutral Hydrogen Absorption
• (Park et al., 9 Dec 2025) A High-resolution Study of the Cold Neutral Medium in and around 30 Doradus
• (Vigoureux et al., 3 Feb 2026) JWST imaging of the Pleiades: anisotropy of turbulence in the cold neutral medium