- The paper establishes that 228 neutral hydrogen clouds trace an accelerating nuclear outflow, with velocities reaching approximately 500 km s⁻¹ by 4 kpc from the Galactic center and kinematics matching ionized gas.
- The paper finds that neutral clouds fade and become more uniform with increasing Galactocentric distance, supporting progressive stripping and ionization as they interact with the hot wind.
- The paper identifies a sharp neutral-gas boundary near z ≈ 2 kpc and an outer azimuthal asymmetry, providing key constraints on models of the Fermi Bubbles and their multiphase outflow.
This paper presents a comprehensive analysis of 228 high-velocity neutral hydrogen (H I) clouds entrained in the Milky Way's nuclear wind, based on new and archival Green Bank Telescope (GBT) surveys covering approximately 500 square degrees around the Galactic center. The work establishes the most detailed kinematic picture of a neutral galactic nuclear wind available for any galaxy, at 9.1′ angular resolution (22 pc at the Galactic center). The principal results are: H I clouds reach Local Standard of Rest (LSR) velocities of −335≤VLSR≤+438 km s−1, the largest ever reported for neutral gas associated with the Galactic disk; the outflow accelerates linearly with distance from the Galactic center, from low velocities near the nucleus to approximately 500 km s−1 at r≲4 kpc; the neutral cloud population terminates abruptly at z≈2 kpc from the Galactic plane; and the highest-velocity, highest-latitude clouds exhibit an azimuthal asymmetry inconsistent with a symmetric outflow.
Observations and cloud sample
The data derive from GBT L-band observations taken between 2020 and 2022, combined with earlier GBT surveys, all re-reduced uniformly. Spectra were acquired with in-band frequency switching, smoothed to 3.02 km s−1 channels, and gridded into a data cube spanning −585≤VLSR≤+584 km s−1. A 100 sq-deg region between −5∘≤ℓ≤+5∘ and +5∘≤b≤+15∘ was observed to an rms noise of 20.5 mK per channel, while the southern portion (−10) has a mean noise of 52 mK. Because the noise is identical at positive and negative velocities in each spectrum, cloud detectability is independent of −11.
Clouds were identified as emission isolated in position and velocity from Galactic disk gas, restricted to −12 km s−13 where blending with foreground disk emission is minimal. Gaussian fits to spectra averaged over −14 regions yielded peak brightness temperature, line width, velocity, and column density under the optically thin assumption. The resulting catalog contains 228 clouds with a median −15 of 0.20 K, median −16 of −17 cm−18, and median FWHM of 24 km s−19. The number of clouds increases toward lower −10 down to the selection threshold, implying that many more lower-velocity or fainter clouds exist below the survey's sensitivity.
A key interpretive assumption, stated plainly by the authors, is that the clouds originated near the Galactic center and are entrained in a hot wind that accelerated them, and that they are confined within the Fermi Bubble (FB) boundaries as defined by the Miller & Bregman (2016) X-ray model. The paper cannot directly connect the acceleration mechanism of the neutral clouds to the mechanism that created the FBs themselves. Notably, the clouds are detected outside the inner filled bubble of that model, so the outer shock locus is adopted as the FB boundary throughout.
Kinematics and the accelerating outflow
Two features of the velocity-latitude distribution carry the central kinematic information. Negative-velocity clouds are found across the full surveyed latitude range (−11), whereas positive-velocity clouds are absent at −12 — a gap the authors demonstrate cannot arise from any observational selection effect. Additionally, positive-velocity clouds extend to roughly 100 km s−13 higher −14 than negative-velocity clouds.
Because projection effects alone would favor high −15 (a "velocity crowding" in which large path lengths through the bubble have −16), the observed preponderance of clouds at low −17 requires that the outflow velocity −18 vary with position. Placing each cloud at the nearest FB wall — where the projection factor is maximal — yields a minimum required −19 that increases with Galactocentric distance r≲40. This cannot be a projection artifact, since projection is most favorable at low latitude and low longitude, i.e., low r≲41. The data are therefore fit with an accelerating model, designated M-1:
r≲42
with r≲43 held at 500 km sr≲44 beyond 4 kpc. The authors note that this model is unconstrained at r≲45 kpc and r≲46 kpc, so the near-center behavior remains speculative. The result is robust to placing clouds 0.5 kpc inside the wall, which shrinks the inferred outflow slightly but preserves the required gradient. A constant-velocity outflow (model M-0) is explicitly rejected by the simulations, confirming an earlier suggestion by Lockman (2020).
Kinematic comparison with ultraviolet absorption lines from highly ionized species (C IV, Si IV, N V) toward the AGN PDS 456 and the star LS 4825 shows that the ionized and neutral gas kinematics are indistinguishable under model M-1, indicating a genuinely multiphase wind.
Cloud evolution within the wind
Once clouds are assigned 3D positions via model M-1, a monotonic decline in median r≲47 with Galactocentric distance emerges, fit by r≲48 K; r≲49 declines identically, as expected given the lack of FWHM variation. Crucially, there is no corresponding dependence on distance from the Sun: bright clouds appear over z≈20 kpc while faint clouds dominate at z≈21 kpc regardless of z≈22. The relative dispersion in z≈23 falls by an order of magnitude, from roughly 1 near the center to about 0.1 at large z≈24 (slope z≈25; Spearman z≈26, z≈27), with no significant correlation with z≈28.
These trends indicate that clouds are structurally evolving as they are advected outward — becoming fainter, more diffuse, and more uniform. This is consistent with high-resolution imaging of individual clouds and with molecular dissociation measurements, and supports a picture in which ram-pressure interaction with the hot wind progressively strips and ionizes the neutral material.
Vertical cutoff at z≈29 kpc
The most striking structural result is the complete absence of positive-velocity H I clouds over more than 100 sq-deg at −10, which under model M-1 corresponds to a hard upper boundary at −11 kpc from the Galactic plane. The authors demonstrate this void is not a sensitivity artifact: negative-velocity clouds are detected in the same region of sky, and simulations with a filled FB volume predict clouds there that are not observed. The boundary tracks distance from the plane rather than distance from the Galactic center — clouds at a given −12 span a wide range of −13. Even the ionized absorption at −14 with −15 km s−16 maps to −17 kpc under the model.
The physical interpretation offered is that clouds are disrupted by the hot wind and their neutral hydrogen converted to ionized gas as they rise, with the ionized material potentially contributing to the diffuse high-latitude absorption phase. This is consistent with hydrodynamical simulations of cloud–wind interactions, though the paper does not demonstrate the conversion directly.
Simulations, selection effects, and remaining discrepancies
Monte Carlo simulations populate the MB16 FB volume with clouds at constant number per −18, assign velocities from M-1, and forward-model the survey's sky coverage, noise variation, beam dilution (−19 empirically), and a −585≤VLSR≤+5840 detection threshold. These simulations show the survey is mainly informative over −585≤VLSR≤+5841 kpc. Three discrepancies persist:
- A void near −585≤VLSR≤+5842 that simulations do not naturally produce, possibly indicating real spatial structure in the neutral outflow.
- An excess of negative-velocity clouds relative to simulations, which geometrically favor positive −585≤VLSR≤+5843 by roughly 65% to 35%. When clouds at −585≤VLSR≤+5844 are excluded, the observed counts are nearly equal (103 negative, 104 positive). The authors attribute this to an imperfect treatment of cloud faintness at large −585≤VLSR≤+5845 and −585≤VLSR≤+5846, and predict many positive-velocity clouds lie just below the current detection limit.
- The longitude–velocity asymmetry discussed below.
Azimuthal asymmetry in the outflow
For the approximately 70% of clouds at −585≤VLSR≤+5847, the longitude–velocity distribution is consistent with azimuthal symmetry and shows no signature of Galactic rotation. Among the highest-velocity clouds at −585≤VLSR≤+5848 and −585≤VLSR≤+5849 km s−10, however, positive-velocity clouds cluster at −11 and negative-velocity clouds at −12. The authors rule out Galactic rotation as the explanation on three grounds: the expected −13 trend is absent, the mean velocities are strongly asymmetric (−14 versus −15 km s−16), and there is no symmetry about −17.
Instead, the pattern is reproduced by a pure radial outflow whose cloud distribution is confined to an ellipse in the −18–−19 plane — semi-major axis approximately 4 kpc along the FB wall, minor axis 1 kpc, rotated to −5∘≤ℓ≤+5∘0 — implying the outflow becomes azimuthally elongated beyond roughly 2.5 kpc from the center. The authors caution this cannot be the complete picture, since both positive and negative velocity ionized components are seen toward PDS 456 at −5∘≤ℓ≤+5∘1, outside the inferred elongated structure.
Limitations and open questions
The analysis carries several stated dependencies. The identification of clouds with the FB volume and the assumption of a purely radial wind are foundational but not directly testable with these data. The outflow model is unconstrained below −5∘≤ℓ≤+5∘2 kpc, where low-velocity clouds are confused with disk emission. The southern FB is undersampled at higher noise (52 mK versus 20.5 mK in the north), leaving open whether the −5∘≤ℓ≤+5∘3 kpc cutoff is symmetric to the northern boundary and whether southern azimuthal asymmetries exist at −5∘≤ℓ≤+5∘4. The discrepancy in the positive-to-negative velocity cloud ratio suggests the brightness evolution model is incomplete. Finally, the relationship between the low-latitude H I-traced clouds and the high-latitude UV-traced ionized population remains physically unclear, and the recent detection of negative-velocity clouds at −5∘≤ℓ≤+5∘5 implies outflow velocities −5∘≤ℓ≤+5∘6 km s−5∘≤ℓ≤+5∘7 at −5∘≤ℓ≤+5∘8 kpc — above the neutral cloud limit found here — which the current model does not fully accommodate.
Conclusion
This work establishes the accelerating character of the Milky Way's neutral nuclear outflow, quantifies its vertical extent, and documents the progressive disruption of entrained clouds. The combination of a linear velocity law reaching 500 km s−5∘≤ℓ≤+5∘9, a sharp neutral boundary at +5∘≤b≤+15∘0 kpc, and azimuthal asymmetry at large radius provides empirical constraints that any model of the Fermi Bubbles and their interaction with the interstellar medium must satisfy. Resolving the southern hemisphere coverage gap and the nature of the ionized-neutral transition are the most immediate outstanding observational needs raised by these results.
