Seven new feedback tests, from episodic UFOs to early-universe neutral winds (2026)

The week of September 7 to September 13, 2026 highlights a central challenge in feedback physics: distinguishing genuine kinematic gas ejection from radiative transfer and resolution artifacts across vastly different physical scales. The six preprints and one open codebase analyzed below track feedback coupling from sub-parsec accretion disks and parsec-scale protostellar core interactions to 20-kiloparsec galactic winds, high-redshift quiescent halos, and cosmological cluster emulators.

At a glance

Item	Design / Method	Main result	Scale or tracer
Victoria-Ceballos et al.	High-resolution XMM-Newton X-ray spectroscopy of low-luminosity quasar Mrk 205	Two outflow components at `~0.07c to ~0.1c`; marginal Fe K detection; indicates episodic wind launches	Sub-parsec to nuclear; ultra-hot gas and Fe K absorption
Luo et al.	ALMA Bands 3 and 6 CO (2-1) and 1.3 mm continuum with 3D dendrogram linking in IRAS 16119-5048	16 radial streamers converging on explosive center; kinetic energy is `≥1 dex` below typical explosive outflows	`~0.1–1 pc`; cold molecular CO and dense fragmented cores
Oh et al.	Stacking analysis of Na I D absorption in 274 galaxies at `z = 2–5` using JWST/NIRSpec	Ubiquitous neutral outflows in quiescent galaxies with mass-loading factors elevated by `2–4 dex` and \(v_{\mathrm{out}} \sim 310\text{--}570\ \mathrm{km\ s}^{-1}\)	Galaxy to inner CGM; cool neutral Na I D gas
Magee et al.	Synthetic down-the-barrel UV absorption spectra from 20-kpc CGOLS simulations (\(\Delta x = 5\text{--}20\ \mathrm{pc}\))	Velocity statistics match HST data, but fiducial equivalent widths are `50%–97%` lower than empirical relations	`5–20 pc` resolution across `20 kpc`; multiphase UV ions
Kym et al.	3D Monte Carlo Lyα radiative transfer in rotating CGM models	Rotation broadens profiles and creates a severe degeneracy between \(V_{\mathrm{rot}}\) and \(N_{\mathrm{HI}}\), mimicking inflow/outflow	Halo-scale CGM; resonant Lyα emission
Lee et al.	Conditional flow matching (BIND) trained on 1,024 CAMELS SB35 hydrodynamic halos across 35 parameters	Gas and stellar masses recovered to percent-level; radial profiles match to `≲10%` in minutes on one GPU	Halo to cluster scale; multi-field baryonic distribution
Di Tella (`aion-flow_v2`)	Open-source Python data pipeline for eROSITA DR2, DESI DR1, and Legacy Survey DR10 cutouts	End-to-end data preparation for probing AGN feedback signatures using the AION foundation model	Survey-scale cross-match; X-ray, optical, and spectroscopic data

Paper briefs

Episodic ultra-fast outflows emerge in a low-luminosity quasar

Authors and date. César Ivan Victoria-Ceballos and collaborators, submitted September 8, 2026. 1

Study design. The authors carry out a detailed spectral analysis of XMM-Newton high-resolution grating data and CCD observations of the low-luminosity quasar Mrk 205, targeting blueshifted absorption signatures in the soft X-ray band and the Fe K band. 1

Result. High-resolution grating data reveal two distinct ultra-fast outflow components with line-of-sight velocities spanning ~0.07c to ~0.1c. A blueshifted absorption feature is also marginally detected in the Fe K band. The spectral variability between observation epochs demonstrates that Mrk 205 undergoes episodic wind launching events rather than sustaining a steady wind. The paper evaluates whether UFO kinetic power, which typically ranges from \(10^{42}\) to \(10^{45}\ \mathrm{erg\ s^{-1}}\) in bright quasars, delivers sufficient energy to regulate star formation in lower-luminosity hosts. 1

Why read. Most ultra-fast outflow surveys focus on luminous quasars operating near the Eddington limit. This study provides empirical evidence that fast wind launching continues down into lower-luminosity regimes, while emphasizing that the duty cycle is episodic.

Caveat. The Fe K absorption feature is a marginal detection. Because line-of-sight absorption samples a single pencil beam, the total outflow solid angle, mass outflow rate, and kinetic coupling efficiency to the surrounding host interstellar medium remain model-dependent. 1

ALMA identifies a low-energy explosive outflow candidate

Authors and date. Yongquan Luo and collaborators, submitted September 7, 2026. 2

Study design. The authors analyze high-resolution ALMA Band 3 (ATOMS) and Band 6 (QUARKS) observations of the massive star-forming region IRAS 16119-5048, complemented by archival ATCA radio continuum and Spitzer mid-infrared imaging. They use dendrogram algorithms to isolate coherent structures in position-position-velocity space. 2

Result. Dendrogram linking of CO (2-1) emission uncovers 16 high-velocity, streamer-like structures whose projected trajectories converge toward a central core. The total kinetic energy of the outflow is at least an order of magnitude lower than that of classical Orion-type explosive outflows, but its mass entrainment and momentum injection rates significantly exceed standard protostellar bipolar jets. The 1.3 mm continuum resolves 27 dense cores along a filamentary network, showing evidence of mass segregation where the most massive cores cluster near the explosion center. The core separations match thermal Jeans or cylindrical fragmentation predictions, whereas collect-and-collapse triggering is unsupported. 2

Why read. The observations establish IRAS 16119-5048 as a prototype for low-energy explosive stellar feedback, likely triggered by close dynamical interactions or stellar mergers within dense protostellar clusters before supernova feedback begins.

Caveat. The complex kinematic environment makes it difficult to definitively rule out superposition from multiple overlapping bipolar outflows driven by individual protostars without higher angular resolution and proper motion measurements. 2

JWST stacked spectra reveal neutral outflows across early quiescent galaxies

Authors and date. Rion Oh and collaborators, submitted September 8, 2026. 3

Study design. The team conducts a spectral stacking analysis of rest-frame optical Na I D λλ5891, 5897 absorption lines in 274 galaxies spanning z = 2–5. The sample combines JWST/NIRSpec medium-resolution spectroscopy and NIRCam imaging from the DAWN JWST Archive (DJA) and the EMBER survey, contrasting quiescent galaxies against matched star-forming systems. 3

Result. Blueshifted Na I D absorption is detected in every quiescent galaxy stack between z = 2 and z = 5. Quiescent galaxies display mass-loading factors \(\eta = \dot{M}{\mathrm{out}}/\mathrm{SFR}\) that are 2–4 dex higher than those of star-forming controls. Their inferred mass outflow rates reach \(\dot{M}{\mathrm{out}} \sim 9\text{--}30\ M_\odot\ \mathrm{yr}^{-1}\) (elevated by ~0.5–1 dex), with velocities of \(v_{\mathrm{out}} \sim 310\text{--}570\ \mathrm{km\ s}^{-1}\) (elevated by a factor of ~2). In the statistically robust redshift bins at 2 <= z < 4, elevated [N II]/Hα and [O III]/Hβ emission ratios are consistent with non-stellar photoionization from an active galactic nucleus. 3

Why read. This study provides the first statistical confirmation that neutral gas outflows are ubiquitous in early quiescent galaxies. Because these galaxies have declining star-formation histories, their extreme mass-loading factors cannot be sustained by ongoing star formation, implicating AGN feedback in preventing gas reaccretion.

Caveat. Stacking averages out line-of-sight geometry and galaxy-to-galaxy variance. The derived mass outflow rates and loading factors depend on assumed gas-to-dust ratios, ionization corrections, and outflow radii. 3

Synthetic UV absorption lines expose resolution limits in wind modeling

Authors and date. Jake Magee, Evan Schneider, Kate H. R. Rubin, S. Alwin Mao, and Cameron Hummels, submitted September 9, 2026. 4

Study design. The authors generate synthetic down-the-barrel UV absorption spectra from 20-kiloparsec-scale multiphase outflow simulations in the CGOLS framework. The hydrodynamic calculations achieve spatial resolutions of \(\Delta x = 5\text{--}20\ \mathrm{pc}\). The resulting synthetic line profiles and velocity metrics are compared directly with high-resolution far-UV spectra from HST treasury programs. 4

Result. The velocity statistics measured from the synthetic data agree well with HST far-UV observations. Outflow velocity correlates strongly with ion ionization potential, mirroring the physical velocity-temperature relationship established in the hydrodynamic grid. However, fiducial equivalent widths are systematically lower than empirical relations by 50%–97%, with the discrepancy most pronounced in high-ionization transitions. Resolution tests demonstrate that equivalent widths are far more sensitive to numerical grid refinement than centroid velocities. 4

Why read. The paper provides a direct benchmark for calibrating hydrodynamic winds against down-the-barrel absorption surveys. It demonstrates that matching observed velocity centroids does not guarantee that the cool and warm gas column densities are correctly resolved.

Caveat. The synthetic spectra in this run employ an idealized uniform ionizing background rather than local radiation transfer from star clusters. The simulations model an isolated disk outflow rather than a cosmological circumgalactic environment. 4

Circumgalactic rotation creates degeneracies in Lyα emission profiles

Authors and date. Hee-Gyeong Kym, Seok-Jun Chang, Max Gronke, and Kwang-Il Seon, submitted September 8, 2026. 5

Study design. The authors perform 3D Monte Carlo Lyα radiative-transfer simulations through rotating circumgalactic gas halos. They systematically investigate how emergent line profiles depend on rotational velocity \(V_{\mathrm{rot}}\), neutral hydrogen column density \(N_{\mathrm{HI}}\), viewing inclination, medium clumpiness, and intrinsic source emission width. 5

Result. Circumgalactic rotation broadens spatially integrated Lyα profiles and widens their peak separation, showing the strongest inclination dependence when rotational Doppler shifts dominate over frequency diffusion. At high neutral hydrogen column densities, repeated resonant scattering washes out rotational sensitivity, creating a strong degeneracy between \(V_{\mathrm{rot}}\) and \(N_{\mathrm{HI}}\). Using peak separation alone to infer column density in rotating halos leads to systematic overestimates. Spatially resolved IFU spectra break this degeneracy through opposing redshifted and blueshifted asymmetries, but these rotational shears can also mimic radial outflow or inflow patterns on projected 2D velocity maps. 5

Why read. Resonant Lyα emission is a primary tracer of cold circumgalactic gas kinematics. This work maps out where rotational kinematics contaminate radial inflow and outflow diagnostics, establishing requirements for interpreting integral field unit observations.

Caveat. The calculations assume idealized rotating spheres or simple clumpy gas media rather than fully turbulent, multiphase hydrodynamic halo structures with realistic velocity shear. 5

BIND emulates field-level baryonic feedback across 35 cosmological parameters

Authors and date. Max E. Lee and collaborators, submitted September 9, 2026. 6

Study design. The authors introduce BIND (Baryonic INpainting with Deep learning), a conditional flow-matching architecture that learns a field-level mapping from dark-matter-only halos to full hydrodynamic halos. Training uses 1,024 paired volumes from the CAMELS 50 \(h^{-1}\) Mpc SB35 suite, spanning a 35-dimensional parameter space of cosmological (ΛCDM) and IllustrisTNG feedback parameters. 6

Result. BIND recovers total dark matter, gas, and stellar masses to percent-level precision. It reproduces azimuthally averaged radial density profiles to within ≲10% across all radii and faithfully reconstructs halo triaxial shapes. When deployed halo-by-halo on a \((50\ h^{-1}\ \mathrm{Mpc})^3\) N-body simulation volume with \(512^3\) particles, BIND replicates the projected matter power spectrum suppression induced by baryonic feedback in minutes on a single GPU, matching the fidelity of direct hydrodynamic halo substitution. 6

Why read. Baryonic feedback represents a major systematic uncertainty for cosmic shear and weak-lensing surveys. BIND provides a computationally tractable route to evaluate baryonic suppression across large simulation volumes without running thousands of expensive hydrodynamic simulations.

Caveat. The model is trained specifically on the IllustrisTNG feedback implementation within CAMELS. It does not evaluate alternative feedback schemes (such as SIMBA or EAGLE), so its predictions remain tied to that subgrid physics framework. 6

Code and data releases

AION-Flow v2 pipelines AGN feedback probes across eROSITA and DESI

Author and date. Rocco Di Tella, open repository released on GitHub September 11, 2026 and updated September 12. 7

Tool scope. The repository provides a reproducible data processing pipeline and model evaluation harness for investigating AGN feedback signatures. The pipeline coordinates cross-matched data from eROSITA DR2 X-ray catalogs, DESI DR1 spectroscopic redshifts, and Legacy Survey DR10 optical cutouts, coupling them to a frozen AION-1-B foundation model encoder with normalizing-flow heads under an open MIT license. 7

Why track. Applying foundation models to large multiwavelength sky surveys requires rigorous data pipelines that preserve provenance and handle multi-survey cross-matching. This codebase provides a concrete reference implementation for probing feedback signatures across surveys.

Caveat. The repository is distributed as companion research code without formal semantic release tags or packaging, and running the pipeline requires downloading primary survey datasets. 7

Data repository status

A targeted search of major astronomy repositories identified a new Zenodo deposition on September 11, 2026 for supplementary data on cool circumgalactic gas around DESI green valley galaxies 8. However, the underlying dataset files are currently restricted under a formal embargo extending to April 1, 2027. Under the channel's standard of including only verified, publicly downloadable datasets, no standalone external dataset qualifies for delivery this cycle.

What the week separates

The items in this issue expose how feedback physics changes character depending on the physical coupling scale and the gas phase chosen to observe it:

Coupling location and physical scale. Feedback energy is injected at radically different radii. In Mrk 205, Victoria-Ceballos et al. trace ultra-fast outflows launched from the accretion disk at sub-parsec scales. In IRAS 16119-5048, Luo et al. resolve stellar-feedback streamers driven by core interactions at sub-parsec to parsec scales. Magee et al. simulate winds propagating across 20 kiloparsecs of halo gas, Oh et al. measure neutral gas outflows extending into the circumgalactic environments of quiescent galaxies at z = 2–5, and Lee et al. map baryonic feedback across megaparsec-scale cosmic structures.
Phase diagnostics and observable tracers. Highly ionized, relativistic iron absorbers in XMM-Newton spectra trace hot, nuclear AGN winds. Millimeter CO (2-1) emission detects cold, dense molecular gas entrained by protostellar explosions. Optical Na I D absorption measures cool neutral gas being swept out of early galaxies. Ultraviolet metal lines (such as C IV and Si IV) sample multiphase cooling interfaces, while resonant Lyα scattering probes diffuse neutral circumgalactic envelopes. Comparing feedback models against only one of these phases inevitably misses the broader mass or energy distribution.
Duty cycle versus persistent memory. Mrk 205 confirms that UFOs can be episodic rather than steady, meaning a quiescent or low-luminosity nucleus may have launched energetic winds in the recent past. Conversely, the JWST stacking results from Oh et al. show that high mass-loading outflows are ubiquitous across z = 2–5 quiescent galaxies, implying that maintenance-mode feedback or lingering gas ejection persists long after primary star formation has ceased.
Kinematic reality versus observational artifacts. The papers from Magee et al. and Kym et al. provide essential cautions for interpreting gas flow data. Magee et al. demonstrate that while hydrodynamic simulations reliably reproduce observed velocity centroids, their equivalent widths are severely suppressed when cooling interfaces are under-resolved. Meanwhile, Kym et al. show that circumgalactic rotation alters Lyα peak separations and line asymmetries, creating profile distortions that can be easily misinterpreted as radial inflows or outflows. Validating feedback models requires treating radiative transfer and resolution effects with the same rigor as the underlying hydrodynamic calculations.

Seven new feedback tests, from episodic UFOs to early-universe neutral winds