Astrophysics papers — 2026-09-17

We finally have a better idea of how to find the stars that vanish without a trace. This is vital because it tells us how black holes actually form.

By modeling how massive stars evolve and collapse, researchers have found that most black hole progenitors are actually hot, blue, and ultraviolet-bright, often in a Wolf-Rayet phase. This means that if we only look for the classic red supergiants, we are likely missing a huge chunk of these direct-collapse events, which occur at a rate of about 0.4 per century in a galaxy like ours.

To catch them, we need to start looking at nearby star-forming galaxies through ultraviolet-sensitive lenses. This search for hidden cosmic signatures extends to the very beginning of the universe, where a slight tweak to fundamental physics might resolve long-standing tensions.

If the Higgs vacuum expectation value was about 1.58 percent higher during Big Bang Nucleosynthesis, we can reconcile the observed helium abundances with the deuterium and lithium levels we see today. This would imply that the universe evolved in a non-standard way between the Big Bang and the era of the cosmic microwave background.

A smaller uplift of 0.2 percent could also help fix the cosmic lithium problem while remaining consistent with current observations. Moving from the dawn of time to the violent deaths of stars, we are learning that gravitational waves can peek inside the dense envelopes that hide gamma-ray bursts.

While light gets trapped by the star, high-frequency gravitational waves can carry information about the central engine's variability. If the engine powering the jet fluctuates rapidly, it produces a signal peaking at tens of hertz that current ground-based detectors could actually see if the burst is nearby.

We can finally start moving through the massive data volumes expected from upcoming spectroscopic surveys by letting machine learning take the heavy lifting of stellar classification. A new framework using tree-based ensemble methods has successfully automated the classification of massive OB-type stars, reaching an impressive 98 percent accuracy for broad spectral types using the LGBM algorithm.

While fine-grained classification is harder, with Random Forest hitting 89 percent, the most complex task of combining spectral type with luminosity class only reached 77 percent accuracy with XGBoost. This drop is likely due to inherent degeneracies in how we classically define these stars rather than a failure of the models themselves.

To speed up our understanding of how galaxies actually grow, researchers have developed a graph neural network surrogate that can predict galaxy properties across cosmic time. By combining merger tree data with semi-analytical models, this tool can predict stellar mass and star formation rates with high precision, such as an R-squared value of up to 0.973 for stellar mass.

This effectively allows us to bypass the massive computational cost of traditional simulations. Moving from the evolution of entire galaxies to the local neighborhood, a new catalog called RMSTAR has mapped the nearest 3352 M dwarf systems.

This census reveals a multiplicity rate of about 8.7 percent for wide stellar companions, providing a clear look at the mass and luminosity functions of these red dwarfs. We are finally getting a clearer picture of how binary black holes might be born in the chaotic environments of active galactic nuclei, which is vital for understanding if these mergers are common or rare.

By combining gravitational wave data from the latest LVK catalogue with the Quaia AGN catalogue into a single statistical framework, researchers have found they cannot yet rule out the idea that a significant fraction of mergers happen within these massive accretion disks. For specific brightness ranges of unobscured AGN, they have been able to set upper limits on this contribution, effectively saying that the AGN channel can't account for more than roughly forty to sixty percent of mergers depending on the luminosity.

This search for hidden cosmic processes extends to the high-energy particles they leave behind. New modeling shows that if a short gamma-ray burst jet travels through the low-density cavity of an AGN disk, it should produce a distinct signature of high-energy neutrinos.

These neutrinos are created when protons hit photons from the burst itself or the disk, and the presence of that disk radiation actually shifts the neutrino energy peak to lower levels. For a standard supermassive black hole, this means next-generation detectors like IceCube-Gen2 could potentially spot these events as far as 400 megaparsecs away.

Detecting these signals requires incredibly clean data, which is where recent improvements in particle identification come in. IceCube has developed a way to better separate gamma rays from the much more common background of cosmic-ray protons by looking at the muon content and the spread of the air shower.

By using the deep detector to spot muons that shouldn't be there in a pure photon shower, they can suppress the hadronic background by more than a factor of a thousand. We finally have a smoking gun for why the solar transition region behaves so strangely.

By analyzing SiIV line profiles from IRIS mosaics, researchers have found that non-Gaussian, k-like velocity distributions are actually pervasive, appearing in 60% of observed profiles. These profiles appear most frequently when the magnetic field is oriented transverse to our line of sight, which strongly suggests these non-Maxwellian distributions are anisotropic and tied to magnetic energy release.

This discovery means our standard magnetohydrodynamic models are missing the vital kinetic physics happening in the solar atmosphere. The need for better modeling extends to how we track the evolution of galaxies and their central black holes.

A new Python framework called CosmoDyn allows researchers to take existing cosmological simulations and reconstruct the complex, time-dependent environments of galaxies. Instead of just looking at a static host, it models the gravitational influence of accreted satellite galaxies, letting us simulate how compact objects move through these shifting landscapes without the massive cost of running a new simulation from scratch.

Moving from the evolution of galaxies to the physics of their centers, we are seeing new ways that magnetic fields might emerge in accretion flows. While we usually rely on magneto-rotational instability to explain these fields, a new model shows that non-conservative radiation fields from a luminous corona can generate and amplify magnetic fields independently.

This radiation-driven mechanism can work just as fast as the standard instability, providing a dominant channel for magnetism in everything from active galactic nuclei to gamma-ray bursts. The eROSITA mission has finally provided the massive statistical foundation we have been waiting for to understand how galaxy groups and clusters evolve.

By analyzing a high-purity sample of 3,061 systems, researchers have mapped the scaling relations between X-ray luminosity, temperature, and gas mass across a vast range of cosmic time. These results show that the way these systems scale deviates significantly from simple self-similar models, though their redshift evolution remains predictable.

This suggests that complex processes like AGN feedback are actively shaping the gas in larger clusters, a finding that helps bridge the gap between observation and simulation. This focus on the energetic feedback within cosmic structures finds a much more localized echo in the study of supernova remnants.

New molecular maps of SN 1987A reveal a bipolar structure in the gas that aligns perfectly with the visible keyhole morphology and iron emissions. This evidence strongly supports the jittering-jets explosion mechanism, suggesting that a pair of energetic jets may be the primary driver for most core-collapse supernovae.

While jets shape the remnants of dead stars, radiation pressure is busy sculpting the very first galaxies. An analytical framework for the super-early universe shows that stellar clumps in high-redshift disks are naturally produced by gravitational instabilities.

These clumps are kept visible by radiation pressure, which clears away natal gas even when supernova feedback fails to do so. The interplay between gas and energy also plays out in the invisible halos surrounding massive galaxies.

By combining X-ray and Sunyaev-Zel'dovich observations of the galaxy NGC 4594, researchers found a hidden component of hot, low-density gas that carries immense thermal pressure. This extra layer is so faint in X-rays that standard analyses might be underestimating the total baryon mass by a factor of 2.5 and the thermal energy by an entire order of magnitude.

We have finally found direct evidence of how the very first galaxies polluted their surroundings with metals during the Epoch of Reionization. By looking at a red quasar, researchers spotted a massive, metal-rich gas absorber at a redshift of 7.03, which turns out to be linked to a heavily dust-obscured, star-forming galaxy called ND1.

This galaxy is sitting incredibly close to the line of sight, just 16 kiloparsecs away, and its intense starburst is driving a powerful wind of gas at 50 kilometers per second. This outflow is moving a massive amount of material, with a mass loading factor of 3, and the specific carbon-to-oxygen ratios suggest we are seeing the chemical fingerprints of primordial Population III stars.

This rapid enrichment of the early universe connects to how we understand the structure of galaxies later in cosmic time. A new model for Hoag-like ring galaxies suggests these detached outer rings are maintained by a specific radial layering mechanism.

By using a shell-deformed Kepler control model, researchers found that a compact inner boundary and a localized shell deformation can create the necessary gap and ring structure seen in objects like Hoag's Object. The physics of these environments often involves complex gas dynamics, much like the cosmic ray acceleration seen near stellar clusters.

Observations of the Berkeley 59 cluster using Fermi-LAT revealed a significant extended gamma-ray emission, which likely comes from cosmic rays accelerated by cluster winds hitting surrounding gas. We are finally getting a clearer picture of how the very first galaxies grew, and it turns out they might have been much more efficient at making stars than our best simulations predicted.

By looking at pairs of galaxies from the cosmic dawn using JWST data, researchers found that these early systems inhabit dark matter halos that aren't unusually massive, yet they possess a surprisingly high stellar-to-halo mass ratio. This suggests a star formation efficiency of about five percent, which is roughly double what the TNG100 simulations suggest.

We still need larger samples to be certain if this efficiency is what drives the abundance of luminous galaxies we see so early on. The mystery of how energy moves through atmospheres is also being tackled in the brown dwarf regime.

There has been a puzzle regarding why certain Y-dwarfs and T-dwarfs show thermal inversions, which implies some hidden heating source, perhaps auroral energy. However, new modeling shows that auroral electron beams, even when scaled to expected fluxes, cannot quite match the magnitude or the pressure levels seen in observations.

The heating happens too high up in the atmosphere, meaning we are likely missing either a much stronger high-energy electron component or other heating mechanisms entirely. Moving from the atmospheres of small objects to the massive engines of galaxies, a new Bayesian analysis of legacy Chandra X-ray data has uncovered a much more complex wind system in the AGN MCG-6-30-15.

The study identified a hot, ultra-fast outflow moving at eight percent the speed of light and found evidence of collisionally ionized absorbers, a process often overlooked. These findings, which also confirm the presence of dust within the wind, provide a vital baseline for how these powerful outflows vary over decades.

The quest for autonomous multi-messenger astronomy is moving closer to reality through the development of the RADAR framework, which aims to bridge the gap between a gravitational-wave alert and a radio telescope's response. By testing large language models against historical data, researchers found that GPT-5.5 achieved an event-level F1 score of 0.893, while Claude-Opus-4.7 reached a precision of 0.978.

These models can now help convert natural-language requests into actual scheduling blocks for the Very Large Array, potentially automating the entire follow-up chain. The search for light from these cosmic mergers is equally urgent, though often frustratingly elusive.

In a follow-up of the subsolar-mass merger candidate S251112cm, the Wide Field Survey Telescope identified four optical transients, but they evolved too slowly to be the expected kilonovae. This allowed researchers to rule out certain models, specifically disfavoring kilonovae with dynamical ejecta masses between 0.01 and 1.0 solar masses for certain viewing angles.

Even when we can't find the light from mergers, we are finding new ways to see planets using the tools meant for something else. By looking at asymmetries in the point spread function of the HD 101452 star, researchers used short, noncoronagraphic JWST MIRI imaging to detect a stellar companion.

This proves that very brief exposures of just 17 to 42 seconds could potentially detect sub-Jupiter-mass planets around nearby stars if we use closely matched reference stars for subtraction. Precision in our observations also relies on understanding the stars themselves, particularly the subtle broadening of their spectral lines.

Using the Keck Planet Finder, a new study found that stellar oscillation frequency, or nu max, is a much better predictor of macroturbulence in dwarf stars than temperature, reducing the error in these predictions to about 90 meters per second. On a much larger scale, we are still refining how we model the very first stars.

New simulations from the Popsicle project show that atomic hydrogen shielding is vital for the growth of Population III protostars, as it prevents Lyman-Werner radiation from shutting down accretion. This shielding keeps gas temperatures low and allows the stars to reach much larger final masses than previously thought.

The physics of these early environments is further complicated by the magnetic fields within accretion disks. New radiation magneto-hydrodynamic simulations show that if we don't resolve the thermal scale height in our models, we might incorrectly predict that strongly magnetized disks undergo runaway cooling.

Instead, properly resolved simulations show these disks settle into a steady state of turbulent heating and dynamo cycles. This turbulence is part of a larger, non-linear evolution of the universe.

New calculations of the first non-linear corrections to cosmic density and velocity variances show that these effects, known as previrialization, occur before structures even fully form. The study found that the coefficient for this leading-order correction is remarkably stable, staying within a narrow range between 1.817 and 1.843 regardless of the linear power spectrum used.

Finally, we are seeing how quantum effects might nudge our fundamental cosmological models. By adding temperature-dependent corrections to the cosmological constant, researchers found that the resulting parameters, Omega L2 and Omega L3, actually improve the fit to the Planck 2018 cosmic microwave background data.

This suggests that the fine structure of the power spectrum might be a window into finite-temperature quantum gravity.

Today's papers

The papers

Important terms

Wolf-Rayet phase
A stage in a massive star's life where it becomes extremely hot, blue, and bright in ultraviolet light. Identifying this phase helps astronomers find the specific types of stars that eventually collapse into black holes.
Higgs vacuum expectation value
A fundamental value related to the Higgs field that influences particle masses. Small changes to this value during the early universe could explain why current observations of helium and lithium don't perfectly match theoretical models.
Gravitational waves
Ripples in spacetime caused by violent cosmic events. Unlike light, which can be blocked by dense gas, these waves can pass through star envelopes, allowing scientists to peek inside the engines driving gamma-ray bursts.
Population III stars
The very first generation of stars formed in the early universe. They were made of primordial material and their chemical signatures, like carbon-to-oxygen ratios, help researchers understand how the first galaxies polluted their surroundings with metals.
Sunyaev-Zel'dovich effect
A phenomenon used to observe hot gas in galaxy clusters by looking at how it affects cosmic microwave background radiation. It helps scientists map hidden, low-density gas that might be missing from standard X-ray observations.