| 2026 |
|---|
 | Machine-learning classification of the kinematic properties of elliptical galaxies. Using integral-field spectroscopic data from the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) galaxy survey, this study analyzes how velocity dispersion changes from the centers to the outer regions of elliptical galaxies. A combination of unsupervised and supervised machine-learning methods was used to build an automated classifier that separates the velocity-dispersion profiles into four characteristic types: Flat, Decline, Ascend, and Irregular. The trained classifier achieved an overall accuracy of 88% on the test sample. When applied to 2,624 MaNGA DR17 elliptical galaxies, approximately 68.87% were found to exhibit Flat velocity-dispersion profiles. This approach provides an efficient way to investigate the kinematic structures of large galaxy samples and offers a new statistical tool for future studies of galaxy formation, evolution, and mass distribution.
Dr. Yi Duann, Dr. Yong Tian, and Prof. Chung-Ming Ko have published a study applying machine-learning techniques to investigate the internal motions of elliptical galaxies observed by the MaNGA survey. The research analyzes velocity-dispersion profiles of elliptical galaxies, which describe how the spread of velocities changes from the centers to the outer regions of galaxies. By combining unsupervised K-means clustering with a supervised TreeBagger classifier, the team identified four characteristic profile types: Flat, Decline, Ascend, and Irregular. The supervised model achieved an overall classification accuracy of 88% on the test set.
The trained model was then applied to 2,624 elliptical galaxies in the MaNGA DR17 sample. Approximately 67.9% of these galaxies were found to exhibit Flat velocity-dispersion profiles, while about 24.5% showed declining profiles. The high fraction of Flat systems is particularly interesting because similar profiles have been reported among brightest cluster galaxies, providing a promising direction for future studies of galaxy formation, evolution, and mass distribution. This work demonstrates how machine learning can efficiently transform large astronomical surveys into statistically meaningful classifications of galaxy kinematics. The study was published in RAS Techniques and Instruments (Duann, Tian & Ko 2023, RASTAI, 2, 649–656).
https://doi.org/10.1093/rasti/rzad044
Show large images and more information |
 | Possible near-ultraviolet early ingress of WASP-189b and its magnetohydrodynamic interpretation. Top: CUTE near-ultraviolet observations of three WASP-189b transits. The third visit (V3) shows a tentative transit-phase offset of approximately 31.5 minutes, consistent with additional absorbing material located ahead of the planet. Bottom: Two-dimensional magnetohydrodynamic simulations illustrating how the interaction between the stellar wind and the planetary magnetosphere changes under different fast-mode Mach numbers. Distinct bow shocks form only in sufficiently fast and dense stellar-wind conditions, whereas slower winds can instead produce a compressed plasma pileup ahead of the planet. The results suggest that such cooled, dense material may provide a more favourable condition for producing detectable near-ultraviolet early-ingress absorption.
Dr. Yi Duann and her team have published a study investigating a possible near-ultraviolet early ingress of the ultra-hot Jupiter WASP-189b using observations from the 6U Colorado Ultraviolet Transit Experiment (CUTE) CubeSat. Among three observed transits, the third visit showed a tentative phase offset of about 31.5 minutes, corresponding to absorbing material extending several planetary radii ahead of the planet. Such early-ingress signatures have often been proposed as possible evidence of interactions between stellar winds and planetary magnetospheres.
To test this scenario, the team performed magnetohydrodynamic simulations across different stellar-wind conditions. The simulations show that classical bow shocks can form when both the wind speed and plasma density are sufficiently high; however, the shocked gas can then remain too hot to efficiently absorb near-ultraviolet light. In contrast, a transition toward slower stellar winds can preserve compressed plasma while allowing it to cool, creating a dense magnetic pileup that may be more readily detectable in transit observations. The study therefore highlights how time-variable stellar winds can influence the observable signatures of exoplanet magnetospheres and provides a new framework for probing star–planet interactions around ultra-hot Jupiters. The work was published in Astronomy & Astrophysics (Duann et al. 2025, A&A, 703, A24).
https://doi.org/10.1051/0004-6361/202556404
Show large images and more information |
 | Connecting observed exoplanet populations to planet-formation pathways with machine learning. Top: Analysis framework combining planets from the NASA Exoplanet Archive with unsupervised Gaussian mixture model (GMM) clustering and pebble-accretion synthetic populations. Observed planets are classified using their physical and dynamical properties and subsequently mapped into the same three-dimensional parameter space as simulated planets. Bottom: Distributions of three formation-related quantities predicted for the mapped populations: the gas availability at the onset of formation (G1), planetary gas mass fraction (fqas), and ice–rock mass ratio. The four populations—hot giants (HG), warm-Jupiter-dominated systems (WJD), lower-mass giants (LMG), and very-massive gas giants (VMGG)—show systematically different formation histories, with the very-massive gas giants preferentially associated with earlier formation in gas-rich disks.
Dr. Yi Duann, in collaboration with researchers at the Center for Star and Planet Formation, Globe Institute, University of Copenhagen, has published a study using machine learning to connect observed close-in exoplanets with theoretical planet-formation models. Rather than assigning planets to predefined categories, the team used a two-stage Gaussian mixture model to identify natural populations from their orbital and dynamical properties. These observational groups were then mapped onto synthetic planets produced by pebble-accretion simulations, allowing otherwise inaccessible properties, such as formation timing, gas-envelope growth, and solid composition, to be statistically inferred at the population level.
The statistical analysis shows that these populations are not simply visual groupings. Differences in formation-related parameters are highly significant, with the strongest inter-population effects found for the gas mass fraction and planet-to-star mass ratio. A complementary multinomial diagnostic separated the mapped populations with an overall accuracy of 98.2%. The very-massive gas giants show the highest median gas availability at formation and a strong correlation between early formation and gas accretion, while lower-mass giants exhibit substantially broader and more diverse formation histories. These results provide a data-driven bridge between present-day exoplanet observations and the physical processes that shaped planetary systems. The study was published in Astronomy & Astrophysics (Duann et al. 2026, A&A, 711, A238).
https://doi.org/10.1051/0004-6361/202659961
Show large images and more information |
 | Figure: Superorbital phase residual distributions for soft (red) and hard (black) X-rays, with the smooth evolution trend removed. Data are divided into two epochs: MJD 53500 - 57000 (top) and MJD 57000 - 60300 (bottom). Mean values and 1σ uncertainties are indicated by dashed lines and horizontal bars. A significant phase shift of 0.044 ± 0.010 cycles between soft and hard X-ray bands is evident in the bottom plot
Professor Yi Chou has published a latest study that explores the evolution of "superorbital" light variations in the high-mass X-ray binary system LMC X-4 by analyzing 33 years of data from multiple space telescopes. LMC X-4, located in the Large Magellanic Cloud, is an accreting binary system consisting of a massive star and a compact neutron star. These two stars orbit each other every 1.4 days, while the system also exhibits a "superorbital period" of approximately 30.5 days. Astronomers point out that the core mechanism of this period lies in the existence of a prominently warped accretion disk around the neutron star. This warped disk undergoes a slow "precession," periodically obscuring the X-rays emitted by the neutron star during its rotation. Professor Chou's research confirms that despite the complexity of the disk structure and superorbital phase variations, the precession period of LMC X-4 is, on average, remarkably stable. Over the past thirty-plus years, the variation in the superorbital period has been only 0.55%, making it the most stable system of its kind currently known.
Intriguingly, the study found that after late 2014 (MJD ~ 57000), observation data revealed a distinct "phase shift" between the soft and hard X-ray bands. This phenomenon indicates that the geometric structure of the warped accretion disk underwent further deformation, transitioning from a relatively symmetric warped state to an asymmetric structure. This change in disk geometry caused a lag in the timing of when rays of different energies were obscured as they passed through the warped edges. This discovery coincided with a decline in the system's hard X-ray intensity, suggesting that the warped disk is now obscuring the central object in a new and more complex manner. These findings provide astronomers with an ideal laboratory for studying accretion disk behavior in extreme physical environments. By understanding how these warped disks precess, evolve, and shift over decades, researchers can further understand how matter flows under extreme gravity and radiation pressure. This marks an important step forward in the understanding of accretion disk dynamics within high-energy astrophysics. This research paper has been published in The Astrophysical Journal (
Chou 2026, ApJ, 1000, 23).
Show large images and more information |
