J/ApJ/998/307 High-confidence Galactic RSGs (Zhang+, 2026)
An Enhanced Sample of Galactic Red Supergiants Reveals Spiral Structures.
Zhang Z., Jiang B., Ren Y.
<Astrophys. J. 998, 307 (2026)>
=2026ApJ...998..307Z 2026ApJ...998..307Z (SIMBAD/NED BibCode)
ADC_Keywords: Milky Way ; Stars, supergiant ; Optical
Keywords: Red supergiant stars - Galaxy structure - Spiral arms -
Neural networks - Stellar classification
Abstract:
Red supergiants (RSGs), representing a kind of massive young stellar
population, have rarely been used to probe the structure of the Milky
Way, mainly due to the long-standing scarcity of Galactic RSG samples.
The Gaia BP/RP spectra (hereafter XP), which cover a broad wavelength
range, provide a powerful tool for identifying RSGs. In this work, we
develop a feedforward neural network classifier that assigns to each
XP spectrum a probability of being an RSG, denoted as P(RSG). We
perform 10 independent runs with randomly divided training and
validation sets, and apply each run to all XP spectra of stars with
G<12mag. By selecting sources with P(RSG)≥0.9, 10 high-confidence
candidate samples are obtained. A star is considered a true Galactic
RSG only if it appears in at least eight of these samples, yielding a
final catalog of 2436 objects. These RSGs show a clear spatial
correlation with OB stars and trace the Galactic spiral arms well,
confirming the reliability of our classification, and highlighting
their potential to serve as powerful tracers of the Milky Way's
structure.
Description:
Gaia BP/RP (XP) spectra are used to identify Galactic red supergiants
(RSGs) with a feedforward neural network. Ten independent training
runs are applied to stars with G<12mag. High-confidence candidates
have an RSG-class probability of at least 0.9 in an individual run.
Sources selected in at least eight of the ten runs, after the
Galactic-membership cuts, form a catalog of 2436 objects. The spatial
distribution of the selected sources is compared with that of OB stars
and with models of the Galactic spiral arms.
This data set contains the the high-confidence Galactic RSG catalog.
The classification uses Gaia DR3 XP spectra and ten independent
neural-network runs. Additional cuts require
classprobdsccombmod_star>0.99 and remove sources in the specified
LMC/SMC sky regions with radial velocity greater than 100 km/s, as
described in Section 4.1 of the paper. Each row gives the Gaia DR3
identifier, ICRS position at epoch J2016.0, mean RSG-class
probability, adopted distance, and the published Galactocentric
Cartesian coordinates where available. All published numerical values
and the original row order are retained. The P(RSG) field is stored as
a dimensionless fraction, correcting the percent unit in the input MRT
description only.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table3.dat 103 2436 High-confidence Galactic RSG catalog
--------------------------------------------------------------------------------
See also:
I/355 : Gaia DR3 Part 1. Main source (Gaia Collaboration, 2022)
Byte-by-byte Description of file: table3.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 19 I19 --- GaiaDR3 Gaia DR3 source identifier
21- 40 F20.16 deg RAdeg Right ascension (ICRS) at Ep=2016.0
42- 61 F20.16 deg DEdeg Declination (ICRS) at Ep=2016.0
63- 68 F6.4 --- PRSG [0/1] Mean RSG-class probability (1)
70- 76 F7.4 kpc Dist Adopted heliocentric distance (2)
78- 86 F9.5 kpc X ? Galactocentric X-coordinate (3)
88- 94 F7.4 kpc Y ? Galactocentric Y-coordinate (3)
96-103 F8.5 kpc Z ? Galactocentric Z-coordinate (3)
--------------------------------------------------------------------------------
Note (1): PRSG is P(RSG), the mean probability across the ten independent
runs. Values are fractions, not percentages; multiply by 100 for the
percentage presentation used in the printed table.
Note (2): Distances primarily use the photogeometric estimates of
Bailer-Jones et al. (2021AJ....161..147B 2021AJ....161..147B; CDS catalog I/352), converted
from pc to kpc. For five sources without such estimates, the paper adopts
inverse-parallax distances.
Note (3): The Galactic centre is the origin and the Sun is at
(X,Y,Z)=(0,8.5,0)kpc. With heliocentric distance d in kpc and
Galactic longitude l and latitude b, the convention is
X=d*cos(b)*sin(l), Y=8.5-d*cos(b)*cos(l), Z=d*sin(b).
--------------------------------------------------------------------------------
Acknowledgements:
From Zehao Zhang, zhzhang(at)mail.bnu.edu.cn
Prepared for Zhang, Jiang & Ren
This work uses data from the ESA Gaia mission, processed by the Gaia
Data Processing and Analysis Consortium (DPAC).
License: CC-BY-4.0
(End) Patricia Vannier [CDS] 08-Oct-2026