J/ApJ/992/93 Classifying cool dwarfs (Zhou+, 2025)
Classifying Cool Dwarfs: Comprehensive Spectral Typing of Field and Peculiar
Dwarfs Using Machine Learning.
Zhou T., Theissen C.A., Feeser S.J., Best W.M.J., Burgasser A.J.,
Cruz K.L., Zhao L.
<Astrophys. J. 992, 93 (2025)>
=2025ApJ...992...93Z 2025ApJ...992...93Z (SIMBAD/NED BibCode)
ADC_Keywords: Stars, brown dwarf ; Stars, M-type ; Stars, L-type ;
Stars, T-type ; Spectral types ; Photometry, infrared ;
Parallaxes, trigonometric
Keywords: Brown dwarfs - T dwarfs - L dwarfs - M dwarf stars -
M subdwarf stars - Stellar classification - Random Forests -
Support vector machine
Abstract:
Low-mass stars and brown dwarfs-spectral types (SpTs) M0 and
later-play a significant role in studying stellar and substellar
processes and demographics, reaching down to planetary-mass objects.
Currently, the classification of these sources remains heavily reliant
on visual inspection of spectral features, equivalent width
measurements, or narrow/wideband spectral indices. Recent advances in
machine learning (ML) methods offer automated approaches for spectral
typing, which are becoming increasingly important as large
spectroscopic surveys such as Gaia, SDSS, and SPHEREx generate data
sets containing millions of spectra. We investigate the application of
ML in spectral type classification on low-resolution (R∼120)
near-infrared spectra of M0-T9 dwarfs obtained with the SpeX
instrument on the NASA Infrared Telescope Facility. We specifically
aim to classify the gravity- and metallicity-dependent subclasses for
late-type dwarfs. We used binned fluxes as input features and compared
the efficacy of spectral type estimators built using Random Forest
(RF), Support Vector Machine, and K-Nearest Neighbor (KNN) models. We
tested the influence of different normalizations and analyzed the
relative importance of different spectral regions for surface gravity
and metallicity subclass classification. Our best-performing model
(using KNN) classifies 95.5%±0.6% of sources to within ±1 SpT, and
assigns surface gravity and metallicity subclasses with 89.5%±0.9%
accuracy. We test the dependence of signal-to-noise ratio on
classification accuracy and find sources with SNR ∼60 have ∼95%
accuracy. We also find that zy band plays the most prominent role in
the RF model, with FeH and TiO having the highest feature importance.
Description:
Table 2 provides the machine-readable version of the SpeX Prism
Library data set used in the paper as the validation and testing
sample for machine-learning spectral typing of cool dwarfs. It lists
1527 unique sources observed with the SpeX spectrograph on the NASA
IRTF in prism mode, covering spectral types M0-T9, including field
dwarfs, low-gravity objects, and metal-poor subdwarfs. Missing values
are left blank in the table.
Table 7 provide the classification results.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table2.dat 234 1527 SpeX Prism Library validation/testing sample
table7.dat 79 1300 Classification Results
--------------------------------------------------------------------------------
See also:
II/246 : 2MASS All-Sky Catalog of Point Sources (Cutri+ 2003)
Byte-by-byte Description of file: table2.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 32 A32 --- Name Source identifier
34- 43 A10 --- ShortName Short name (JMHHMM+DDMM)
45- 57 F13.9 deg RAdeg Right Ascension (J2000)
59- 71 F13.9 deg DEdeg Declination (J2000)
73- 91 A19 --- AName Designation (JHHMMSSss+DDMMSSs)
93- 99 A7 --- SpType Adopted NIR spectral type
101-108 F8.2 --- Chi2 Chi-square of spectral standard fit
110-111 I2 --- DwarfType Dwarf type flag
113-116 I4 --- SNR SpeX prism S/N
118-124 A7 --- SpTypeB06 Spectral type from Burgasser et al.
(2006ApJ...637.1067B 2006ApJ...637.1067B, Cat. J/ApJ/637/1067)
126-132 A7 --- SpTypeC18 Spectral type from Cruz et al.
(2018AJ....155...34C 2018AJ....155...34C, Cat. J/AJ/155/34)
134-140 A7 --- SpTypeK10 Spectral type from Kirkpatrick et al.
(2010ApJS..190..100K 2010ApJS..190..100K. Cat. J/ApJS/190/100)
142-151 A10 --- SpTypeSimbad Spectral type from SIMBAD
153-171 A19 --- r_SpType Bibcode for SpType
173-175 I3 --- Refs Reference code
177-183 F7.3 mas plx ? Trigonometric parallax
185-190 F6.3 mas e_plx ? Parallax uncertainty
192-210 A19 --- r_plx Bibcode for parallax
212-216 F5.2 mag Jmag ? 2MASS J magnitude
218-222 F5.2 mag Hmag ? 2MASS H magnitude
224-228 F5.2 mag Kmag ? 2MASS Ks magnitude
230-234 F5.2 mag JMAG ? Absolute J magnitude from Jmag and plx
--------------------------------------------------------------------------------
Byte-by-byte Description of file: table7.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 43 A43 --- Name Source identifier
45- 51 A7 --- SpTypeRFS Spectral type from Random Forest model
53- 59 A7 --- SpTypeSVM Spectral type from Support Vector Machine
61- 67 A7 --- SpTypeKNN Spectral type from K-Nearest Neighbor model
69- 71 I3 --- SNR SpeX signal-to-noise ratio
73- 79 A7 --- SpType Adopted spectral type
--------------------------------------------------------------------------------
Acknowledgements:
From Tianxing Zhou, t4zhou(at)ucsd.edu
This work makes use of data products from the SpeX Prism Library and
the SpeX Prism Library Analysis Toolkit (SPLAT), based on observations
obtained with the NASA Infrared Telescope Facility (IRTF). We
acknowledge the use of the 2MASS survey, SIMBAD, and the VizieR
catalogue access tool at CDS. We also thank the original discovery and
classification papers referenced through the r_SpType and r_plx
columns.
References:
Burgasser & Splat Development Team 2017, ASInC, 14, 7
(SpeX Prism Library overview)
Rayner et al., 2003, PASP, 115, 362 (SpeX instrument)
(End) T. Zhou [Univ. California San Diego], P. Vannier [CDS] 26-Nov-2025