J/MNRAS/538/605 Gaia RVS stellar labels using The Cannon (Das+, 2025)
The GALAH Survey: Stellar parameters and abundances for 800000 Gaia RVS spectra
using GALAH DR4 and The Cannon.
Das P.B., Zucker D.B., De Silva G.M., Borsato N.W., Mura-Guzman A.,
Buder S., Ness M., Nordlander T., Casey A.R., Martell S.L.,
Bland-Hawthorn J., de Grijs R., Freeman K.C., Kos J., Stello D., Lewis G.F.,
Hayden M.R., Sharma S.
<Mon. Not. R. Astron. Soc. 538, 605 (2025)>
=2025MNRAS.538..605D 2025MNRAS.538..605D (SIMBAD/NED BibCode)
ADC_Keywords: Surveys ; Milky Way ; Spectroscopy ; Abundances
Keywords: methods: data analysis - methods: statistical -
techniques: spectroscopic - surveys - stars: abundances -
stars: fundamental parameters
Abstract:
Analysing stellar parameters and abundances from nearly one million
Gaia DR3 Radial Velocity Spectrometer (RVS) spectra poses challenges
due to the limited spectral coverage (restricted to the infrared CaII
triplet) and variable signal-to-noise ratios of the data. To address
this, we use The Cannon, a data-driven method, to transfer stellar
parameters and abundances from the GALAH Data Release 4 (DR4; R∼28000)
catalogue to the lower resolution Gaia DR3 RVS spectra (R∼11500). Our
model, trained on 14484 common targets, predicts parameters such as
Teff, logg, and [Fe/H], along with several other elements across
approximately 800000 Gaia RVS spectra. We utilise stars from open and
globular clusters present in the Gaia RVS catalogue to validate our
predicted mean [Fe/H] with high precision (∼0.02-0.10dex).
Additionally, we recover the bimodal distribution of [Ti/Fe] versus
[Fe/H], reflecting the high and low alpha-components of Milky Way disk
stars, demonstrating The Cannon's capability for accurate stellar
abundance determination from medium-resolution Gaia RVS spectra. The
methodologies and resultant catalogue presented in this work highlight
the remarkable potential of the RVS dataset, which by the end of the
Gaia mission will comprise spectra of over 200 million stars.
Description:
We have demonstrated the capability of The Cannon; a data-driven
machine learning approach to accurately determine stellar parameters
(Teff and logg) and elemental abundances across a multi-dimensional
label space, including [Fe/H], [Ca/Fe], [Si/Fe], [Ni/Fe], and [Ti/Fe],
for 796633 Gaia DR3 RVS stars using stellar labels from GALAH DR4.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
rvs_gal.dat 431 796633 The Cannon predictions for stellar labels
--------------------------------------------------------------------------------
See also:
I/355 : Gaia DR3 Part 1. Main source (Gaia Collaboration, 2022)
Byte-by-byte Description of file: rvs_gal.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 19 I19 --- GaiaDR3 Gaia DR3 source_id
21- 32 F12.7 --- GaiaRVSSNR Signal-to-Noise of RVS spectra
34 A1 --- flagCannon [0/1] Cannon flag (1)
36- 53 F18.13 K Teff Effective temperature
55- 75 F21.16 K e_Teff rms uncertainty on Teff
77- 95 F19.16 [cm/s2] logg Surface gravity
97-119 E23.15 [cm/s2] e_logg rms uncertainty on logg
121-144 E24.16 km/s vsini Broadening velocity
146-164 F19.16 km/s e_vsini rms uncertainty on vsini
166-189 E24.16 [-] [Fe/H] Fe abundance
191-213 E23.15 [-] e_[Fe/H] rms uncertainty on [Fe/H]
215-238 E24.16 [-] [Ca/Fe] Ca abundance
240-262 E23.15 [-] e_[Ca/Fe] rms uncertainty on [Ca/Fe]
264-287 E24.16 [-] [Ti/Fe] Ti abundance
289-311 E23.15 [-] e_[Ti/Fe] rms uncertainty on [Ti/Fe]
313-336 E24.16 [-] [Si/Fe] Si abundance
338-360 E23.15 [-] e_[Si/Fe] rms uncertainty on [Si/Fe]
362-385 E24.16 [-] [Ni/Fe] Ni abundance
387-410 E24.16 [-] e_[Ni/Fe] rms uncertainty on [Ni/Fe]
412-431 F20.16 --- chisq Chi2 value of stellar label fitting
--------------------------------------------------------------------------------
Note (1): Cannon Flag as follows:
0 = Label estimates within the acceptable label space of the training sample
1 = Label estimates outside the label space of the training sample
--------------------------------------------------------------------------------
Acknowledgements:
Pradosh Barun Das, pbdrohan(at)gmail.com
(End) Patricia Vannier [CDS] 05-Feb-2025