TY - JOUR
T1 - The RAdial Velocity Experiment (RAVE)
T2 - Parameterisation of RAVE spectra based on convolutional neural networks
AU - Guiglion, G.
AU - Matijevič, G.
AU - Queiroz, A. B.A.
AU - Valentini, M.
AU - Steinmetz, M.
AU - Chiappini, C.
AU - Grebel, E. K.
AU - McMillan, P. J.
AU - Kordopatis, G.
AU - Kunder, A.
AU - Zwitter, T.
AU - Khalatyan, A.
AU - Anders, F.
AU - Enke, H.
AU - Minchev, I.
AU - Monari, G.
AU - Wyse, R. F.G.
AU - Bienaymé, O.
AU - Bland-Hawthorn, J.
AU - Gibson, B. K.
AU - Navarro, J. F.
AU - Parker, Q.
AU - Reid, W.
AU - Seabroke, G. M.
AU - Siebert, A.
PY - 2020
Y1 - 2020
N2 - Context Data-driven methods play an increasingly important role in the field of astrophysics In the context of large spectroscopic surveys of stars, data-driven methods are key in deducing physical parameters for millions of spectra in a short time. Convolutional neural networks (CNNs) enable us to connect observables (e.g. spectra, stellar magnitudes) to physical properties (atmospheric parameters, chemical abundances, or labels in general). Aims. We test whether it is possible to transfer the labels derived from a high-resolution stellar survey to intermediate-resolution spectra of another survey by using a CNN. Methods. We trained a CNN, adopting stellar atmospheric parameters and chemical abundances from APOGEE DR16 (resolution Ra22 500) data as training set labels. As input, we used parts of the intermediate-resolution RAVE DR6 spectra (R ∼ 7500) overlapping with the APOGEE DR16 data as well as broad-band ALLWISE and 2MASS photometry, together with Gaia DR2 photometry and parallaxes. Results. We derived precise atmospheric parameters Teff, log(g), and [M/H], along with the chemical abundances of [Fe/H], [α/M], [Mg/Fe], [Si/Fe], [Al/Fe], and [Ni/Fe] for 420 165 RAVE spectra. The precision typically amounts to 60 K in Teff, 0.06 in log(g) and 0.02-0.04 dex for individual chemical abundances. Incorporating photometry and astrometry as additional constraints substantially improves the results in terms of the accuracy and precision of the derived labels, as long as we operate in those parts of the parameter space that are well-covered by the training sample. Scientific validation confirms the robustness of the CNN results. We provide a catalogue of CNN-Trained atmospheric parameters and abundances along with their uncertainties for 420 165 stars in the RAVE survey. Conclusions. CNN-based methods provide a powerful way to combine spectroscopic, photometric, and astrometric data without the need to apply any priors in the form of stellar evolutionary models. The developed procedure can extend the scientific output of RAVE spectra beyond DR6 to ongoing and planned surveys such as Gaia RVS, 4MOST, and WEAVE. We call on the community to place a particular collective emphasis and on efforts to create unbiased training samples for such future spectroscopic surveys.
AB - Context Data-driven methods play an increasingly important role in the field of astrophysics In the context of large spectroscopic surveys of stars, data-driven methods are key in deducing physical parameters for millions of spectra in a short time. Convolutional neural networks (CNNs) enable us to connect observables (e.g. spectra, stellar magnitudes) to physical properties (atmospheric parameters, chemical abundances, or labels in general). Aims. We test whether it is possible to transfer the labels derived from a high-resolution stellar survey to intermediate-resolution spectra of another survey by using a CNN. Methods. We trained a CNN, adopting stellar atmospheric parameters and chemical abundances from APOGEE DR16 (resolution Ra22 500) data as training set labels. As input, we used parts of the intermediate-resolution RAVE DR6 spectra (R ∼ 7500) overlapping with the APOGEE DR16 data as well as broad-band ALLWISE and 2MASS photometry, together with Gaia DR2 photometry and parallaxes. Results. We derived precise atmospheric parameters Teff, log(g), and [M/H], along with the chemical abundances of [Fe/H], [α/M], [Mg/Fe], [Si/Fe], [Al/Fe], and [Ni/Fe] for 420 165 RAVE spectra. The precision typically amounts to 60 K in Teff, 0.06 in log(g) and 0.02-0.04 dex for individual chemical abundances. Incorporating photometry and astrometry as additional constraints substantially improves the results in terms of the accuracy and precision of the derived labels, as long as we operate in those parts of the parameter space that are well-covered by the training sample. Scientific validation confirms the robustness of the CNN results. We provide a catalogue of CNN-Trained atmospheric parameters and abundances along with their uncertainties for 420 165 stars in the RAVE survey. Conclusions. CNN-based methods provide a powerful way to combine spectroscopic, photometric, and astrometric data without the need to apply any priors in the form of stellar evolutionary models. The developed procedure can extend the scientific output of RAVE spectra beyond DR6 to ongoing and planned surveys such as Gaia RVS, 4MOST, and WEAVE. We call on the community to place a particular collective emphasis and on efforts to create unbiased training samples for such future spectroscopic surveys.
KW - Galaxy: Abundances
KW - Galaxy: stellar content
KW - Methods: data analysis
KW - Stars: Abundances
KW - Techniques: spectroscopic
U2 - 10.1051/0004-6361/202038271
DO - 10.1051/0004-6361/202038271
M3 - Article
AN - SCOPUS:85098008804
SN - 0004-6361
VL - 644
JO - Astronomy and Astrophysics
JF - Astronomy and Astrophysics
M1 - 202038271
ER -