Gaia Data Release 3_All-sky classification of 12.4 million variable sources into 25 classes.pdf

OpenAtomOpenTenBase

105页

2次

2024-10-13

免费下载

A&A 674, A14 (2023)

https://doi.org/10.1051/0004-6361/202245591

 The Authors 2023

Astronomy

Astrophysics

Gaia Data Release 3 Special issue

Gaia Data Release 3

All-sky classiﬁcation of 12.4 million variable sources into 25 classes

Lorenzo Rimoldini

1,?

, Berry Holl

1,2

, Panagiotis Gavras

, Marc Audard

1,2

, Joris De Ridder

Nami Mowlavi

1,2

, Krzysztof Nienartowicz

, Grégory Jevardat de Fombelle

, Isabelle Lecoeur-Taïbi

Lea Karbevska

1,6

, Dafydd W. Evans

, Péter Ábrahám

8,9

, Maria I. Carnerero

, Gisella Clementini

Elisa Distefano

, Alessia Garofalo

, Pedro García-Lario

, Roy Gomel

, Sergei A. Klioner

Katarzyna Kruszy

nska

, Alessandro C. Lanzafame

12,17

, Thomas Lebzelter

, Gábor Marton

Tsevi Mazeh

, Roberto Molinaro

, Aviad Panahi

, Claudia M. Raiteri

, Vincenzo Ripepi

László Szabados

, David Teyssier

, Michele Trabucchi

, Łukasz Wyrzykowski

Shay Zucker

, and Laurent Eyer

(Aﬃliations can be found after the references)

Received 30 November 2022 / Accepted 17 December 2022

ABSTRACT

Context. Gaia DR3 contains 1.8 billion sources with G -band photometry, 1.5 billion of which with G

and G

photometry, complemented by

positions on the sky, parallax, and proper motion. The median number of ﬁeld-of-view transits in the three photometric bands is between 40 and

44 measurements per source and covers 34 months of data collection.

Aims. We pursue a classiﬁcation of Galactic and extra-galactic objects that are detected as variable by Gaia across the whole sky.

Methods. Supervised machine learning (eXtreme Gradient Boosting and Random Forest) was employed to generate multi-class, binary, and meta-

classiﬁers that classiﬁed variable objects with photometric time series in the G, G

, and G

bands.

Results. Classiﬁcation results comprise 12.4 million sources (selected from a much larger set of potential variable objects) and include about

9 million variable stars classiﬁed into 22 variability types in the Milky Way and nearby galaxies such as the Magellanic Clouds and Andromeda,

plus thousands of supernova explosions in distant galaxies, 1 million active galactic nuclei, and almost 2.5 million galaxies. The identiﬁcation of

galaxies was made possible by the artiﬁcial variability of extended objects as detected by Gaia, so they were published in the galaxy_candidates

table of the Gaia DR3 archive, separate from the classiﬁcations of genuine variability (in the vari_classifier_result table). The latter contains

24 variability classes or class groups of periodic and non-periodic variables (pulsating, eclipsing, rotating, eruptive, cataclysmic, stochastic, and

microlensing), with amplitudes from a few milli-magnitudes to several magnitudes.

Key words. catalogs – galaxies: general – methods: data analysis – quasars: general – stars: variables: general

1. Introduction

Time-dependent brightness variations of celestial objects may be

caused by diﬀerent phenomena: intrinsic physical changes, such

as pulsations, eruptions, and cataclysmic outbursts, or extrin-

sic reasons that depend on the direction of observation, such

as eclipsing binaries, stars rotating with spots or with ellip-

soidal shapes, and microlensing events, as shown in Fig. 1 of

Gaia Collaboration (2019). The detection of variability requires

multi-epoch observations and, depending on the signal sampling,

a certain set of classes can be identiﬁed. Gaia’s sparse sampling

allows for the detection of periodic signals ranging from min-

utes to years and for medium to long-term non-periodic variabil-

ity. The chance of detection of up to approximately six-week

long transient phenomena, for example, crucially depends on

the sampling at a given location in the sky (see Appendix A in

Eyer et al. 2017), which follows from the scanning law proper-

ties (Gaia Collaboration 2016b). Although the scanning law of

Gaia was designed for astrometric goals, it allows for the iden-

tiﬁcation of a broad variety of variability types, with diﬀerent

possible levels of completeness (Eyer et al. 2023).

Corresponding author: L. Rimoldini,

e-mail: Lorenzo.Rimoldini@unige.ch

As the time span of Gaia data collection progressively

increased from Gaia data release 1 (DR1) to DR2 and DR3 (14,

22, and 34 months, respectively; Gaia Collaboration 2016a, 2018,

2023c), the classiﬁed variability types increased from Cepheids

and RR Lyrae stars in a limited region of the sky (in DR1;

Eyer et al. 2017), to an all-sky

classiﬁcation of the DR1 classes

plus long-period variables and δ Scuti or SX Phoenicis stars (in

DR2; Rimoldini et al. 2019), and 20 further variability classes

in DR3 (presented in this article and listed in Sect. 3.1.1). For

brevity, we refer to Table 1 for selected publications related to

these classes, with representatives identiﬁed in various surveys.

Machine learning is a practical tool to automate classiﬁca-

tion tasks that involve multiple known classes and a possibly high

number of attributes to identify such classes and distinguish them

from others (e.g. see Debosscher et al. 2007; Sarro et al. 2009;

Blomme et al. 2010; Richards et al. 2011; Dubath et al. 2011).

Herein, we present how a supervised classiﬁcation was applied to

Gaia DR3 data to classify variable sources into two dozen classes

(plus galaxies). In particular, we describe the details concerning

the construction of the training set and of the classiﬁers, the veriﬁ-

cation of the results, and the generation of an overall classiﬁcation

Due to the scanning law coverage, only from Gaia DR2 onwards were

suﬃcient epochs available at all locations on the sky, for example, see

Fig. 1 in Holl et al. (2018).

Open Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0),

which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This article is published in open access under the Subscribe to Open model. Subscribe to A&A to support open access publication.

A14, page 1 of 105

Rimoldini, L., et al.: A&A 674, A14 (2023)

score. Selection procedures, parameter distributions, and assess-

ments of candidates are presented for each class.

Some of the classiﬁcation results are further processed

by speciﬁc object studies (SOSs) dedicated to single classes,

typically describing a subset of the most reliable candidates in

detail. Such single-class processing modules are available in

DR3 for active galactic nuclei (AGNs Carnerero et al. 2023),

Cepheids (Ripepi et al. 2023), compact companions (Gomel et al.

2023), eclipsing binaries (Mowlavi et al. 2023), long-period

variables (Lebzelter et al. 2023), main-sequence oscillators

(Gaia Collaboration 2023b), planetary transits (Panahi et al.

2022), and RR Lyrae stars (Clementini et al. 2023). Other SOS

modules, such as microlensing events (Wyrzykowski et al. 2023),

short-timescale variables (see Sect. 10.12 of the Gaia DR3

documentation; Rimoldini et al. 2022), and solar-like rotation

modulation stars (Distefano et al. 2023), were executed indepen-

dently of the classiﬁcation results, as they relied on their own

candidate selection. A summary of the variability results from all

modules is presented in Eyer et al. (2023).

This article is organised as follows. The classiﬁcation input

data are outlined in Sect. 2; the preparation, application, and veri-

ﬁcation ofsupervised learning procedures are described in Sect. 3;

the results for each class are presented in Sect. 4; and conclusions

are drawn in Sect. 5. Special training selections applied to a sub-

set of classes are detailed in Appendix A; selected classiﬁcation

attributes are listed in Appendix B; additional class labels from

the literature (among the false positive classes listed in Table 3)

are deﬁned in Appendix C; some examples of queries to facilitate

the exploitation of classiﬁcation results in the Gaia archive

are

provided in Appendix D; and common diagrams for all classes,

including a summary of trained and classiﬁed sources, an assess-

ment of the results with respect to the literature, and sample light

curves, are presented in Appendix E. All table names in the Gaia

archive that are mentioned in the text assume the preﬁx gaiadr3

(as shown in Appendix D).

2. Data

As part of the Gaia variability pipeline (Eyer et al. 2023), the gen-

eral classiﬁcation module received – as input – sources with pho-

tometric time series in the G, G

, and G

bands (Riello et al.

2021) that had at least ﬁve ﬁeld-of-view (FoV) measurements

in the G band, which were already identiﬁed as potential vari-

able sources and characterised by basic statistics and periodic-

ity parameters. Before any computation, sources and associated

epoch FoV transits were processed by the chain of operators

described in Sect. 10.2.3 of the Gaia DR3 documentation

(Rimoldini et al. 2022) and Sect. 3.1 of Eyer et al. (2023), which

selected, transformed, and cleaned time series from spurious or

doubtful observations. The balance between outlier removal and

signal preservation favoured the latter, considering that some of

the targeted variability types relied on a small number of outlier-

like measurements (such as Algol-type eclipsing binaries and

microlensing events). All time series and derived statistical num-

bers hereafter refer to these cleaned time series. The median

number of FoV measurements in the three photometric bands is

between 40 and 44 per source (Eyer et al. 2023), within a time

span of typically 900–1000 days in the G band.

While the processing of Gaia (early) DR3 photometry

included signiﬁcant calibration improvements with respect to

DR2 (Riello et al. 2021), some low-level uncalibrated system-

atic eﬀects remained and their impact on epoch photometry are

https://gea.esac.esa.int/archive/

described in Evans et al. (2023). Among instrumental eﬀects,

scan-angle dependent signals were induced mainly by asym-

metric extended sources (such as barred spiral galaxies and

tidally distorted stars) and multiple close pairs (.1

) of point-

like sources (Holl et al. 2023). Although such signals helped the

identiﬁcation of galaxies from photometric variations, in general

data artefacts might interfere with the correct identiﬁcation of

classes with genuine variability, especially those associated with

low signal-to-noise ratios.

The classiﬁcation of variables employed also astromet-

rically derived parameters such as parallax and proper

motion (Lindegren et al. 2021b). However, Gaia DR3 astro-

physical parameters (Andrae et al. 2023; Creevey et al. 2023;

Delchambre et al. 2023; Fouesneau et al. 2023) could not be

included as they were processed in parallel and became avail-

able after the results of the variability pipeline were ﬁnalised.

A subset of classiﬁed sources were analysed in more detail

by subsequent SOS modules, typically focusing on speciﬁc

classes, as mentioned in Sect. 1. The results of all variability

modules were subject to additional source ﬁltering before their

ingestion into the public Gaia archive (Babusiaux et al. 2023).

Statistical parameters of all the photometric time series pub-

lished in Gaia DR3 are available in the vari_summary table.

3. Method

For Gaia DR3, general classiﬁcation relied on supervised

machine learning, that is, training classiﬁers with sources of

known variability types and applying the resulting models to

classify sources of an unknown variability type. Known vari-

ables in the literature are cross-matched with Gaia sources,

veriﬁed, selected, and characterised by attributes derived from

the Gaia data. The use of both cross-match sources and (opti-

mised) classiﬁcation attributes for training was described in

Rimoldini et al. (2019) and it is not repeated herein.

An extensive cross-match of Gaia sources was compiled by

Gavras et al. (2023), which provided millions of variable objects

from the literature and represented over 100 variability types. The

robustness of the cross-match method, which included astromet-

ric and photometric information in the identiﬁcation of matches,

and the veriﬁcation of the genuineness of literature classiﬁ-

cations ensured the reliability of training sources (critical to

supervised classiﬁcation) and of the validation of the results.

3.1. Training set

Potential training sources from literature were vetted for each

class to ensure the correct class membership. This was repeated

for every catalogue that was deemed trustable for training the

class under investigation. The reliance of supervised classiﬁca-

tion on known objects makes it vulnerable to biases from the

literature, for instance, related to their data acquisition and clas-

siﬁcation methods. Thus, in addition to class veriﬁcation, the

cross-matched objects were probed in several dimensions to

identify intrinsic biases, such as limited sky coverage or appar-

ent magnitude range with respect to the ones of Gaia, in order

to prevent (or minimise) the transfer of literature selection func-

tions to the Gaia classiﬁcations.

3.1.1. Published classes

Since it was diﬃcult to know a priori the full list of classes

that could be identiﬁed in Gaia DR3, the veriﬁcation of liter-

ature classiﬁcations and source selection for training purposes

A14, page 2 of 105

of 105

免费下载

opentenbase

关注

评论