École Evry Schatzman 2023: Stellar physics with Gaia

C. Babusiaux, C. Reylé (eds)

THE VARIABLE SOURCES IN THE GAIA ARCHIVE

L. EYER

Département d’Astronomie de l’Université de Genève, Chemin Pegasi 51, 1290 Versoix, Swizterland
e-mail: laurent.eyer@unige.ch
arXiv:2412.02744v1 [astro-ph.IM] 3 Dec 2024

Abstract. At the core of the Gaia mission is a multi-epoch survey consisting of astrometric, photometric,
spectrophotometric, and spectroscopic measurements. The astrometric time series provides parallaxes and
proper motions, along with information on astrometric binary systems. The photometric time series offers
a means to investigate the variability of the sources. Due to their whole-sky, multi-epoch nature, multiple
instruments, their magnitude range covering 21 magnitudes, and their remarkable photometric precision,
these data allow us to describe the variability of celestial phenomena in an unprecedented manner. For the
third Gaia Data Release (DR3), the data collection spanned 34 months, with a median number of field-of-
view measurements in the G band of about 44, reaching up to 270. At publication time, DR3 delivered
the largest collection of variable sources with an associated classification across the entire sky. All these
sources have their G, GBP , GRP epoch data published and accessible in the Gaia ESA archive. In summary,
there are 10.5 million variable sources, including 9.5 million variable stars and 1 million QSOs. Additionally,
2.5 million galaxies were identified thanks to spurious variability caused by the non-axisymmetric nature of
galaxies and the way Gaia collects data. Moreover, all the epoch data and time series of nearly 1.3 million
sources in a pencil beam around the Andromeda galaxy are published, regardless of their status (constant or
variable); This dataset is known as the Gaia Andromeda Photometric Survey. We also introduce the citizen
science project, GaiaVari to classify variable stars, the Focused Product Release delivered on October 10,
2023. In the future, DR4 will cover 66 months, and we hope DR5 will have accumulated 10.5 years of data.

Keywords: stars: variables: general - Galaxy: stellar content - catalogues - stars: oscillations - binaries:
eclipsing - starspots

1 Introduction

In astronomy, from its beginnings to now, the primary means of learning about the Universe and its constituents
has been through the photons. We can find the following texti by Evry Schatzman in his book ”Astrophysique”
Schatzman (1963):
The fundamental basis of astrophysics lies in the radiation that reaches us from celestial objects.
Whether in the visible or invisible spectrum, it is always a form of radiation that we measure.
Only from the late 20th century were alternative methods used to learn about our Universe, first as exceptions
to the sentence above, such as the cosmic rays (Hess 1912), then with important implications, such as the
neutrinos (from the Sun, Davis et al. 1968), and the gravitational waves in 2015 (Abbott et al. 2016).
The classical subdivision in observational astronomy about the three main observation branches are:
• Astrometry: Knowledge of celestial objects obtained from their positions, motions, and shapes. Probably
this discipline is as old as humanity.
• Photometry: Knowledge of celestial objects obtained from the measurement of their integrated light over
a range of wavelengths. Historically, photometry went hand in hand with astrometry as they are the most
basic perceptions of the human eye. The magnitude scale, as it is well known, is inherited from the ancient
greeks (and formalized by Norman Pogson, cf. Pogson 1856).

i Translated from French.

©Ecole Evry Schatzman 2023

2 EES 2023

• Spectroscopy: Knowledge of celestial objects obtained from their spectra. This technique emerged in
the 19th century, with Joseph von Fraunhofer discovering dark lines in the spectra of the Sun and other
stars, and Gustav Kirchhoff and Robert Bunsen correctly interpreting these dark lines as being caused by
specific atomic elements (for a brief historical account see Appenzeller 2013).
It should be remarked that we gain knowledge not only from the source but also from what is between
the source and the observer, e.g. the interstellar medium. Observations from the ground can also allow us to
determine the properties of the atmosphere and its evolution, such as in Burki et al. (1995).
Each of these three main observation branches benefits from multi-epoch observations. The most knowledge
is obtained when these different multi-epoch domains are combined. Such merging is fruitful for finding and
calibrating standard candles, identifying stellar populations, probing the invisible with microlensing events (e.g.
merging the photometry and the astrometry as in Sahu et al. 2022) and determining astrophysical parameters
thanks to binaries (Eyer et al. 2015), to pulsating stars (Baade-Wesselink method, Carney et al. 1992), etc.
It should be noted that on the side of photometric surveys, there has been a booming period for these past
30 years with OGLE (Udalski 2003), HAT (Bakos et al. 2004), Kepler (Borucki et al. 2010), and ZTF (Bellm
et al. 2019), to name just a few. The future is also bright with the LSST (Ivezić et al. 2019) and PLATO
(Rauer et al. 2014), again to name just a few. Spectroscopy is entering into the era of systematic multi-epoch
global surveys, obviously with Gaia, but also with 4MOST (de Jong et al. 2019), and in the future with WST
(Mainieri et al. 2024).
Now, if we turn to the Gaia cornerstone mission, the ”tour de force” of Gaia is to assemble multi-epoch data
for the entire sky on these three main observation branches: astrometry, photometry, and spectroscopy.
In this short text, we will mostly restrict the presentation to what has been put in the Gaia archive for the
variable sources.

2 The Gaia mission: An introduction

Let us review key aspects of the Gaia mission: Gaia is a cornerstone mission of the European Space Agency
science program (Gaia Collaboration et al. 2016), tasked with performing a systematic survey of all objects
brighter than G=20.7, attaining a survey of more than 2 billion celestial sources. The measurements collected
by Gaia consist of position (astrometry), brightness, and colours (photometry and spectrophotometry), along
with radial velocity (spectroscopy). It should be noted that the survey for radial velocities is restricted to
brighter sources (about G = 15). Gaia was launched on December 19, 2013 by a Soyuz rocket. The CCD
camera aboard Gaia is the largest ever deployed in space, reaching nearly one billion pixels. Over its ten-year
mission, each source, on average, will be observed 140 times in each of the 9 CCDsii in the G band, as well as
in each of the BP and RP CCDs. The number of measurements for the radial velocity spectrometer instrument
will be comparatively fewer, averaging approximately 80 times. The results of the Gaia mission, that the Data
Processing and Analysis Consortium processes, are made accessible to the public through Data Releases. DR4
and DR5 are anticipated to be delivered in 2026 and no later than 2032, respectively.
We will focus on this article on the DR3 results, which compiles 34 months of observations and nearly one
trillion CCD measurements over the whole celestial sphere.
Several factors contribute to Gaia’s exceptional status as a mission:
• Unprecedented Astrometry: Gaia provides unparalleled astrometric data, i.e. positions, parallaxes, and
proper motions for more than a billion stars over the whole celestial sphere. The astrometric precision is
available on the Gaia webpage (https://www.cosmos.esa.int/web/gaiaiii ): For Gaia DR4, that is, for
the nominal mission, the parallax uncertainty is 22 µas at magnitude 15. This precision can be compared
to the 20-25 µas estimates from pre-launch assessments in 2006 for a G2 star at magnitude 15 (see e.g. Eyer
2006). It’s important to note that the errors in position and proper motion need to be multiplied by 0.8
and 0.5, respectively. The astrometry is also able to detect astrometric binaries, exoplanets, microlensing
effects, etc. . .
• Three Instruments on a Single Platform: astrometry, [spectro-]photometry, spectroscopy with the deriva-
tion of radial velocities. This results in a unprecedented uniformity of data across the entire celestial
sphere.

ii for each of the 7 raws at the exception of raw 5, for which the number is 8.
iii Note: This page has undergone changes in URL and error estimates following the better understanding of the performance.
 L. Eyer: The Variable Sources in the Gaia archive 3

• Multi-Epoch Measurements: The mission conducts multi-epoch measurements of the entire sky, providing
a comprehensive view of variable celestial objects, see also Section 3.

• 10-Year Time Baseline: With a 10-year mission duration (if all goes as planned), Gaia offers high frequency
precision, particularly beneficial for periodic objects.

• Quasi-Simultaneous Measurements: Gaia achieves quasi-simultaneous measurements. Other multi-band

photometric surveys do not have this simultaneity and this complicates somewhat the analysis. For more
details, see Section 3.

• Time-Domain Selection Function: Because of its predefined scanning law, Gaia allows us to determine the
time-domain selection function, enabling the understanding of observational biases.

• Extensive Number of Measurements: Gaia’s data volume is significant, with nearly 1 trillion CCD mea-
surements for DR3 (i.e. 34 months).

• High Dynamical Range: Gaia covers a high dynamical range, capturing data from the brightest sources
(G ≈ 1.7 in DR3) to magnitudes as faint as G ≈ 21/22.

• Space-based Operation: Gaia’s position in space provides stability and allows us to access the entire ce-
lestial sphere from a single platform, which is not achievable by any single ground-based optical telescope.
However, operating in space presents its own challenges, such as the high cost of space missionsiv , the
impossibility of repairs, micrometeoroid hits, and solar flares/the radiation damages. Indeed Gaia’s loca-
tion at L2, 1.5 million kilometers from Earth in the direction opposite to the Sun, leaves it unshielded by
Earth’s magnetic field.

• Cyclic Improvements: The Gaia consortium implements cyclic improvements, performing systematic data
analyses, with not only more data at each cycle, but with improved calibrations, and enhanced outlier
detection.

Each of these elements makes Gaia extremely fruitful for variable and binary stars. These are among the
objects that benefit the most from Gaia. Due to the semi-regular Gaia sampling, strictly periodic objects can
be studied in detail despite the gaps in the scanning law. Also transient sources are and will be classified in the
data releases (Eyer et al. 2023; Rimoldini et al. 2023) and are also detected in real time by the Gaia Science
Alert System (Hodgkin et al. 2021). Indeed, since the beginning of its science operations, Gaia has continuously
(with few interruptions) provided alerts, identifying potentially time-sensitive events. If not addressed promptly,
these events could lead to significant scientific loss. To date (June 2024), the Science Alerts team has released
over 25,000 of these alerts. About 27% of the alerts are classified, with approximately 60% of these classified
alerts identified as supernovae.

3 The Gaia time sampling

The time samplingv of Gaia is very particular and semi-regular. It is optimized to achieve astrometric precision
as uniform as possible across the entire celestial sphere. This sampling is known as the Nominal Scanning Law
(NSL): Gaia continuously sweeps the sky, with the spacecraft’s rotation axis precessing on a cone with a 45-
degree opening angle to the Sun. A full precession cycle on the cone is completed every 63 days. The spacecraft
rotates on its axis every 6 hours, and the 106.5-degree angle between Gaia’s two fields-of-view (preceding and
following fields-of-view) results in a time interval of 1 hour and 46 minutes between them (the exemplary angle
of 253.5-degree between the following and preceding fields-of-view takes thus 4 hours 14 minutes).
Figure 1 shows the sky plot for the number of observations (the field-of-view transit). We took the field
matched transits, i.e. the total number of field–of–view transits matched to this source. This number is higher
or equal to the num selected g fov of gaiadr3.vari summary table, because it does not take into account the
number of observations that are removed.
The second plot, Fig. 2, presents the number of CCD observations as a function of ecliptic latitude. On a
field crossing the CCD observations are separated by 4.85 seconds. The Annex A of Eyer et al. (2017) provides

iv As Prof. Bohdan Paczynski humorously noted, ”One dollar in space is worth less than one dollar on Earth”.
v referred to as the cadence in LSST/Vera C. Rubin observatory.
4 EES 2023

detailed properties of the scanning law for 5 years, including various representations, histograms of the lags,
and spectral windows for different regions of the sky.
Because the scanning law is well-defined, it is possible to study the selection function. Eyer & Mignard
(2005) conducted a study on periodic signals based on an earlier design of the spacecraft and a 5-year scanning
law. They concluded that the ability to detect the correct period of a periodic signal strongly depends on the
ecliptic latitude. For DR3, significant differences in detection exist even at fixed ecliptic latitudes due to the
scanning law creating regions with a low number of observations; see Fig. 1.
There are two exceptions to the Nominal Scanning Law:
1. At the beginning of the mission, a 28-day period was planned to assist with photometric calibrations.
During this time, the same stars and regions of the sky were scanned nearly continuously. The axis of
rotation of the spacecraft was kept in the ecliptic plane (at a 45-degree angle to the Sun) without precession,
allowing the ecliptic poles to be scanned regularly during these 28 days, with repeated sequences of intervals
1h46m, 4h14m, 1h46m, etc. As the spacecraft orbits around the Sun, stars at lower ecliptic latitudes (in
absolute value) are observed for shorter sequences.
2. The second exception occurred when, for 12 monthsvi , a modified NSL was introduced to improve the
astrometric solution. As a positive side effect it reduces aliasing peaks. In this mode, the precession of the
spacecraft’s rotation axis was reversed. However, the spacecraft’s rotation direction remained unchanged
due to the constraints of the Time Delay Integration (TDI) mode. Note that the reverse motion of the
rotation axis on the cone increased the consumption of cold gas.
Although the scanning law determines which stars are observed, understanding the temporary absence of an
observed source is challenging. Several factors could explain a missing observation: the star might have truly
disappeared (the most interesting case) by becoming very faint due to an eclipse or occultation, it could be just
below the detection threshold of the less precise sky mapper and thus not observed, the star can be perturbed
by the other field-of-view (e.g. bright star/extended object) which is superposed on the same focal plane, the
data might have been deleted due to excessive data volume onboard, or technical problems might have occurred.

4 The Gaia G-band photometric precision

Fig. 3 shows the uncertainty in the mean G-band photometry as a function of the G magnitude. There are
two plots: the results from DR1 and DR3. DR2 is not displayed to avoid too many figures and it falls between
the two. The plot is based on a selection of stars with absolute ecliptic latitudes below 60 degrees to avoid the
perturbation of ecliptic pole scanning. The uncertainty is estimated from the flux standard deviation divided by
the square root of the number of measurements (after some clipping). Naturally, this uncertainty decreases as
the number of measurements increases. In case of variable sources, the standard deviation is obviously larger.
Although more measurements may reveal more instrumental calibration features, this is not the case here. On
the contrary, some features visible in DR1 are less prominent or have disappeared in DR3 demonstrating the
improvements of the calibrations. It should be noted that in the Gaia archive, uncertainties are provided in flux
rather than in magnitude. Therefore, these uncertainties need to be converted into magnitude uncertainties for
Fig. 3. We used the formula Uncertainty(magnitude) = 1.086 × Uncertainty(flux)/flux, where the term 1.086 is
coming from 2.5/ ln(10).
We will not enter into the debate on using flux versus magnitudes to evaluate the brightness of a celestial
source here, opening Pandora’s box. The main point is to avoid being too dogmatic about these issues and
to use the approach that makes the most sense for the problem at hand. As a side note, negative fluxes
were unfortunately excluded in the mean calculation process (Riello et al. 2021; Evans et al. 2023). Negative
fluxes can occur due to statistical variations or overestimation of the background for example. This exclusion
introduces a bias in the magnitude estimation at the faint end: the sources are brighter than they are in reality.
Starting from DR4, negative fluxes will be included.

5 The Gaia data releases and the variability

The data release process is iterative and advances with each step in both quantity and quality, this is particularly
evident for the variability processing and analysis. So let us have a quick look at the DR1 and DR2 data releases.

vi 6 months will be included in DR4, the rest in DR5.

 L. Eyer: The Variable Sources in the Gaia archive 5

Fig. 1. Number of fields-of-view data per source in ecliptic coordinates (top) and Galactic coordinates (bottom) for
DR3 in Aitoff projections. We see the complex sampling which is clearly not completed yet, with blue regions, i.e. low
number of transtis, around the ecliptic plane. We see that one of these low number of measurements is in the direction
of the lower bulge at positive Galactic longitudes.

A global pipeline was set up for the variability processing and analysis of Gaia data. Initially, it was applied to
the cycle 1 data, the details of this processing can be found in Eyer et al. (2017). The DR1 publication also covers
the data properties of cycle 1, along with certain aspects of the 5-year sampling obtained through simulations.
Based on 14 months of data, the DR1 variable star catalogue showcased a selection of 3,194 variable stars. This
data release consisted of two types of variability, namely Cepheids and RR Lyrae stars in the direction of the
Large Magellanic Cloud (Clementini et al. 2016). The Large Magellanic Cloud, situated near the South Ecliptic
pole, benefited from the regular sampling during the 28 days of the Ecliptic Pole Scanning Law. In addition,
a performance verification paper (Gaia Collaboration et al. 2017) presented the TGAS parallaxes for Cepheids
and RR Lyrae stars, demonstrating that already the improvements with respect to Hipparcos were impressive.
6 EES 2023

Fig. 2. Number of CCD observations per source as a function of ecliptic latitude (β) for DR3. A large number of
observations are visible at ±45◦ due to the Nominal Scanning Law. Additionally, stars at the ecliptic poles (both ends
of the plot) have an exceptionally high number of observations, exceeding 2000, due to the Ecliptic Pole Scanning Law,
which was in effect for the first 28 days of the mission.

0:01 0:01
Uncertainty of the Mean

Uncertainty of the Mean

0:005 0:005

0:002 0:002
0:001 0:001
5e ¡ 4 5e ¡ 4

2e ¡ 4 2e ¡ 4
1e ¡ 4 1e ¡ 4

6 8 10 12 14 16 18 20 6 8 10 12 14 16 18 20
Mean G Magnitude Mean G Magnitude

Fig. 3. Uncertainty on the mean versus the mean G magnitude. Two plots are displayed side by side, for the comparison
of DR1 (left) and DR3 (right) photometry. We see that not only has the noise gone down thanks to the square root of
the number of measurements but also that the calibration has significantly improved. The various changes in windowing
schemes or gates are better taken into account in DR3 than in DR1.

The DR2 based on 22 months of operation data reached 550,737 variable stars (Holl et al. 2018). The DR2
contained the following variability types: RR Lyrae stars, Cepheids, long period variables, δ Scuti/SX Phoenicis
stars, stars with rotation modulation and stars with short time scale variability. A performance verification
paper (Gaia Collaboration et al. 2019) presented the properties of variable sources across the Hertzsprung-
Russel Diagram. As a remarkable result, we can follow the path of stars, their motion in the Hertzsprung-Russel
Diagram as their magnitude and colour change.
More generally, information and details for all data releases can be found in various articles (many are
listed in this section for details on variability processing and analysis) or in the Gaia documentation associated
with the data releases on the archive website, https://gea.esac.esa.int/archive/documentation/GDR3/
for DR3. This document, available online and also in PDF format, contains more than 1,500 pages.
It is also important to consult the ”known issues” page. For each data release, there is a section on the
 L. Eyer: The Variable Sources in the Gaia archive 7

Data tab of https://www.cosmos.esa.int/web/gaia/data. Under these sections, you can find the known
issues for each data release. For example, the known issues for DR3 can be found at the following address:
https://www.cosmos.esa.int/web/gaia/dr3-known-issues.
A note on the archive names: you have in the ESA Gaia archive (https://gea.esac.esa.int/archive/)
the different data sets associated with the different data releases, gaiadr1.table name, gaiadr2.table name,
gaiaedr3.table name, gaiadr3.table name, gaiafpr.table name which relates the Gaia Data Release 1
(DR1), Gaia Data Release 2 (DR2), Gaia Early Data Release 3 (EDR3), Gaia Data Release 3 (DR3), Fo-
cused Product Release (FPR) respectively. For a given source, the source id can change. A last remark on the
nomenclature related to variability: for the tables names associated to the variability processing and analysis,
starting from the second data release, these table names start with ”vari table name”.
In the following sections, we will focus more on the DR3 as it has the most numerous outputs and diversity
in variability types.

5.1 The DR3 results

For the third data release of Gaia, the variability analysis is described in Eyer et al. (2023). In summary, as of
the data release in June 2021, Gaia provided the largest catalogue ever of classified variable sources across the
entire sky. In DR3, we have 10.5 million variable sources, comprising 9.5 million variable stars and 1 million
AGN. Additionally, thanks to a spurious perturbation in the photometry, we identified 2.5 million galaxies.
A general overview of the DR3 result, as mentioned above, can be found in Eyer et al. (2023). In the
Gaia archive https://gea.esac.esa.int/archive/, the different associated tables related to the variability
processing and analysis can be seen in Fig. 4.

Fig. 4. Tables available related to the variability processing and analysis in the Gaia archive. We opened the
vari summary table with just the first three lines.

A vast and systematic compilation of the literature was done in Gavras et al. (2023); it combines the sources
of 152 catalogues crossmatched with the Gaia data and contains 4.9 million variable sources. The catalogue
8 EES 2023

Variability types Variability types

(1) ACV—CP—MCP—ROAM—ROAP—SXARI: (13) MICROLENSING: Star with microlensing event
α2 CVn, (Magnetic) Chemical Peculiar, Rapidly
Oscillating Am/Apvii SX Arietis stars
(2) ACYG: Alpha Cygni-type variable (14) RCB: R Coronae Borealis stars
(3) AGN: Active Galactic Nuclei (including Quasars) (15) RR: RR Lyrae stars
(4) BCEP: Beta Cephei type variable (16) RS: RS Canum Venaticorum type variable
(5) BE—GCAS—SDOR—WR: Subset of eruptive (17) S: Set of stars with short timescale variability
variable types: B-type emission line star, Gamma
Cassiopeiae, S Doradus, and Wolf-Rayet
(6) CEP: Cepheid variable types: delta Cepheid, (18) SDB: Subdwarf B stars
anomalous Cepheid, and type-II Cepheid
(7) CV: Cataclysmic variable (19) SPB: Slowly Pulsating B-star variable
(8) DSCT—GDOR—SXPHE: delta Scuti, gamma (20) SN: Supernovae
Doradus, and SX Phoenicis
(9) ECL: Eclipsing Binaries of types: Beta Persei (21) SOLAR LIKE: Stars with solar-like variability
(Algol), Beta Lyrae, and W Ursae Majoris due to flares, spots, and rotational modulation
(10) ELL: Ellipsoidal variable (22) SYST: Symbiotic variable star
(11) EP: exoplanetary transits (23) WD: Variable White Dwarf of types: ZZ Ceti
(DAV, ZZA), V777 Her (DBV, ZZB), and GW Vir
(DOV, ZZO)
(12) LPV: Long Period Variable stars of types: omi- (24) YSO: Young Stellar Object
cron Ceti (Mira), OGLE Small Amplitude Red Gi-
ants, and semiregular

Table 1. The 24 variability types considered in DR3. Please note that in some cases, a denomination encompasses
several different variability types.

is available at the CDS (Centre de Données astronomiques de Strasbourg, https://cds.unistra.fr, Gavras

et al. (2022)).
Part of this compilation was used as a training set for the supervised classification (Rimoldini et al. 2023).
For the classification, the methods employed are Random Forest and eXtreme Gradient Boosting models in
a mixture of multi-class and binary classifiers. There are 24 variability (grouped) classes. Some of these are
combined types. Therefore, when counting the subclasses, we reach a total of 35 variability types. The variability
classes list can be found in Table 1.
The table that contains the variability type class from the Machine Learning is gaiadr3.vari classifier result.
A validation of the classification for the young stellar object class was done by Marton et al. (2023).
After the classification, Specific Object Studies are done for various topics. They act as a validation of the
classification, and also these activities compute further parameters that describe the class. It is important to
note that there are two sources for a given variability type: (1) The Machine Learning classification and (2) the
classification of Specific Object Studies. There may be differences between these two classifications.
The Specific Object Studies covers the following topics (listed in the alphabetical order, as in the Gaia
archive). Here we just give the number of sources and also some selected highlights:
• AGN: gaiadr3.vari agn Carnerero et al. (2023). There are 872,228 sources in this table, this number
is obtained with the ADQL query (just shown for this example):

SELECT COUNT(*)
FROM gaiadr3.vari_agn

We can compare this number from the number of AGN from classification, we have 1,035,207 sources, see
following ADQL command:

SELECT COUNT(*)
FROM gaiadr3.vari_classifier_result
WHERE best_class_name = ’AGN’
 L. Eyer: The Variable Sources in the Gaia archive 9

A surprising result is that even with the gaps of the Gaia time series, the time delay for a source, the lens
system DESJ0501-4118, was derived.
• Cepheids: gaiadr3.vari cepheid Ripepi et al. (2023). There are 15,021 sources. A Fourier series
model of the light curve is fitted and the parameters are published in this table. The estimation of the
mean should be taken for this table and not from gaia source table. A subclassification is also given.
Radial velocities time series are published for 798 sources. It is the largest catalogue of Cepheids having
radial velocities.
• Compact companion: gaiadr3.vari compact companion Gomel et al. (2023). There are 6,306 sources.
If the variability is caused by ellipsoidal deformation, a mass ratio can be determined from the Fourier
parameters of the light curve. This mass ratio helps identify potential candidates for white dwarfs, neutron
stars, and black holes. Other methods to detect black holes are based on astrometry and radial velocities
(El-Badry et al. 2023; Chakrabarti et al. 2023; El-Badry et al. 2023; Gaia Collaboration et al. 2024).
• Eclipsing binaries: gaiadr3.vari eclipsing binary Mowlavi et al. (2023). There are 2,184,477
sources. The largest catalogue of eclipsing binaries ever published on the whole sky. In this table,
we adopt a geometrical modelling of the light curve and extract some relevant statistics. A sub-sample is
then studied with a physical modelling, see Pourbaix et al. (2022). The results are published in the non-
single star table gaiadr3.nss two body orbit. It contains 86,918 systems with the flag nss solution type
EclipsingBinary and 155 for EclipsingSpectro.
• Long Period Variables: gaiadr3.vari long period variable Lebzelter et al. (2023). There are
1,720,588 sources. Thanks to the RP spectrophotometry, 546,468 stars were classified as carbon rich
candidates.
• Microlensing: gaiadr3.vari microlensing Wyrzykowski et al. (2023). This is the first whole sky
microlensing event list, which contains 363 sources.
• Main Sequence Oscillators: gaiadr3.vari ms oscillator Gaia Collaboration et al. (2023a). There
are 54,476 sources. See Section 5.1.1.
• Planetary transits: gaiadr3.vari planetary transit Panahi et al. (2022b). The table contains 214
sources. Please note that the table gaiadr3.vari planetary transit 13june2022 is incorrect. For
details, please see Section 5.1.3.
• Rotation modulation: gaiadr3.vari rotation modulation Distefano et al. (2023). There are 474,026
sources. The Period Amplitude diagram shows different regimes.

• RR Lyrae stars: gaiadr3.vari rrlyrae is taking care of the RR Lyrae variables and is described in
Clementini et al. (2023). There are 271,779 sources. RR Lyrae stars are well known standard candles.
There are several subtypes of them: the Bailey’s ab and c types corresponding to fundamental mode
pulsation and first overtone, respectively. The ab type light curves have a very recognisable shape with
a sharp rise and slower declined with a period typically of half a day, the c types have a more sinusoidal
light curve. An example of the quite astonishing results is the metallicity map derived from the RR Lyrae
light curves in Figure 5. There are 1,100 sources with radial velocities time series which are in the table
of RR Lyrae stars.
• Short Time scales: gaiadr3.vari short timescale is described in Eyer et al. (2023). There are
471,679 sources. This procedure did not go through removal of sources and reveals also instrumental
and/or calibration artefacts.
For the completeness of the archive description, three additional tables are listed under the variability folder:
one is a summary of the variability analysis, and the other two are associated with the radial velocities of
Cepheids and RR Lyrae stars:
• gaiadr3.vari summary table provides a global overview. It helps determine which table contains a given
source (Eyer et al. 2023), and provides statistical attributes of the time series.
10 EES 2023

• gaiadr3.vari epoch radial velocity: the table contains the time series of the radial velocities.

• gaiadr3.vari rad vel statistics: the table contains statistics of the radial velocities.

Fig. 5. Distribution of RR Lyrae stars in a 3D projection using the distance modulus. The LMC and SMC can be seen
behind the Milky Way RR Lyrae stars. The metallicity near the Galactic plane is higher, lower in the halo, then lower in
the LMC and then even lower in the SMC, as expected. We also see the Sagittarius stream. We can guess the presence
of other structures in the halo.

5.1.1 The Performance Verification Paper on variability

A performance verification paper (PVP) authored by Gaia Collaboration et al. (2023a) investigate the pulsation
properties in main sequence OBAF-type stars. Among its noteworthy findings are: (1) the establishment of
a period-luminosity relation for δ Scuti stars. (2) It is further shown that stellar rotation seems to reduce the
amplitude of variability. (3) Discrepancies arise when comparing the observed locations of these variables with
the predictions of theoretical models. Such inconsistencies may be attributed to inaccuracies in determining
astrophysical parameters using Gaia instruments, and/or it is conceivable that rotational effects play a significant
role, potentially altering the instability regions within the Hertzsprung-Russell diagram. (4) A quite remarkable
result (see Appendix B of the work by Gaia Collaboration et al. (2023a)) which demonstrates Gaia performance:
with as few as 40 measurements, Gaia detects two accurate frequencies in certain cases when compared with
the reference observations from the Kepler mission. As Gaia extends its measurement duration to a decade, the
precision of frequency determination, which scales inversely with the observation time basis, is anticipated to
reach exceptional levels of accuracy.
 L. Eyer: The Variable Sources in the Gaia archive 11

There is also a publication on the spurious effect of spurious photometric variability introduced by the angle
by which the satellite scans a (non axi-symmetric) source (Holl et al. 2023a).

5.1.2 The Gaia Andromeda Photometric Survey

To offer a teaser of DR4 and DR5, which will include epoch photometry/time series for all sources, we provide
in the Gaia archive a pencil beam with a 5.5-degree opening angle centred toward the Andromeda galaxy,
ensuring photometry availability for all sources within this region (Evans et al. 2023). This selection was made
to provide a sufficiently broad sample of celestial sources reaching nearly 1.3 million sources; in addition to the
Andromeda galaxy stars, the foreground stars, with significant parallaxes, allow us to provide a well-populated
Hertzsprung-Russell Diagram. This dataset contains all the G, GBP , GRP time series in the Gaia archive. The
Evans et al. (2023) article also presents a renormalisation of the photometric uncertainties (as these uncertainties
are usually underestimated). We also show a way to combine the three photometric bands with a Principal
Component Analysis method to enhance the detection and the characterisation of the variable sources, an
approach similar to what was done with SDSS data (Süveges et al. 2012).

5.1.3 The exoplanets from Gaia with the transit method

Although the time span of DR3 was limited to 34 months, the quality of the photometry is sufficient to discover
exoplanets by the transit method (Panahi et al. 2022b). Currently, there are two exoplanets detected by the
transit method and confirmed by radial velocity measurements, Gaia-1b and Gaia-2b. The list of candidates is
provided in the Gaia table vari planetary transit and contains 214 sources: 173 systems are known, and 41
are new (among them the two above already cited); the list of these 41 candidates can be found in Eyer et al.
(2023). There is further work with the exoplanets with the transit methods, there has been a special agreement
with the TESS mission (Ricker et al. 2015) to verify the TESS candidates by Gaia. Again, with only 34 months,
we are able (1) to show that about 5% of the TESS candidates are background eclipsing binaries and also (2) to
confirm about 5% of the cases where the transit can be identified (Panahi et al. 2022a).
As a side note, it is noteworthy to mention that: (1) astrometric exoplanet candidates were proposed by Holl
et al. (2023b). At the time of writing, one of these candidates, named Gaia-3b, was confirmed thanks to ground-
based radial velocity measurements (Sozzetti et al. 2023). (2) Exoplanets have also been discovered through
microlensing. Gaia21dnc (Ban et al., in Prep) and Gaia22dkv (Wu et al. 2023) are the first two exoplanets
detected through follow-ups of microlensing events alerted by the Gaia Science alert system (Hodgkin et al.
2021).

5.2 The focused product release: Time domain radial velocities for Long Period Variable stars
After the third data release, on October 10, 2023, there was a focused product release by the Gaia consortium,
consisting of several products:

• Astrometry and photometry from engineering images taken in the ω Centauri region.

• The first results of quasars’ environment analysis for gravitational lenses search.

• Extended radial velocity epoch data for Long Period Variables.

• Diffuse Interstellar Bands from aggregated RVS spectra.

• Updated astrometry for Solar System objects.

Here we present the Long Period Variables (LPV) (Gaia Collaboration et al. 2023b). The results of this
focused product release are based on the DR3 processing both in photometry and radial velocities. A reminder
of DR3 on Long Period Variables: It contains 1.7 million LPV. A selection was performed of stars fainter than
G = 6 and brighter than G = 14. The total number of sources is 9,614 stars with radial velocities times series.
Though this number may look small, it is a record holder: it is the largest sample ever published of Long Period
Variable stars with radial velocity time series. It should be noted that the knowledge of radial velocities allows
us to distinguish between pulsation (so true LPVs) and ellipsoidal variability.
The tables in the Gaia archive are located in the directory Gaia Focused Product Release and the subdirectory
Variability. There are three tables:
12 EES 2023

• gaiafpr.vari epoch radial velocity: Contains the radial velocity time series.

• gaiafpr.vari long period variable: Similar to the DR3 table of the same name, but with added radial
velocity data (frequencies and amplitudes are recomputed in both the radial velocity and G photometric
band time series).

• gaiafpr.vari rad vel statistics: Contains statistics on the radial velocities.

5.3 Citizen Science Projet: GaiaVari

GaiaVari is a citizen science project available on the Zooniverse platform that invites the community to classify
variable stars from the public Gaia DR3 results based on the following various visual elements:

• The light curve, representing the G band magnitude measurements as a function of time (colour-coded
with time).

• The folded curve, displaying the G band magnitude measurements as a function of phase computed from
the observed time and an estimated period (the colour-coded according to the light curve, so according
to the time, not the phase).

• The position in the Hertzsprung-Russell Diagram (the absolute G magnitude as a function of GBP −GRP )
and the motion of the variable star within this diagram. There are no corrections from the extinc-
tion/reddening.

• The position in the Milky Way (the distance is taken as the inverse of the parallax).

The project is limited to stars with sufficiently good parallax measurements, allowing for the determination of
a distance so that it can be placed within the Milky Way. The sample to classify is not exclusively focused on
periodic objects, although the majority of objects chosen fall into this category. An example of the representation
is given in Figure 6 for an RR Lyrae star.
Up to now, there have been one beta campaign and two official campaigns of GaiaVari. The first one classified
the source into the possible following variability types: Eclipsing binaries, Cepheids, RR Lyrae stars and long-
period variable stars. In the second campaign, the following variability types were added to the previous list:
δ Scuti stars and stars with ellipsoidal variability. In both campaigns, a ”None of the above” option was also
introduced.
After making a classification choice, and only then, there is the possibility to leave comments. By choosing
this option, one can also discover the Gaia sourceID and the classification from the automated supervised
machine learning method (Rimoldini et al. 2023). There is also a link towards ESASky (Merı́n et al. 2015),
and then one can obtain information from databases like Simbad at ”Centre de Données astronomiques de
Strasbourg” (CDS) (Wenger et al. 2000), etc.
Some citizen scientists systematically classify RR Lyrae stars and Cepheids into subgroups, such as RRab,
RRc types, or Type-I, Type-II Cepheids. Such sub-classifications can be compared with the Gaia classification
(Clementini et al. 2023; Ripepi et al. 2023) or the literature.
The community easily identifies the Blazhko effect, although, at times, it is challenging to distinguish Blazhko
from an RRd type (double-mode RR Lyrae stars). Additionally, systematic differences in classification between
what is published in DR3 and the classifications by some highly educated citizen scientists, particularly in RR
Lyrae and Cepheids and their sub-classification, are noteworthy. It will be very interesting to compare these
differences.
Due to demand we also created a permanent educational campaign for teachers and schools that can use it
as an introduction resource to variable stars and HR diagram understanding.
We plan to conduct additional campaigns. At a point, the plan is to include radial velocity information for
a more comprehensive dataset offered by Gaia.
Furthermore, there are plans to write articles on the results of GaiaVari from the first two campaigns,
GaiaVari is funded by ESA and driven by Sednai Sarl with a collaboration with the University of Geneva,
ESA and ScienceNow, see https://www.zooniverse.org/projects/gaia-zooniverse/gaia-vari.
 L. Eyer: The Variable Sources in the Gaia archive 13

Fig. 6. RR Lyrae star of Bailey’s type ab in GaiaVari. Upper left panel: the time series: Lower left panel: the folded
curve, with the period fo 16 hours. Upper right panel: the motion of the star in the HR diagram. Lower right panel:
the position in the Milky Way.

6 Conclusions

The quantity and quality of Gaia’s data are impressive. Among the many scientific subjects that Gaia is
touching, one is particularly positively impacted: time domain astronomy. In this context, we mostly presented
the analysis of photometrically variable sources of DR3, covering 34 months of data with its multitude of results.
Even with the relatively short DR3 time span, numerous records were broken in the number of variable sources,
and the unprecedented description of them sets a new standard. Looking ahead, Gaia’s data will nearly double
its number of measurements in DR4 and, again, nearly double it for DR5, reaching about 10.5 years of science
operational data (if all goes as planned). The Gaia variability pipeline for processing and analysing data in DR4
and DR5 will receive additional variability classes, features and improvements to the current ones, ensuring not
only an increase in quantity but also an enhancement in quality. This holds the promise of even more interesting
results for both the fourth and fifth Gaia data releases. Undoubtedly, Gaia’s data alone provide remarkable
insights into variable sources. However, we can anticipate exciting developments, such as combining Gaia data
with other photometric (e.g. Vera Rubin observatory/LSST (Ivezić et al. 2019), ZTF (Bellm et al. 2019),
PLATO (Rauer et al. 2014), etc.) or spectroscopic surveys (e.g. WST (Mainieri et al. 2024)).

I am grateful to the organisers of this School, in particular to Carine Babusiaux. I also appreciated the discussions with PhD
students and postdocs. I would like to thank Mathieu Van der Swaelmen, whose question on the zero point parallax correction
prompted a change in the Gaia consolidated catalogue for DR4. I thank Nicholas Chornay and Krzysztof Nienartowicz for reading
and commenting on this manuscript.
I am thankful to the Geneva Gaia team, the Gaia Variability Processing and Analysis international team (CU7), and DPAC
consortium. The success of such a mission lies in the competence and dedication of these Consortium members. Such an article
would not exist without their contributions.
14 EES 2023

During the writing of this document, I had to face the declining health and the passing of my mother, Andrée Eyer Pétermann.
Throughout my life, she was a constant source of support for my high aspirations and my work. I therefore dedicate this text to
her.
I met Evry Schatzman only once when I was a young Ph.D. student. I was deeply moved by his diligence and interest in the
young researcher I was, working on Hipparcos variable stars at the time. This is why I am quite certain that Evry Schatzman
would have been most enthusiastic about the Gaia mission and its results.
The texts and figures are original; several software were used to improve the readability and the quality of the text, such as
ChatGPT, Grammarly and Google Translate. This work made use of TOPCAT software (Taylor 2005). This work has made use
of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data
Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for the DPAC
has been provided by national institutions, in particular, the institutions participating in the Gaia Multilateral Agreement.

References

Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016, Phys. Rev. Lett., 116, 061102
Appenzeller, I. 2013, Introduction to Astronomical Spectroscopy
Bakos, G., Noyes, R. W., Kovács, G., et al. 2004, PASP, 116, 266
Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2019, PASP, 131, 018002
Borucki, W. J., Koch, D., Basri, G., et al. 2010, Science, 327, 977
Burki, G., Rufener, F., Burnet, M., et al. 1995, A&AS, 112, 383
Carnerero, M. I., Raiteri, C. M., Rimoldini, L., et al. 2023, A&A, 674, A24
Carney, B. W., Storm, J., & Jones, R. V. 1992, ApJ, 386, 663
Chakrabarti, S., Simon, J. D., Craig, P. A., et al. 2023, AJ, 166, 6
Clementini, G., Ripepi, V., Garofalo, A., et al. 2023, A&A, 674, A18
Clementini, G., Ripepi, V., Leccia, S., et al. 2016, A&A, 595, A133
Davis, R., Harmer, D. S., & Hoffman, K. C. 1968, Phys. Rev. Lett., 20, 1205
de Jong, R. S., Agertz, O., Berbel, A. A., et al. 2019, The Messenger, 175, 3
Distefano, E., Lanzafame, A. C., Brugaletta, E., et al. 2023, A&A, 674, A20
El-Badry, K., Rix, H.-W., Quataert, E., et al. 2023, MNRAS, 518, 1057
Evans, D. W., Eyer, L., Busso, G., et al. 2023, A&A, 674, A4
Eyer, L. 2006, Mem. Soc. Astron. Italiana, 77, 549
Eyer, L., Audard, M., Holl, B., et al. 2023, A&A, 674, A13
Eyer, L. & Mignard, F. 2005, MNRAS, 361, 1136
Eyer, L., Mowlavi, N., Evans, D. W., et al. 2017, arXiv e-prints, arXiv:1702.03295
Eyer, L., Rimoldini, L., Holl, B., et al. 2015, in Astronomical Society of the Pacific Conference Series, Vol. 496, Living
Together: Planets, Host Stars and Binaries, ed. S. M. Rucinski, G. Torres, & M. Zejda, 121
Gaia Collaboration, Clementini, G., Eyer, L., et al. 2017, A&A, 605, A79
Gaia Collaboration, De Ridder, J., Ripepi, V., et al. 2023a, A&A, 674, A36
Gaia Collaboration, Eyer, L., Rimoldini, L., et al. 2019, A&A, 623, A110
Gaia Collaboration, Panuzzo, P., Mazeh, T., et al. 2024, A&A, 686, L2
Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016, A&A, 595, A1
Gaia Collaboration, Trabucchi, M., Mowlavi, N., et al. 2023b, A&A, 680, A36
Gavras, P., Rimoldini, L., Nienartowicz, K., et al. 2023, A&A, 674, A22
Gavras, P., Rimoldini, L., Nienartowicz, K., et al. 2022, VizieR Online Data Catalog, J/A+A/674/A22
Gomel, R., Mazeh, T., Faigler, S., et al. 2023, A&A, 674, A19
Hess, V. F. 1912, Phys. Z., 13, 1084
Hodgkin, S. T., Harrison, D. L., Breedt, E., et al. 2021, A&A, 652, A76
Holl, B., Audard, M., Nienartowicz, K., et al. 2018, A&A, 618, A30
Holl, B., Fabricius, C., Portell, J., et al. 2023a, A&A, 674, A25
Holl, B., Sozzetti, A., Sahlmann, J., et al. 2023b, A&A, 674, A10
Ivezić, Ž., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111
Lebzelter, T., Mowlavi, N., Lecoeur-Taibi, I., et al. 2023, A&A, 674, A15
Mainieri, V., Anderson, R. I., Brinchmann, J., et al. 2024, arXiv e-prints, arXiv:2403.05398
Marton, G., Ábrahám, P., Rimoldini, L., et al. 2023, A&A, 674, A21
Merı́n, B., Salgado, J., Giordano, F., et al. 2015, arXiv e-prints, arXiv:1512.00842
 L. Eyer: The Variable Sources in the Gaia archive 15

Mowlavi, N., Holl, B., Lecoeur-Taı̈bi, I., et al. 2023, A&A, 674, A16
Panahi, A., Mazeh, T., Zucker, S., et al. 2022a, A&A, 667, A14
Panahi, A., Zucker, S., Clementini, G., et al. 2022b, A&A, 663, A101
Pogson, N. 1856, MNRAS, 17, 12
Pourbaix, D., Arenou, F., Gavras, P., et al. 2022, Gaia DR3 documentation Chapter 7: Non-single stars, Gaia
DR3 documentation, European Space Agency; Gaia Data Processing and Analysis Consortium. Online at ¡A
href=“https://gea.esac.esa.int/archive/documentation/GDR3/index.html”¿https://gea.esac.esa.int/archive/documentation/GD
id. 7
Rauer, H., Catala, C., Aerts, C., et al. 2014, Experimental Astronomy, 38, 249
Ricker, G. R., Winn, J. N., Vanderspek, R., et al. 2015, Journal of Astronomical Telescopes, Instruments, and Systems,
1, 014003
Riello, M., De Angeli, F., Evans, D. W., et al. 2021, A&A, 649, A3
Rimoldini, L., Holl, B., Gavras, P., et al. 2023, A&A, 674, A14
Ripepi, V., Clementini, G., Molinaro, R., et al. 2023, A&A, 674, A17
Sahu, K. C., Anderson, J., Casertano, S., et al. 2022, The Astrophysical Journal, 933, 83
Schatzman, E. 1963, Astrophysique (Masson et Cie)
Sozzetti, A., Pinamonti, M., Damasso, M., et al. 2023, A&A, 677, L15
Süveges, M., Sesar, B., Váradi, M., et al. 2012, MNRAS, 424, 2528
Taylor, M. B. 2005, in Astronomical Society of the Pacific Conference Series, Vol. 347, Astronomical Data Analysis
Software and Systems XIV, ed. P. Shopbell, M. Britton, & R. Ebert, 29
Udalski, A. 2003, Acta Astron., 53, 291
Wenger, M., Ochsenbein, F., Egret, D., et al. 2000, A&AS, 143, 9
Wu, Z., Dong, S., Yi, T., et al. 2023, arXiv e-prints, arXiv:2309.03944
Wyrzykowski, L., Kruszyńska, K., Rybicki, K. A., et al. 2023, A&A, 674, A23