© The Authors 2024

1. Introduction

With the release of its third catalogue (DR3, Gaia Collaboration 2023a), the European Space Agency Gaia mission has already revolutionised different fields of astrophysics. In particular, the view of the Milky Way stellar populations is being upgraded substantially (Gaia Collaboration 2023b). Regarding stellar physics and variable stars, Aerts et al. (2023) assessed the fundamental parameters and mode properties of 15 602 newly found Gaia/DR3 gravity-mode (g-mode) pulsator candidates, among the more than 1 00 000 new pulsators along the main sequence identified from their Gaia photometric light curves by Gaia Collaboration (2023c). These g-mode pulsators are low- and intermediate-mass main-sequence stars (masses between 1.3 and 9 M_⊙), and they are ideal laboratories for asteroseismology, within the broad landscape of such modern studies on the subject (see Aerts 2021; Kurtz 2022, for recent reviews).

Meanwhile Hey & Aerts (2024) extracted light curves assembled with the Transiting Exoplanet Survey Satellite (TESS; Ricker et al. 2015) for more than 60 000 of the candidate pulsators discovered by Gaia Collaboration (2023c). They confirmed the pulsational nature for a large majority and found them to be multiperiodic, with about 70% of them even sharing the same dominant frequency in the totally independent Gaia DR3 and TESS data. By comparing the astrophysical properties of the Gaiag-mode pulsators with those studied asteroseismically from Kepler data by Pedersen et al. (2021, for the B-type stars) and by Van Reeth et al. (2016), Li et al. (2020, for the F-type stars), Aerts et al. (2023) classified them either as Slowly Pulsating B (SPB) or γ Doradus (γ Dor) candidate stars. Our current work is focused on this last category of pulsating F-type dwarfs, covering masses 1.3 M_⊙ ≲ M ≲ 1.9 M_⊙. In the Hertzsprung-Russell diagram (HRD), they are found in a rather small main-sequence area (Fritzewski et al. 2024, Fig. 2), close to the cool limit of δ Scuti instability strip (Murphy et al. 2019). However, we do point out that many of the γ Dor stars have actually turned out to be hybrid pulsators when observed in high-cadence space photometry. Indeed a good fraction of these pulsators exhibit not only high-order g modes, but also acoustic waves known as pressure (or p) modes (Grigahcène et al. 2010; Hareter et al. 2011; Sánchez Arias et al. 2017; Audenaert & Tkachenko 2022). This makes their instability strip overlap with the one of the δ Scuti stars (Gaia Collaboration 2023c; Hey & Aerts 2024).

One of the new products published with Gaia/DR3 are the stellar atmospheric parameters derived from the analysis of the Gaia/Radial Velocity Spectrometer (RVS) spectra by the DPAC/GSP-Spec module (Recio-Blanco et al. 2023). The RVS spectra cover the Ca II IR domain (846–870 nm) and have a resolution around 11 500. By automatically analysing these spectra, Recio-Blanco et al. (2023) parametrised about 5.6 million single low-rotating stars belonging to the FGKM spectral type. Hotter stars (T_eff > 8000 K), or highly-rotating stars (more than ∼30–50 km s⁻¹, depending on the stellar type), were disregarded in Gaia/DR3 (this will be updated in the next Gaia Data Releases). These limitations come from the adopted reference grids that the parametrisation algorithms of GSP-Spec rely on. The derived stellar atmospheric parameters are the effective temperature T_eff, surface gravity log(g), global metallicity [M/H], and enrichment in α-element with respect to iron [α/Fe]. Moreover, up to 13 individual chemical abundances were also estimated for most of these stars. This analysis led to the first all-sky spectroscopic catalogue and the largest compilation of stellar chemo-physical parameters ever published. Moreover, radial velocities (V_Rad) of about 33 million stars were published in the DR3 catalogue (Katz et al. 2023). These data have allowed for the study of the Galactic kinematics and orbital properties of this huge number of stars, along with their atmospheric and chemical characteristics, keeping in mind that they belong to various populations in the Milky Way. This multitude of data also facilitates the constraining of stellar evolution models for a broad range of masses. In particular, the Gaia spectroscopic data give us the opportunity to unravel the properties of different kinds of variable stars, among which are the g-mode pulsators characterised photometrically by Aerts et al. (2023).

One of the main goals of the present study is to apply spectroscopic techniques to characterise the g-mode and hybrid pulsators identified in Aerts et al. (2023) and Hey & Aerts (2024). This perspective adds to their photometrically deduced properties and may help future asteroseismic modelling of the most promising pulsators. Some contamination by other or additional variability could also occur, such as rotational modulation. This is expected since Gaia Collaboration (2023c) and Aerts et al. (2023) could infer only one secure frequency for many of the 15 602 candidates. Moreover, Hey & Aerts (2024) found a fraction of the g-mode and hybrid pulsators reveal Ap/Bp characteristics in addition to their pulsational behaviour. Therefore, some of them might be Ap/Bp pulsators with anomalous chemical abundances from spots, as previously found from Kepler space photometry (e.g. Bowman et al. 2018; Henriksen et al. 2023). Gaia/GSP-Spec spectroscopic parameters could therefore help confirm the nature of these pulsators and facilitate the building of sub-catalogues of g-mode and hybrid pulsating stars with respect to their chemical properties. In addition, the main properties of these stars can be constrained from the Gaia spectroscopy and compared with the values deduced from the photometry, such as their effective temperature, surface gravity, luminosities, radius, etc.

Another goal is to explore the metallicity (and possibly the chemical abundances) of genuine Gaiag-mode and hybrid pulsators. This is an important input for asteroseismic modelling, once a sufficient number of oscillation frequencies has been identified (Aerts et al. 2018; Mombarg et al. 2021, 2022). Aside from the Kepler sample of γ Dor pulsators modelled by Li et al. (2020), the confirmed g-mode and hybrid pulsators from Hey & Aerts (2024) are currently under study to derive their global parameters (mass, convective core mass, radius, and evolutionary stage; Mombarg et al. 2024), as well as all of their significant oscillation mode frequencies from high-cadence TESS photometry. Future studies will point out whether their TESS data allow one to find period spacing patterns to assess their suitability for asteroseismology, as has been possible for γ Dor stars in the TESS Continuous Viewing Zones (Garcia et al. 2022a,b).

We note that the present work focuses only on the γ Dor stars or hybrid pulsators with dominant g-modes in the catalogue by Aerts et al. (2023), for several reasons. First, only ∼12% of the SPB candidates in that paper have a DR3 radial velocity (V_Rad) and, when available, the associated V_Rad uncertainties are quite large, suggesting possible problems in their RVS spectra. Moreover, most of the SPB candidates are too faint to have a high enough S/N to be analysed meaningfully with GSP-Spec. Finally, we remind that GSP-Spec was initially constructed to achieve proper parametrisation of slowly rotating rather cool stars. In particular, the reference training grid does not contain model spectra for stars hotter than T_eff ∼ 8000 K and the lines identified in the stellar spectra are assumed not to be too broadened. These restrictions led to the rejection of A-type or hotter stars and/or to lower quality parametrisation for the fastest rotators, as indicated by specific quality flags (and even rejection for the extreme cases). This is confirmed by examining the few SPB stars with available GSP-Spec parameters: most of them are indeed flagged as doubtful. We therefore postpone the spectroscopic study of these SPB stars until the next Gaia Data Releases in which the spectral parametrisation of hotter and/or fast rotators, such as many of the γ Dor and SPB pulsators (see Aerts et al. 2019; Li et al. 2020; Pedersen et al. 2021; Aerts 2021, for summaries), will be optimised.

This article is structured as follows. In Sect. 2, we present the sample of γ Dor pulsators with available Gaia spectroscopic data and we study their spatial distribution, kinematics, and orbital properties in the Milky Way. This led to the definition of a sub-sample of stars with very-high probability of being genuine γ Dor stars belonging to the Galactic thin disc. Subsequently, we present in Sect. 3 the physical parameters of the γ Dor candidates analysed by the GSP-Spec module and discuss their properties in Sect. 4. Our conclusions are summarised in Sect. 5.

2. The Gaia/DR3 spectroscopic sample of γ Dor stars

Among the 11 636 γ Dor pulsator candidates presented in Aerts et al. (2023), 4383 stars (38%) have a published Gaia/DR3 radial velocity (Katz et al. 2023) and only 650 of them are found in the GSP-Spec catalogue. These numbers can be explained by taking into account that (i) 58% of these pulsating stars are too faint to have an S/N spectrum high enough to estimate their V_Rad and/or to be parametrised by GSP-Spec (G ≲ 13.5 mag for this later case) and (ii) many of them are too hot, preventing their GSP-Spec parametrisation (see the limitation caused by the reference grid hot boundary, described in the introduction). In the following, we discuss the Galactic properties of these candidate stars, derived from their available distances and V_Rad, in order to define a sub-sample of bona-fide γ Dor stars based only on kinematic and dynamical criteria.

2.1. Spatial distribution, kinematics, and Galactic orbits

We derived the spatial (Cartesian coordinates) and kinematic properties of all the 4383 γ Dor candidate stars from their Gaia coordinates, proper motions and V_Rad, adopting the distances of Bailer-Jones et al. (2021). Their Galactic orbital properties (eccentricity) were computed as described in Palicio et al. (2023) using the Solar Galactic constants presented in Gaia Collaboration (2023b). For the Local Standard of Rest (LSR) velocity of the Sun, we also adopted V_LSR = 238.5 km s⁻¹.

Among all these stars with Galactic data, 560 were parametrised by GSP-Spec. We note that the parametrisation is available for 90 more stars (see Sect. 3) because GSP-Spec parametrised some spectra whose V_Rad was finally not published within Gaia/DR3, and hence their Galactic properties were not computed.

The following quality selections were then applied to define a sub-sample of 2721 stars (405 of them with GSP-Spec parameters) with high-quality Galactic parameters. (1) The best astrometric data were selected thanks to the ruwe parameter (ruwe < 1.4) and the identification of the non-spurious solutions (fidelity_v2 > 0.5, Rybizki et al. 2022). This filtered out 521 stars. (2) We then rejected 152 stars with a distance uncertainty larger than 10%. (3) The V_Rad determination of several of these γ Dor candidates was found to be of poor quality, mostly because of the low S/N of their spectra. We therefore disregarded 1258 stars with a relative V_Rad error larger than 50%.

The Galactic location of this high-quality sub-sample of 2721 γ Dor candidates with astrometric and V_Rad information is presented in Fig. 1. The colour-code of the middle and right panels represents their rotational velocity in the Galactic plane (V_ϕ, whose typical value for thin disc stars in the Solar vicinity is around ∼240 km s⁻¹). In the following, we adopt for this last quantity the velocity of the LSR at the Sun’s position. The location of the same stars in a Toomre diagram, colour-coded with their Galactic orbit eccentricity is shown in Fig. 2. The regions enclosed by the circular dotted lines denote those populated by stars with thin disc kinematics, that is, typical total velocity found within ±40–50 km s⁻¹ around the LSR value (see, for instance, Gaia Collaboration 2023b, for more details). From these figures, it is clear that most stars have the expected thin disc kinematics, although some γ Dor candidates actually belong to the Galactic thick disc or halo (low V_ϕ-values, high Galactic latitudes, large eccentricities or total velocities). Among those with a total velocity larger than ±50 km s⁻¹ compared with the LSR value (i.e. velocities too large to belong to the thin disc, first dotted line in Fig. 2), very few have published GSP-Spec atmospheric parameters. Their spectra S/N are indeed low (around 25), leading to rather large V_Rad uncertainty. Moreover, their rotational rate is high, and their metallicity could be too low for thin disc stars, but more importantly, the associated [M/H] uncertainties are very large. All of this reveals that their spectra were probably not properly analysed by GSP-Spec. In any case, the fact of not belonging kinematically to the thin disc is contradictory to our understanding of the evolutionary stage of this class of variable stars, and they were later rejected from the studied sample (see below). In contrast, most of the brightest candidates with GSP-Spec parameters (i.e. those with the highest S/N spectra) were found closer to the Galactic plane and with kinematics and orbital properties typical of thin disc stars (circular orbits, V_ϕ ∼ 215–260 km s⁻¹ and/or abs(V_Tot − V_LSR)≲25 km s⁻¹.

Fig. 1. Galactic location of the 2721 γ Dor candidates with high-quality astrometric and radial velocity data. The Solar position is indicated by the dotted lines. The left panel is a density plot of the whole sample and the colour-coding in the middle and right panels shows their Galactic rotational velocity (Vϕ). The right panel illustrates the location of the stars parametrised by GSP-Spec. The domain of shown (R, Z) values in each panel decreases from left to right.

Fig. 2. Toomre diagrams of the 2721 γ Dor candidates with high-quality astrometric and radial velocity data, colour-coded with their orbital eccentricity. The upper and lower panels show the whole sample and a closer view of the sub-sample of stars parametrised by GSP-Spec, respectively. The dotted lines correspond to a total velocity equal to 50 and 100 km s−1 with respect to the LSR value.

2.2. γ Dor candidates belonging to the thin disc

γ Dor stars are known to be late A-type to early F-type stars, located on the main-sequence below the classical instability strip. As already mentioned, they have masses between ∼1.3 and ∼1.9 M_⊙ (see, for instance, Mombarg et al. 2019; Ouazzani et al. 2019; Fritzewski et al. 2024) and therefore have rather young ages. We refer to Mombarg et al. (2021, see their Fig. 17) who deduced the ages of the 37 best-characterised γ Dor stars from asteroseismic modelling of their identified oscillation modes and found ages ranging from ∼0.15 up to 2 Gyr. Similarly, Fritzewski et al. (2024) also report asteroseismic ages for 490 γ Dor that are always smaller than 3.0 Gyr, their mean age being close to 1.5 Gyr with a dispersion of 0.5 Gyr. On the contrary, thick disc stars are older than ∼8 Gyr (see, for instance, Hayden et al. 2017; Santos-Peral et al. 2021; Xiang & Rix 2022), and halo stars are even older. Therefore, most of the Galactic γ Dor stars are expected to belong to the thin disc of the Milky Way.

One can thus use the above-described kinematic and orbital properties of the candidates from Aerts et al. (2023) to select those with the highest probability of being bona-fide γ Dor pulsators. This is an entirely complementary and independent selection to the one based on the DR3 or TESS light curves. We therefore selected all the above candidates with a high probability of belonging to the thin disc, that is, those having close to circular orbits (eccentricity lower than 0.2) and (V_Tot − V_LSR) < 25 km s⁻¹. These kinematic criteria¹ led to the selection of 2245 γ Dor candidates belonging to the Galactic thin disc (i.e. 83% of the stars with high-quality Galactic parameters), 385 of them having GSP-Spec parameters. This sub-sample, called the ‘Thin Disc’ sample hereafter, is discussed below. We therefore concluded that most stars of the initial sample are thin disc members.

3. Physical parameters of the γ Dor pulsators

All the 650 γ Dor pulsators with parameters from GSP-Spec have a published effective temperature (T_eff) and a surface gravity (log(g)). In addition, one can also access their global metallicity ([M/H]) and abundances in α-elements with respect to iron ([α/Fe]) for 602 and 595 of them, respectively. We remind that, within GSP-Spec, [M/H] is estimated from all the available atomic lines in the RVS spectra and is a good proxy of [Fe/H]. Similarly, α-element abundances are derived from all the available lines of any α-elements. The [α/Fe] abundance ratio is, however, strongly dominated by the huge Ca II infrared lines that are present in the Gaia/RVS wavelength domain and is thus strongly correlated to [Ca/Fe]. Moreover, because of the complex analysis of these rather hot and, usually, fast-rotating stars, very few other atomic lines are present in their RVS spectra. Therefore, individual chemical abundances were derived for only very few tens of them. Only 20 stars have an estimate of their [Ca/Fe] and the number of stars with other published chemical abundances is even lower. As a consequence, we only consider hereafter the [α/Fe] abundance ratio for this γ Dor sample. We calibrated all the above-mentioned atmospheric parameters and abundances as a function of T_eff, adopting the prescriptions recommended in Recio-Blanco et al. (2023, 2024). In addition, we emphasize that associated with the GSP-Spec parameters, there are several quality flags (flags_gspspec) that have to be considered to assess the quality of this parametrisation.

We also adopted other parameters derived from the analysis of the RVS spectra. For example, we partially used the Gaia/DR3 spectral type provided by the spectraltype_esphs parameters. In addition, 409 of the GSP-Specγ Dor candidates have a published line-broadening measurement (vbroad, related to their rotational velocity, see Frémat et al. 2023), confirming the fast-rotating nature for most of them. We refer to Aerts et al. (2023) for a detailed study of the Gaiavbroad properties of the g-mode pulsators. Even if vbroad was not published in the end for most stars in Gaia/DR3, GSP-Spec has published three quality flags (vbroadT, vbroadG, vbroadM, see Table 2 in Recio-Blanco et al. 2023) that depend directly on vbroad (internally delivered within DPAC for the spectra analysis). The possible biases in the GSP-Spec parameters that could be induced by rotational line-broadening can therefore be explored by future users thanks to these three vbroadTGM flags. Contrary to vbroad, these flags are available for the whole sample and are used later in this work to complement the rotational broadening information, when necessary. In the following and for convenience, all of these line-broadening quantities are referred to as ‘rotational velocities’.

Finally, we note that the Gaia/DR3 spectroscopic parameters were derived by the GSP-Spec module by assuming that the rotation rate of the analysed stars are rather low. Therefore, highly rotating stars were rejected. Depending on the stellar type, vbroad limitations are around ∼30–40 km s⁻¹, and the parametrisation quality degrades quickly above ∼25 km s⁻¹. Because of this parametrisation limitation, we warn that the γ Dor candidates parametrised by GSP-Spec thus have lower rotational velocities than typical values for these variable stars. Indeed, the vbroad distribution of our stars has a mean value around ∼25 km s⁻¹ is associated with a standard deviation equal to ∼10 km s⁻¹, and has a maximum value of 58 km s⁻¹. As a comparison, these stars are known to rotate at 40–100 km s⁻¹ and some can reach up to ∼150 km s⁻¹, see for instance Gebruers et al. (2021), Aerts et al. (2023). This rotational velocity of the stellar atmosphere is related to the internal rotation rate, as discussed for instance by Li et al. (2020). As a consequence, we are conscious that our sample is biased towards γ Dor with a rather low rotation rate, and is therefore not fully representative of these specific class of variable stars.

3.1. Stellar luminosities and radii:

To complement the spectral parametrisation, we computed the luminosity (L_⋆) and radius (R_⋆) for each star. For that purpose, we first estimated the extinction E(B_P − R_p) in the Gaia bands by subtracting the observed (B_P − R_p) colour from a theoretical one. The latter was calculated from the GSP-SpecT_eff, log(g) and [M/H], inverting the Casagrande et al. (2021) relation that predicts stellar colour from atmospheric parameters (see also Sect. 10.2 in Recio-Blanco et al. 2023). This procedure did not converge for a few stars. Hence, these were rejected hereafter. We then estimated the coefficients k_TGMA = A_G/E(B_P − R_p), A_G being the absorption in the G-band. These coefficients depend on the four atmospheric parameters and have been estimated thanks to the tables provided with the Gaia stellar parameters². The value of A_G allows one to deduce the absolute magnitude in the G-band from the Gaia DR3 G-magnitude and the Bailer-Jones et al. (2021) geometric distances. We derived the luminosities, adopting the bolometric corrections (BC) from Casagrande & VandenBerg (2018). We note that the considered relations to estimate k_TGMA and BC do not depend on [α/Fe]. When available, we adopted the relation of Salaris et al. (1993) to include the α-element content into the global metallicity. Finally, thanks to the GSP-SpecT_eff, we directly obtained the stellar radius. The quality of these radii simply computed from Gaia photometry, distances and spectroscopic parameters is excellent. We refer to de Laverny et al. (in prep.) and Recio-Blanco et al. (2024) for a detailed comparison with interferometric and/or asteroseismic radii, that confirmed the high-quality of our R_⋆ estimates. One could also obtain the stellar mass (M_⋆) from the surface gravities, but we favoured fixing this quantity thanks to the known typical masses of γ Dor (see the discussion on the adopted log(g) below). For all of these parameters, the uncertainties were estimated by performing 1000 Monte Carlo realisations, propagating the uncertainties on each atmospheric parameter (that reflect the S/N spectra), distance, and Gaia magnitudes.

3.2. Effective temperatures:

Since γ Dor are known to be early-F spectral type stars and in order to define a sub-sample of high-quality parametrised stars, we first checked the types provided by the Gaia/DR3. Among the 650 candidates parametrised by GSP-Spec, 562, 82, and six were found by the DPAC/ESP-HS module to belong to the ‘F’, ‘A’ or ‘B’ spectral types, respectively. The GSP-SpecT_eff also confirmed the too hot temperature with respect to typical γ Dor stars of a few other stars (see, for instance, Van Reeth et al. 2015; Aerts et al. 2023), and hence they were filtered out. Moreover, we also found that ∼6% of the candidates have a GSP-SpecT_eff much cooler and not compatible with an early spectral type. All the spectra of these outliers suffer from large V_Rad uncertainties and/or very large rotational velocities and/or low S/N that may lead to an erroneous parametrisation. Therefore, all these too cool stars were also rejected. The remaining 598 spectroscopic candidates have a Gaia (B_p − R_p)₀ colour (corrected from extinction by us) fully compatible with their T_eff. The median of their (B_p − R_p)₀ colour is 0.5, and the associated dispersion is found to be extremely small (0.04 mag), confirming their early F-type nature. Hereafter, we refer to these 598 stars as the ‘F-type’ GSP-Spec sample.

3.3. Stellar surface gravities:

By knowing both the stellar luminosity and the GSP-Spec effective temperature and adopting a γ Dor typical mass, the surface gravity can simply be estimated for almost all the GSP-Spec pulsators (called log(g)_Lum, hereafter). Practically, we randomly chose the mass of each star within the γ Dor mass range 1.3–1.9 M_⊙ (see, for instance, Fritzewski et al. 2024), assuming an uniform distribution. The associated mass uncertainty was fixed to half of this range (±0.3 M_⊙). We note that varying the mass over the entire range covered by genuine γ Dor stars changes our log(g)_Lum estimates by about 0.1 dex and thus does not affect our conclusions. The log(g)_Lum uncertainties were computed from 1000 Monte Carlo realisations, propagating the uncertainties on L_⋆, T_eff and M_⋆. We finally point out that 30 stars (among the 650) have no log(g)_Lum because no extinction was available for them (see above).

We found that most of these log(g)_Lum are in very good agreement with the GSP-Spec log(g) for stars with a rather low rotational velocity (vbroad ≲ 15–20 km s⁻¹) or for stars with slightly larger rotational rate but with high S/N spectra (≳100). However, larger discrepancies between the two surface gravity estimates were found for the highest rotators and/or stars with low S/N spectra (mean difference of log(g_{GSP − Spec}/g_Lum)∼0.35 dex, with a standard deviation of 0.4 dex). This results from the DR3 GSP-Spec pipeline that is optimised for non-rotating stars and some parameter biases could exist for stars with large rotation rates³

In the following and in order to avoid surface gravities potentially affected by the stellar rotation, we adopt these log(g)_Lum and their associated uncertainties. Thanks to this procedure, we obtained a sub-sample of about 600 γ Dor pulsators with rather accurate surface gravity values (found in the range from ∼3.5 to ∼4.5) and effective temperatures typical of γ Dor stars, according to Van Reeth et al. (2015), Fritzewski et al. (2024). We emphasize that all the above-derived quantities do not rely on any stellar isochrones, but only on Gaia astrometric and photometric data and RVS spectra.

3.4. Global metallicities:

We simply adopted the GSP-Spec calibrated metallicities for all the γ Dor pulsators with a T_eff and a log(g)_Lum, as defined above.

3.5. Parameter uncertainties:

The RVS spectra of all of these γ Dor pulsators belonging to the F-type GSP-Spec sample have rather low S/N. The median S/N is 34 and only 9% of the spectra have a S/N greater than 100. The resulting median uncertainties⁴ on (T_eff, log(g), [M/H], L_⋆, R_⋆) are therefore rather high (182 K, 0.13 dex, 0.21 dex, 0.13 L_⊙, 0.11 R_⊙) albeit with a dispersion that is still reasonable of (62 K, 0.02 dex, 0.07 dex, 0.07 L_⊙, 0.04 R_⊙).

4. Properties of the Gaiaγ Dor pulsators

The distribution of the main stellar parameters derived above T_eff, log(g) (both derived from the spectrum analysis and from the luminosity), [M/H], L_⋆, and R_⋆ are shown in Fig. 3. In the different panels of this figure, we considered the sub-sample of 371 slowly rotating stars belonging to both the F-type GSP-Spec and Thin Disc samples, that is those having effective temperatures typical of γ Dor stars and having typical thin disc kinematics as defined in Sect. 2.2. We call this sub-sample ‘F-type Thin Disc’ (FTD) hereafter. It is shown as light-blue histograms in Fig. 3 and should contain good Gaiaγ Dor candidate stars.

Fig. 3. Distribution of the GSP-Specγ Dor effective temperatures, luminosities, surface gravities, global metallicities and stellar radii for the FTD (light-blue histograms) and the HQ (in dark-blue) sub-samples (see text for details). The red, black and green distributions are for the parameters derived by Gebruers et al. (2021), Aerts et al. (2023) and Fritzewski et al. (2024), respectively. These distributions have been artificially scaled since their statistics differ from ours. The term log(g)GSP − spec refers to the stellar surface gravities derived from the spectra analysis by GSP-Spec whereas log(g)Lum is the estimation from the stellar luminosity, assuming typical γ Dor masses. The number of stars in each panel may differ since not all the parameters are available for every source.

For comparison purposes, we have over-plotted in Fig. 3 the parameter distributions derived by three recent studies that provide parameters for large numbers of γ Dor stars. First, Gebruers et al. (2021, red-line distributions in Fig. 3) report T_eff, log(g) and [M/H] estimated from the analysis of high-resolution spectra (R ∼ 85 000) for 77 bona-fide γ Dor stars with asteroseismic modelling from Kepler observations. None of these stars were found in the F-type GSP-Spec sample. Secondly, Fritzewski et al. (2024) published luminosities and asteroseismic radii of 490 γ Dor derived from Gaia and Kepler observations. Their parameter distributions are shown as green lines in Fig. 3. There is only one star of Fritzewski et al. (2024) that was parametrised by GSP-Spec, but it is not included in the F-type GSP-Spec sample because of its too large rotational broadening and associated low-quality flags. Finally, we also show the T_eff, log(g), L_⋆ and R_⋆ distributions of the γ Dor candidates of Aerts et al. (2023, black histograms) as another comparison. Some of their stars suffer from rather large uncertainties on their parameters and we therefore filtered out the stars with T_eff and log(g) errors larger than 100 K and 0.2 dex, respectively, before constructing the black histograms. Moreover, stars with relative uncertainties in L_⋆ and R_⋆ larger than 25% and 50% were also rejected.

Some other γ Dor stars were also analysed by high-resolution spectroscopy, for instance in Tkachenko et al. (2012), Niemczura et al. (2015, 2017). The spectra collected by these authors have a resolution of 32 000, 85 000, and 25 000/45 000, respectively, but their samples are much smaller than ours. None of our stars are found in these samples, but we discuss below the atmospheric parameter properties derived by these studies with respect to those of GSP-Spec (see also the discussion about Fig. 4).

Fig. 4. Luminosity versus effective temperature diagram of the HQ sub-sample of γ Dor stars, colour-coded with their global metallicity. Error bars on L⋆ and Teff are shown in light-grey and the dotted lines represent the iso-radius relations.

Regarding the effective temperatures, it can be seen in Fig. 3 that our T_eff distribution is nearly compatible with the estimates reported in the literature. We do report a large number of cooler γ Dor stars than Gebruers et al. (2021) and Aerts et al. (2023). We checked that most of these cooler stars had bad quality GSP-SpecvbroadTGM flags, meaning that their parametrisation was not optimal because of the broadened lines in their spectra. Moreover, we remind that most RVS γ Dor spectra are of rather low quality, leading to large T_eff uncertainties. As a consequence, if we select only the 151 stars belonging to the F-type GSP-Spec sample and with a T_eff error less than 200 K (the mean S/N ratio of the selected spectra is then around 60) and whose three vbroadTGM flags are strictly smaller than two, we then obtain a T_eff distribution fully compatible with the two comparison ones. This High-Quality (HQ) sub-sample of γ Dor stars is shown as dark-blue histograms in Fig. 3. Our T_eff distribution is also fully compatible with the ones derived from high-resolution spectroscopy (see above-cited references), if one excludes binaries and/or hybrid stars, as their T_eff are always found in the ∼6700–8000 K range.

Our stellar luminosity distributions (FTD and HQ sub-samples) peak around 5 L_⊙, with few bright stars that go up to ∼25 L_⊙. They are close to the distribution of the two comparison samples (similar peak in L_⋆) although Aerts et al. (2023) and Fritzewski et al. (2024) report a larger number of more luminous stars. We have checked that considering only stars with good Gaia astrometric ruwe parameters (ruwe < 1.4) in the different comparison samples (about 10% stars would be rejected) does not modify these distributions. Moreover, we remind that these two other studies computed their stellar luminosity by adopting a Gaia/DR3 interstellar reddening that could differ from our own absorption estimate. Since we did not find any specific differences between these two reddening flavours, the lack of high-luminosity stars in our sample could be due to selection bias effects, namely a lack of too hot and, hence more luminous γ Dor stars, not parametrised by GSP-Spec; and/or a lack of low surface gravity stars, hence with a high radius and luminosity, as seen in the log(g) and radius panels of Fig. 4.

The GSP-Spec FTD and Gebruers et al. (2021) spectroscopic surface gravities are in good agreement (left panel, second row in Fig. 3). γ Dor surface gravities derived by other analysis of high-resolution spectra studies also agree, although we note that the samples of Niemczura et al. (2015, 2017) have log(g) within ∼3.6–4.0; that is, their surface gravities are slightly smaller than ours and those of Gebruers et al. (2021), which reach up to log(g) = 4.4. However, more numerous lower gravity values are seen in Fig. 3 compared to Aerts et al. (2023). Such stars with a surface gravity lower than ∼3.5 should have left the main sequence (or are close to this evolutionary stage). The agreement between the HQ spectroscopic gravities and those of Aerts et al. (2023), however, is excellent, confirming again that low-quality and line-broadened spectra are more difficult to parametrise. Moreover, the results from Aerts et al. (2023) are also in good agreement with our surface gravities estimated from the stellar luminosity (log(g)_Lum shown in the right panel, second row) and assuming γ Dor masses in the range 1.3–1.9 M_⊙. This is quite normal since the Aerts et al. (2023) log(g) values were estimated by the DPAC GSP-phot module (Andrae et al. 2023) from the luminosity, T_eff and stellar isochrones, which except for the use of isochrones is a rather similar method to ours. From these different comparisons, we conclude that γ Dor surface gravities derived from the stellar luminosity and assuming typical γ Dor masses should be preferred over the spectroscopic ones because of the difficulty involved in analysing spectra of fast-rotating stars, except if one considers the HQ stars for which the spectroscopic gravities are excellent.

As for the γ Dor stars’ mean metallicities, our FTD sample covers a larger range in [M/H] than the comparison samples. More importantly, it reveals an excess of low-[M/H] stars with respect to the spectroscopic sample of Gebruers et al. (2021). The latter peaks around −0.2 dex, whereas ours peaks at about −0.7 dex. We note that the confirmed γ Dor of Tkachenko et al. (2012) with parameters derived from high-resolution spectra have [M/H] that cover a distribution close to the one of Gebruers et al. (2021). We remind that such low-metallicities are not expected to be numerous for thin disc stars. This shift towards lower metallicities could again be partially explained by the stellar rotation and/or low-quality RVS spectra, as it might induce a bias in the parametrisation. Again, rejecting such complex spectra and considering HQ sub-sample stars with a [M/H] uncertainty less than 0.25 dex leads to a metallicity distribution that is more compatible with Gebruers et al. (2021). However, we still have more stars with higher and lower metallicities than them by about ±0.5 dex. They could be present since our sample has a larger spatial coverage in the Milky Way.

Finally, the compared stellar radius FTD and HQ distributions are in very good agreement with the asteroseismic radii determined by Fritzewski et al. (2024), which are expected to be the most accurate. The spectroscopic radii, whatever the sample considered, are indeed very well representative of those expected for γ Dor stars. Our radius median value is close to 1.7 R_⊙ and 90% of the sample is found in the range 1.35–2.35 R_⊙. The largest stars in our sample have R ∼ 3.5–4 R_⊙, so they must be rather close to leaving the main-sequence, which is confirmed by their effective temperatures and high luminosities (T_eff ≃ 7000 K and L ≃ 25 L_⊙).

In summary, our reported T_eff, log(g), [M/H], L_⋆ and R_⋆ are in very good agreement with recent literature values, in particular when considering the High-Quality sub-sample defined by selecting high-quality spectra or low-rotating stars. Furthermore, our L_⋆, log(g)_Lum, and R_⋆ values are trustworthy, even without considering the membership of the Galactic thin disc as an extra good measure of the star being a genuine γ Dor pulsator and/or spectra properties.

Moreover, since we have shown that the agreement on the effective temperature and mean metallicity is improved by selecting high-quality S/N stellar spectra or stars without large rotational broadening velocities, we show these HQ γ Dor stars in a luminosity-effective temperature diagram (Fig. 4), colour-coded with their metallicity. This figure is similar to those found in the literature, as for example in Niemczura et al. (2015, 2017), Tkachenko et al. (2012), Gebruers et al. (2021). Excluding binaries and hybrid stars, all of our γ Dor stars are very well concentrated in the same small region of the L-T_eff diagram (or log(g)-T_eff, in some of the above-cited works). We note that there are no γ Dor stars in this figure with T_eff hotter than ∼7750 K⁵. This bias is caused by the GSP-Spec parametrisation that was optimised for FGKM type stars (we recall that the reference grid is based on spectra models cooler than 8000 K). We therefore cannot exclude the possibility that hotter γ Dor could exist but were rejected during the parameterisation (this will be updated for Gaia/DR4). Finally, it can be seen in Fig. 4 that more metal-rich γ Dor are found at higher T_eff, whatever their luminosity is. This could be partly due to some possible parametrisation biases. For instance, metal-poor hot star spectra show very few lines and are thus more difficult to parameterise, particularly when their rotation rate is high, likely explaining the absence of such stars in the present sample. But this could also be real and could be a signature of the different evolution of stars with slightly different masses and metallicities. For instance, by exploring BaSTI evolutionary tracks (Pietrinferni et al. 2021), it can be seen that metal-rich γ Dor stars with masses around 1.6–1.9 M_⊙ appear hotter than lower mass (∼1.3 M_⊙) more metal-poor stars. More specifically, we have estimated that a difference of ∼0.3 M_⊙ implies a shift in T_eff of similar amplitude as a difference of ∼0.7 dex in metallicity. This corresponds rather well to what is seen in Fig. 4.

The same HQ sub-sample but with stars with an [α/Fe] uncertainty greater than 0.15 dex filtered out is plotted in an [α/Fe] versus [M/H] diagram in Fig. 5. Such a filtering again reduces the number of stars but reveals more accurate chemo-physical properties of γ Dor pulsators. We remind that only the global [α/Fe] abundances (which is a good indicator of [Ca/Fe] for RVS spectra) of these γ Dor stars parametrised by GaiaGSP-Spec are available in the DR3 catalogue because of too low statistics of individual chemical abundances (see above). Fig. 5 confirms that the selected sample has chemical properties consistent with the Galactic disc population, that is a constant decrease of [α/Fe] with the metallicity for [M/H] >  − 1.0 dex. Moreover, since the membership of most of these stars to the thin disc was based purely on kinematics and dynamical criteria, such an [α/Fe] versus [M/H] trend is an independent proof that the chemo-physical properties of the γ Dor pulsators can be safely adopted.

Fig. 5. Distribution of the HQ sub-sample of γ Dor stars in the [α/Fe] versus [M/H] plane. Only stars with an uncertainty smaller than 0.15 dex in [α/Fe] are shown. The Solar location is indicated by the intersection of the dotted lines.

Thus, the GSP-Spec/DR3 analysis of γ Dor pulsators provides useful physical and/or chemical parameters once an optimised filtering of the poorly parametrised spectra (low S/N or too fast rotating stars) is performed. The parameters of these HQ γ Dor stars are provided in an electronic table whose content is presented in Table 1.

5. Conclusions

We have studied the Gaia/DR3 spectroscopic parameters derived from the analysis of the RVS spectra for the large sample of γ Dor candidate pulsators composed by Aerts et al. (2023) and confirmed in Hey & Aerts (2024). About 38% of these stars have a published radial velocity and ∼6% of them were actually analysed by the GSP-Spec module in charge of analysing their Gaia spectra.

Thanks to the available V_Rad and astrometric Gaia information, we have been able to compute kinematics and orbital information for all these stars. This allowed us to identify that 2245 of them (i.e. most of the candidates with high-quality kinematics) belong to the thin disc of the Milky Way, which is expected since these g-mode pulsator should have typical ages lower than 2–3 Gyr.

We then computed their luminosity and stellar radius from Gaia astrometric and photometric data, adopting the GSP-Spec effective temperature and without considering any stellar evolutionary models or isochrone priors. A comparison with recently published values of well-studied γ Dor stars revealed that the derived luminosities, stellar surface gravities derived from L_⋆ and assuming typical γ Dor masses as well as the stellar radii are of high quality. Moreover, strict filtering that reject stars with a large T_eff uncertainty (caused by too low S/N RVS spectra) or a high rotational velocity led to having a sample of pulsators with the best derived parameters, including T_eff, [M/H], and [α/Fe]. All of these observables were found to be fully consistent with typical values of genuine slowly rotating γ Dor pulsators. Indeed, the GSP-Spec HQ γ Dor stars have effective temperatures between ∼6500 and ∼7800 K and surface gravities around 4.2. Their luminosities and stellar radii peak at ∼5 L_⊙ and ∼1.7 L_⊙, whereas their metallicity distribution is centred close to the Solar value, covering the range [−0.5, +0.5] dex. Their [α/Fe] properties are consistent with the chemical properties of the Galactic disc population. We note that the final number of parametrised stars is smaller compared to the initial sample because of the low S/N spectra of many of them, together with the fact that most of them are fast rotators, which the GSP-Spec analysis pipeline of the Gaia/DR3 was not optimised for. Nevertheless, the number of newly spectroscopically parametrised γ Dor presented in this work is about a factor of two larger than in previous studies.

Finally, it can be concluded that the GSP-Spec analysis of γ Dor stars provides a significant added value to the study of these g-mode pulsators, delivering their physical and chemical properties. This will be even more important with future Gaia data releases, in which the RVS spectra S/N values and the number of analysed stars will be significantly increased. Indeed, it is expected that the S/N increase between DR3 spectra and those released in DR4 and DR5 will correspond to a factor of $\sqrt{2}$ and two, respectively. Moreover, it is also anticipated that the analysis of fast-rotating stars by the GSP-Spec module will be improved by considering reference grids of synthetic spectra representative of a wide variety of stellar rotational values, contrary to the present study that is biased towards slowly rotating γ Dor. Finally, hot star spectra will also be considered for reference, allowing for better parametrisation for stars with high T_eff. These anticipated improved analyses of g-mode pulsators performed by the GSP-Spec module will therefore be of prime interest as input for asteroseismology of such stars. They will indeed allow for a much larger and thus more statistically significant number of bona-fide γ Dor stars with physical and chemical properties to be defined. Furthermore, they will also enable the study of the spectroscopic parameters of hotter g-mode pulsator, such as the SPB stars with the aim of improving their asteroseismic modelling (Pedersen et al. 2021).

Data availability

The parameters of HQ Dor stars are available at the CDS via anonymous ftp to cdsarc.cds.unistra.fr (130.79.128.5) or via https://cdsarc.cds.unistra.fr/viz-bin/cat/J/A+A/691/A182

Acknowledgments

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. This work has also made use of the SIMBAD database, operated at CDS, Strasbourg, France (Wenger et al. 2000), the IPython package (Pérez & Granger 2007), NumPy (Harris et al. 2020), Matplotlib (Hunter 2007), Pandas and TOPCAT (Taylor 2005). PdL and ARB acknowledge funding from the European Union’s Horizon 2020 research and innovation program under SPACE-H2020 grant agreement number 101004214: EXPLORE project. CA acknowledges funding from the European Research Council (ERC) under the Horizon Europe programme (Synergy Grant agreement number 101071505: 4D-STAR project). While partially funded by the European Union, views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. Finally, we are grateful to the anonymous referee for their constructive remarks.

References

All Tables

All Figures

Fig. 1. Galactic location of the 2721 γ Dor candidates with high-quality astrometric and radial velocity data. The Solar position is indicated by the dotted lines. The left panel is a density plot of the whole sample and the colour-coding in the middle and right panels shows their Galactic rotational velocity (Vϕ). The right panel illustrates the location of the stars parametrised by GSP-Spec. The domain of shown (R, Z) values in each panel decreases from left to right.

In the text

	Fig. 2. Toomre diagrams of the 2721 γ Dor candidates with high-quality astrometric and radial velocity data, colour-coded with their orbital eccentricity. The upper and lower panels show the whole sample and a closer view of the sub-sample of stars parametrised by GSP-Spec, respectively. The dotted lines correspond to a total velocity equal to 50 and 100 km s−1 with respect to the LSR value.
In the text

Fig. 3. Distribution of the GSP-Specγ Dor effective temperatures, luminosities, surface gravities, global metallicities and stellar radii for the FTD (light-blue histograms) and the HQ (in dark-blue) sub-samples (see text for details). The red, black and green distributions are for the parameters derived by Gebruers et al. (2021), Aerts et al. (2023) and Fritzewski et al. (2024), respectively. These distributions have been artificially scaled since their statistics differ from ours. The term log(g)GSP − spec refers to the stellar surface gravities derived from the spectra analysis by GSP-Spec whereas log(g)Lum is the estimation from the stellar luminosity, assuming typical γ Dor masses. The number of stars in each panel may differ since not all the parameters are available for every source.

In the text

	Fig. 4. Luminosity versus effective temperature diagram of the HQ sub-sample of γ Dor stars, colour-coded with their global metallicity. Error bars on L⋆ and Teff are shown in light-grey and the dotted lines represent the iso-radius relations.
In the text

	Fig. 5. Distribution of the HQ sub-sample of γ Dor stars in the [α/Fe] versus [M/H] plane. Only stars with an uncertainty smaller than 0.15 dex in [α/Fe] are shown. The Solar location is indicated by the intersection of the dotted lines.
In the text