Insights from Super-Metal-Rich Stars: Is the Milky Way bar young?

Super-metal-rich (SMR) stars, currently in the solar neighbourhood, are expected to originate only in the inner Galaxy and have definitely migrated. We aim at studying a large sample of SMR stars to provide constraints on the epoch of the bar formation and its impact on the MW disc stellar populations. We investigate a sample of 169,701 MSTO and SGB stars with 6D phase space information and high-quality stellar parameters coming from the hybrid-CNN analysis of the Gaia-DR3 RVS stars. We compute distances and ages using the StarHorse code with a mean precision of 1% and 11%, respectively. From these, 11,848 stars have metallicity ([Fe/H]) above 0.15 dex. We report a metallicity dependence of spatial distribution of stellar orbits shown by the bimodal distribution in the guiding radius at 6.9 and 7.9 kpc, first appearing at [Fe/H]~0.1 dex, becoming very pronounced at larger [Fe/H]. In addition, we've observed a trend where the most metal-rich stars, with [Fe/H]~0.4 dex, are predominantly old (9-12 Gyrs) but show a gradual decline in [Fe/H] with age, reaching around 0.25 dex at about 4 Gyrs ago, followed by a sharp drop around 3 Gyrs ago. Furthermore, our full dataset reveals a clear peak in the age-metallicity relationship during the same period, indicating a SF burst around 3-4 Gyrs ago with slightly sub-solar [Fe/H] and enhanced [$\alpha$/Fe]. We show the SMR stars are good tracers of the bar activity. We interpret the steep decrease in number of SMR stars at around 3 Gyr as the end of the bar formation epoch. In this scenario, the peak of bar activity also coincides with a peak in the SF activity in the disc. Although the SF burst around 3 Gyr ago has been reported previously, its origin was unclear. Here, we suggest the SF burst to have been triggered by the high bar activity, 3-4 Gyr ago. According to these results and interpretation, the MW bar could be young.


Introduction
Stars are the luminous storytellers of our Galactic saga.By studying the properties of these stars, their ages, chemical abundances and motions, we can trace back the history of our Galaxy (Pagel 1997;Matteucci 2001Matteucci , 2021;;Freeman & Bland-Hawthorn 2002).Among the different stellar populations in the Milky Way (MW), super-metal-rich (SMR1 ) stars are very interesting as they are expected to be formed only in the inner regions of our Galaxy out of materials enriched by previous generations of stars (Grenon 1972;Steinmetz & Mueller 1994;Trevisan et al. 2011;Casagrande et al. 2011;Kordopatis et al. 2015;Anders et al. 2017;Miglio et al. 2021).These SMR stars currently residing in the SNd have definitely migrated from the inner Galaxy (Feuillet et al. 2018;Chen et al. 2019;Dantas et al. 2023) and their study can inform us about the processes that bring them to the outer disc.Indeed, Queiroz et al. (2021) have shown the inner kpcs of the Galaxy to provide a large reservoir of highmetallicity stars (either as part of an inner-thin disc or on bar shape orbits), and these are primary good candidates for migration.
The Galactic bar is considered an important perturber for generating radial motion of stars and gas in the galactic disc (e.g.Sellwood 2014).According to N-body simulations, during the formation and the phase of strong bar activity, stars in the disc are significantly redistributed, with the highest probability of migration at the bar resonances (Minchev & Famaey 2010;Halle et al. 2015;Khoperskov et al. 2020a).However, an open question still remains on the epoch of MW bar formation.
Old to young formation times for the MW bar have been proposed in both observational and simulation studies conducted so far.Bovy et al. (2019) and Wylie et al. (2022), using red giant stars in the inner Galaxy with ages estimated using astroNN (Leung & Bovy 2019), proposed an older bar at least 7 to 8 Gyr old based on the mean and peak of the age distribution.Cole & Weinberg (2002), using Carbon stars that trace the bar, suggested that bar likely formed ∼3 Gyr ago.However, it is important to recognise that the bar is a cumulative entity, it will indiscriminately contain older stars and stars born during bar formation.So, care has to be taken to differentiate older stars from those formed during bar formation (e.g.see de Sá-Freitas et al. 2023a).Tepper-Garcia et al. (2021), using a tailored N-body model of barred MW, suggested that the bar formed 3-4 Gyrs ago.In context of external galaxies, de Sá-Freitas et al. (2023b) recently reported the discovery of young bars, formed 4.5 and 0.7 Gyr ago, for spiral galaxies NGC 289 and NGC 1566 using the SF history of nuclear discs.This suggests that some disc galaxies, with stellar mass comparable to MW, could settle on longer time-scales.
However, because SMR stars are rare, we still lack a statistically significant sample to constrain this process.This situation has radically changed thanks to third data release (DR3) of the ESA Gaia mission (Gaia Collaboration et al. 2023).Gaia-DR3 has provided about one million spectra from the Radial Velocity Spectrometer (RVS)2 of which only ∼178,000 stars have the good quality stellar parameters after the applying the recommended flags (Recio-Blanco et al. 2023).A large portion of the published RVS spectra are of low signal-to-noise (S/N; 15 -25 per pixel) and can be challenging for traditional spectroscopic pipelines.Guiglion et al. (2023, G23) using a hybrid Convolutional Neural-Network (CNN) method, reanalyzed the RVS sample to derive atmospheric parameters (T eff , log(g), and overall [M/H]) and chemical abundances ([Fe/H] and [α/M]), by supplementing extra information from Gaia magnitudes (Riello et al. 2021), parallaxes (Lindegren et al. 2021) and XP coefficients (De Angeli et al. 2023).The hybrid-CNN was trained with a high-quality training sample based on APOGEE DR17 (Abdurro'uf et al. 2022) labels and show precision and accuracy comparable to external data-sets such as GALAH and asteroseismology.Thanks to the novel method, G23 significantly improved the number of reliable targets that can be used for Galactic archaeology.The G23 catalog, has a large number of SMR stars, including a sample of MSTO and SGB stars, enabling the use of these traces to constrain the epoch of MW bar formation.
In this Letter, we explore the age-[Fe/H] (AMR) and the age-[α/Fe] (AAR) relationships along with the chemo-dynamics for a sample of MSTO and SGB stars selected from the RVS-CNN catalog of G23.In Sect.2, we describe our sample focusing on the methods used to obtain the stellar ages and kinematics.In Sect. 3 we present our results and in Sect. 4 we present the main conclusions.

Data
We obtain a sample of stars, with −0.7 < [Fe/H]< 0.5 including a large set of the SMR stars, from the catalog of G23.We selected MSTO and SGB stars (see Queiroz et al. 2023) after the appli-cation of the G23 recommended flags.We keep only the MSTO-SGB stars as the stellar ages from isochrone fitting methods are most reliable for these evolutionary stages (e.g.Soderblom 2010).We computed the distances and stellar ages with the StarHorse Bayesian isochrone-fitting method (Queiroz et al. 2018(Queiroz et al. , 2023) ) and integrate the orbits of the stars using Galpy (Bovy 2015).For details on these computations see Appendix A.
We selected stars with relative age uncertainty less than 20%, distance uncertainty less than 5% and extinction uncertainty less than 0.2 Mag.We are left with a sample with mean uncertainty of 11% for age and 1% for distance due to very low parallax errors in the extended SNd thanks to Gaia.We further remove any stars with poor astrometric solutions by limiting RUWE < 1.4 and also remove known variable stars by using Gaia flag 'phot_variable_flag' 'VARIABLE' (see Gaia Collaboration et al. 2023).In Fig. 1, we present our sample properties.Although the stars are widely distributed in Z vs R space, the number of stars decreases as we move away from SNd, as expected for a sample of MSTO+SGB stars (see Queiroz et al. 2023).Therefore the sample is essentially dominated by thin disc stars (low-[α/Fe] population).

Super-metal-rich stars as tracers of MW bar activity
Figure 2 shows the distribution of guiding radius (R g ), in bins of metallicity, for stars in the range −0.3<[Fe/H]<0.5.The [Fe/H] values increase as we move from panel a) to panel i).These histograms show that a clear bimodality in the distribution of R g Fig. 2. Distribution of the guiding radii (R g ) in bins of metallicity ([Fe/H]).The [Fe/H] values increase from top left to bottom right, i.e. from panel a) to panel i) (number of stars and the [Fe/H] range shown).The two dotted lines at 6.9 and 7.9 kpc represent the two peaks of the R g distribution.appears as we move from sub-solar [Fe/H] to SMR stars.The bimodality visibly appear first in the bin 0.10<[Fe/H]<0.15,as seen in panel (d), and becomes stronger for the higher [Fe/H] bins.An inner and an outer peak is seen at 6.9 and 7.9 kpc, for all super-solar bins.We validate this finding with an external sample of red-giant stars in Appendix D. We note that the two peaks are not equally populated, with higher number of stars near the solar radius, and attribute this to the Gaia selection function.
Unlike SMR stars, sub-solar to solar [Fe/H] stars are equally plausible to be formed in the SNd as well as the inner and outer galaxy, so their actual birth location can be uncertain due to radial migration (e.g.Minchev et al. 2013, 2014 andreferences therein).We also know the innermost regions of the MW to be populated of very metal-rich stars (e.g see Barbuy et al. 2018 and references therein).In particular, Queiroz et al. (2021), using APOGEE and Gaia surveys to study bar and bulge region, found a large cache of metal-rich stars in inner Galaxy.These metal-rich stars, found in cold to highly eccentric orbits, could be a source of origin for our local SMR stars.
What could be the mechanism that brought them here?The SMR stars in our sample, most of them in cold orbits with 96% with ecc<0.3,distributed favourably at certain R g entices us to view their distribution in context of evolution and dynamical process in the Galaxy (See Appendix B for a comparison of velocity dispersion of metal-rich stars with the full sample.).The Galactic bar has been long considered an efficient source of radial migration (Combes & Sanders 1981;Pfenniger & Norman 1990).Monari et al. (2019).See Appendix F for test on how metal-rich stars are better suitable for probing bar activity.Therefore, the inner and outer R g peaks, observed for the SMR stars at 6.9 and 7.9 kpc, most probably correspond to the imprints of the Galactic bar.
From our observations (see Fig. 2) we conclude that the SMR stars, currently in the SNd, trace the signatures of bar activity.In the next subsection we employ these SMR stars, with good ages, to study the formation epoch of the MW bar.

Timing the bar formation epoch with the youngest super-metal-rich stars
In the left panel of Fig. 3 we present the stellar ages as a function of R g , for the 19 367 metal-rich stars for which bimodality is clearly seen in Fig. 2, i.e. for [Fe/H]>0.1 dex.The plot shows that the metal-rich stars have a wide range of ages with higher prevalence at older ages (11 to 6 Gyr).This hints to the fact that a large number of metal-rich stars were formed at early times in our Galaxy's history.The bimodal nature of R g , with peaks at 6.9 and 7.9 kpc, is clearly seen for all ages and the spread in R g is narrower for the younger stars.Although some of these older stars could have already dispersed to the outer galaxy at earlier times through various processes such as merger events (e.g.Helmi 2020), most of them are still in cold orbits.Interestingly, we see a near absence of metal-rich stars with ages younger than ∼2.7 Gyr old (see the dashed line in Fig. 3, also see Fig.  sample, see Figs. 4 and E.1.Therefore, the dearth of younger SMR stars could be related to the mechanism that brought these stars formed in the inner Galaxy to the SNd. In the right panel of Fig. 3 we present the time evolution of the upper envelope of the metallicity for our whole sample.The plot reveals that our Galaxy quickly attained the highest levels of metallicity enrichment, levelling around [Fe/H]=0.4 dex, already at ∼10 Gyr ago.Then starting at 8.5 Gyr ago we observe a slow decreases, at −0.04 dex/Gyr, up to 4 Gyr.Then for a period of 1 Gyr, from 4 to 3 Gyr ago, we see a plateau at [Fe/H]=0.25 dex, highlighted by an orange band.After 3 Gyrs we see a steep decline in the [Fe/H] envelope at a rate of −0.15 dex/Gyr (see Appendix D for a validation with an asteroseismic sample).This decline of the [Fe/H] envelope further confirms that the mechanism, that brought majority of these metal-rich stars to SNd, has ceased.We note here that Minchev et al. (2011) identified a radial migration mechanism, via bar and spiral structure interaction, that can migrate stars throughout the disc of MW-like galaxy in < 1 Gyr period.However, the migration induced by the bar is most effective at the epoch of bar growth and, therefore, stars in the inner Galaxy can be moved en masse to larger R g (e.g.Halle et al. 2015).
Considering MW bar growth and outward propagation of the resonances as the transport mechanism of these SMR stars to the SNd (Minchev & Famaey 2010;Khoperskov et al. 2020a;Iles et al. 2023), we can use the stellar ages of these youngest SMR stars to place a constraint on the epoch of intense bar growth following bar formation.It is worth noting that the MW bar, and inner Galaxy in general, when currently observed will contain stars of all ages from very old in higher number to very young, including those formed before, during and after the epoch of bar formation.However, only the youngest metal-rich stars, formed during the epoch of bar formation, can be confidently employed to constrain the age of bar.Based on this absence of younger SMR stars, we place the intense bar growth for the MW ending at 2.5 to 3 Gyrs ago.

Enhanced star formation triggered by the bar formation
Considering that intense bar growth for MW ended at ∼3 Gyrs ago (see previous Section), we can expect an epoch of bar enhanced SF succeeding this event (e.g.Baba & Kawata 2020).In this section, we explore SF in the Age-Metallicity and Age-Alpha Relationships (AMR and AAR) and look for possible imprints of such an event.
Fig. 4 shows, the AMR (panel a) and AAR (panel b) stellar density distributions for the whole sample.The AMR shows three distinct density features, at correspondingly three age regimes.
At the oldest regime, from 12 to ∼9-10 Gyrs, an oblique feature shows a steep rise in [Fe/H] with age with the broadest [Fe/H] distribution.This oblique feature depicts an epoch of intense SF that leads to a rapid chemical enrichment in the early MW from the lowest to the highest [Fe/H] , i.e. from −0.7 to 0.4 dex.In the accompanying AAR, at this regime, we see the high values of [α/Fe] corresponding to the SF burst in the early galaxy leading to the formation of the chemical thick disc and old bulge (e.g.Fuhrmann 1998;Anders et al. 2018;Miglio et al. 2021;Montalbán et al. 2021;Queiroz et al. 2023;Xiang & Rix 2022).In addition, this epoch is also attributed to the merger of MW with Gaia-Sausage-Enceladus (GSE; Belokurov et al. 2018;Helmi et al. 2018).
Between 9 to 5 Gyrs we see a horizontal blob with most of the stars between −0.2<[Fe/H]<0.0.This epoch reflects a slow and steady SF and is attributed to the growth of the thin disc (e.g.Chiappini et al. 1997;Minchev et al. 2013).This regime shows a birth of a large number of stars with sub-solar [Fe/H].A trend of decreasing [α/Fe] with age is seen, also a sign of slow but continued SF.Growth of the MW thin disc in an inside-out manner, due to steady gas infall is expected in this regime (see Matteucci 2021 and references therein).
Between 5 to 4 Gyrs we see a decline in SF, followed by another feature between 4 to 2.7 Gyr.This oblique feature shows an enhanced SF with a slight decrease in metallicity at younger ages.An increase in the [α/Fe] is also seen, showing a signature of SF burst.This result further confirms the enhanced SF during similar epoch previously reported (Rocha-Pinto et al. 2000;Isern 2019;Mor et al. 2019;Sysoliatina & Just 2021;Sahlholdt et al. 2022;Imig et al. 2023).These mentioned previous works have not conclusively attributed any physical process to this SF phase.Considering that intense bar growth period ended at ∼3 Gyr ago, see Sect.3.2, we deduce that this star-burst is caused by the high activity of the bar evolution.
A likely scenario one could consider is that this high bar activity is simply a buckling of an old bar.But, there is still no consensus if the bar-buckling triggers SF in the disc (See Fragk- Interestingly, in their study of the outer bar region, Wylie et al. ( 2022) also find an abrupt decline of younger stars followed by a significant fraction of stars at 2-4 Gyr for the inner disc and ring region (see their Fig. 3).Additionally, bar formation triggered star burst have been seen to typically last for a duration of ∼1 Gyr (e.g.see Baba & Kawata 2020), this corroborates our assessment.Furthermore, bars have been observed to enhance SF in external galaxies.For example, see study by Lin et al. (2020) on low-redshift galaxies using integral field spectroscopy (also see Ellison et al. 2011).Hence, we conclude that MW bar formation occurs at ∼4 Gyr ago with end of strong bar activity at ∼2.7 Gyr ago.

Conclusion
We have explored a large and homogeneous sample of 169 701 MSTO and SGB stars with 6D phase space information and high-quality stellar parameters coming from the Gaia-DR3 RVS analysis of Guiglion et al. (2023).We supplemented the chemical abundances with stellar ages, distances and kinematics to study the epoch of MW bar formation.Thanks to Gaia DR3 we obtain a mean distance uncertainty of 1% which greatly contributes to the findings of this Letter.
The new data shows two new results: a) a clear bimodality in R g at all ages (>3 Gyrs) of SMR stars, and b) a dearth of SMR stars younger than 3 Gyr.These results imply: -Milky Way's bar had a strong activity phase lasting ∼1 Gyr ending at ∼3 Gyr ago.During the phase of strong bar activity, stars formed in the inner region (bulge/bar), are significantly redistributed across the outer disc, with the highest probability of migration around bar resonances.We verify this with a dearth of SMR stars younger than 3 Gyrs and an observed bimodality in the R g around the bar resonances for the SMR stars, which are, with high confidence, formed in the inner galaxy and brought here during the strong activity phase of bar formation.-We detect an enhancement in the global SF (around −0.3 < [Fe/H]< 0.0) at around 3 Gyr which, by that time had already declined in the local thin disc.Although this SF enhancement has been detected previously in the literature, the age coinci-dence with our estimate of the bar age suggest these events to be related.-We suggest that, due to mixing and strong gas inflow due to bar there was an epoch of enhanced SF which is seen in Fig. 4.This gas inflow during this phase lowers the upper floor of [Fe/H] vs age distribution while causing an increase in the [α/Fe] due to intense SF.
Future spectroscopic surveys such as 4MIDABLE-LR Chiappini et al. (2019) will enable a large increase in the number of super metal rich stars with full 6D phase space and chemical information.In addition, the Japan Astrometry Satellite Mission for INfrared Exploration (JASMINE - Kawata et al. 2023) will be able to provide additional constraints on the age of the MW bar by precisely measuring the age of the Nuclear Stellar Disc.

Fig. 1 .
Fig. 1.The properties of our selected sample -a) Distance from the galactic mid-plane (Z) vs the Galactocentric distance (R); b) the Kiel diagram (log(g) vs T eff ); and c) the [α/Fe] vs [Fe/H] diagram for the sample, colored by number density.This gives us a final sample of 169 727 stars, including 19 367 stars with [Fe/H]>0.1 dex.As a test, we apply the 13 recommended flags on the GSP-Spec parameters (Recio-Blanco et al. 2023) to find a total sample reduced to 20 269 stars with only 4 853 stars with [Fe/H]>0.1 dex.In Fig.1, we present our sample properties.Although the stars are widely distributed in Z vs R space, the number of stars decreases as we move away from SNd, as expected for a sample of MSTO+SGB stars (seeQueiroz et al. 2023).Therefore the sample is essentially dominated by thin disc stars (low-[α/Fe] population).
Galaxy models and N-body simulations have shown that slow down of Galactic bars can lead to migration of stars from inner to outer galaxy by trapping them in resonances as they travel outwards in the disc (e.g.Athanassoula 2003; Khoperskov et al. 2020a; Chiba & Schönrich 2021).Although dependent on the exact dynamical recipe, many recent studies have placed bar Corotation (CR) between radius of 6-7 kpc (e.g.Portail et al. 2017; Khoperskov et al. 2020b; Chiba & Schönrich 2021) and the Local arm at around 8 kpc (e.g.Palicio et al. 2023) linking it to Outer Lindbald Resonances (OLR; Fragkoudi et al. 2019; Khoperskov et al. 2020b).Furthermore, Chen et al. (2022) using SMR stars, find ridges and undulations in the ϕ vs L Z plane similar to the orbits trapped in resonances of a slow bar as in the model of

Fig. 3 .
Fig. 3. Left panel: Stellar ages as a function of guiding radii for stars with [Fe/H]>0.1 dex.The colors represent number of stars per bin.A Kernel Density Estimate (KDE) has also been over-plotted.A dashed black line at 2.7 Gyr is also shown.Right panel: Upper envelope of the age-metallicity relation traced by our full sample.The triangles represent the median [Fe/H] of 20 most metal-rich stars in each age bin and the error-bars show the scatter in [Fe/H].

Fig. 4 .
Fig. 4. Top (a): 2D density distribution showing [Fe/H] as a function of stellar ages for the full sample.Bottom (b): 2D density distribution showing [α/Fe] as a function of stellar ages.A Kernel Density Estimate (KDE) have also been overplotted to highlight the density features in the AMR and AAR relationships.The colors represent number of stars per bin in log scale.