High-resolution spectroscopy of the young open cluster M 39 (NGC 7092)

The abundance and spatial distribution of the different chemical elements as well as their variation over time, are keys to understanding the origin and evolution of our Galaxy. In the era of $Gaia$ and the large spectroscopic surveys, such as for example $Gaia$-ESO (Gilmore et al., 2022; Randich et al., 2022), GALAH (De Silva et al., 2015) and APOGEE (Majewski et al., 2017), Galactic Archaeology is experiencing an unprecedent advance, significantly improving our knowledge of the structures and history of the Milky Way. Open clusters (OCs) are among the best tracers of the Galactic thin disc, since they are disseminated throughout it, covering a wide range of distances and ages. Furthermore, the chemical cluster composition, usually based on the analysis of a large number of members, is more reliable than that obtained from individual field stars. The best tool for deriving accurate chemical abundances is high-resolution spectroscopy. However, despite these large surveys and some other that are about to come into operation like WEAVE (Jin et al., 2023) and 4MOST (de Jong et al., 2019), only a small fraction of OCs (around 10%) have been observed in this way to date. This fraction could be even lower if we consider the recently discovered new OCs (Hunt & Reffert, 2023) in light of the $Gaia$ third data release ($Gaia$-DR3, Gaia Collaboration et al., 2022). Thus, high-resolution spectroscopy, such as the one conducted in this work, is highly required to continue studying our Galaxy in depth.

M 39 (also known as NGC 7092) is a sparse and nearby OC in Cygnus. It was first studied by Trumpler (1928), who classified it as a type 1a object. That means the cluster hosts only main sequence (MS) stars, i.e. with no late-type giants, not earlier than spectral type A. The most relevant works focused on M 39 date back several decades (see e.g., Ebbighausen, 1940; Eggen, 1951; Johnson, 1953; Weaver, 1953; Abt & Sanders, 1973; Abt & Levato, 1976; McNamara & Sanders, 1977; Mohan & Sagar, 1985; Manteiga et al., 1991; Platais, 1994). The cluster has been mostly studied by combining photographic and photoelectric $UBV$ photometry, spectral types obtained with low-resolution slit spectroscopy and membership identification based on proper motions derived from photographic plates. These works determined the cluster properties by observing a small number of stars, a few dozens at best. Nonetheless, Platais (1994) provided photometry and proper motions for 7931 stars in the field of M 39, assigning spectral types for 511 of them.

The aforementioned studies agree that the cluster lies at around 300 pc and that is practically unaffected by interstellar extinction. Among the MS stars the earliest spectral type is found to be B9-A0 (e.g. Abt & Levato, 1976), confirming the pioneering result of Trumpler (1928). As there is no evolved members it has not been easy to accurately determine the age of the cluster, which appears to be in the 200–500 Ma¹¹1Throughout the text we follow the recommendations of the IAU using the symbol $a$ (from the Latin word $annus$) as the unit of time rather than $yr$. range (Mohan & Sagar, 1985; Manteiga et al., 1991).

Recently, Cantat-Gaudin et al. (2020), based on the $Gaia$ second data release, placed M 39 at 311 pc, found no appreciable reddening in the cluster field and estimated an age of about 400 Ma. To date, M 39 has not yet been observed with high-resolution spectroscopy and its chemical composition remains unknown. The main aim of this paper is to fill this gap. We conducted the present investigation using high-resolution spectra, archival photometry and $Gaia$-DR3 data with the aim of characterizing the main properties of the cluster. We payed particular attention to the determination of the chemical abundances of a large number of elements that allow us to perform the proper chemical tagging of our target. The paper is structured as follows. We present our observations in Sect. 2 and explain the search performed to identify new (evolved) cluster members in Sect. 3. Then, in Sect. 4 we describe our spectral analysis and in Sect. 5 we study the cluster reddening. The investigation of the age is outlined in Sect. 6, while the discussion and comparison of our main results with the literature are presented in Sect. 7. Finally, we summarise our results and present the conclusions of this work in Sect. 8.

A fundamental step when studying a stellar cluster is to carefully choose a representative sample of members. In this work the target selection was performed before the publication of the $Gaia$-DR2 catalogue. Therefore, as a starting point, we made use of the list of members identified by Platais (1994), the most comprehensive study focused on M 39 available in the literature. Then, we completed our target list taking advantage of the selection made by Cantat-Gaudin et al. (2018). As noted above, the cluster does not host any cool giants and, therefore, our sample consists of MS stars only. For this reason, we payed special attention to the brightest stars located in the upper MS and the turn-off point (TO) that can help us to constrain the cluster age.

The disentangling of cluster members from field stars is not a trivial task. However, to succeed in the characterisation of the cluster, a correct identification of its members is essential. The arrival of a large amount of high-precision photometric and, specially, astrometric data from the $Gaia$ mission, has been of great help in simplifying this task.

Platais (1994) observed about 8000 stars in the field of M 39 and identified 511 likely members among them. His analysis was based on $UBV$ photometry and proper motions measured on photographic plates. As above mentioned, we selected our targets from this list. Specifically, we took 19 high-probability members, that is, with an assigned membership probability, $P\geq$0.7, among the brightest.

Cantat-Gaudin et al. (2018) taking advantage of the $Gaia$-DR2 identified 198 members, out of which 154 bona fide members (i.e. with $P\geq$0.7) were found. Among our targets, 14 are also high-probability members according to them, while the remaining five stars, namely 3004, 3423, 4294, 5045, and 5609, are not considered by them as cluster members. In a second run, we observed the star ucac, a likely member reported in Cantat-Gaudin et al. (2018), out of the field studied by Platais (1994).

With the intention of finding new members, and specially some giant that allow us to improve the estimation of the cluster age when performing the isochrone fitting, we carried out a search around the nominal centre of the cluster. We took all the sources from the $Gaia$-DR3 catalogue brighter than $G$=18 within a radius of 250′around it. In total we collected $\approx$1 250 000 objects with a complete set of astrometric (i. e. $\mu_{\alpha*}$, $\mu_{\delta}$, $\varpi$) and photometric ($G$, $G_{\textrm{BP}}$, $G_{\textrm{RP}}$) parameters.

Cantat-Gaudin et al. (2018) placed M 39 in the astrometric space at ($\mu_{\alpha*}$,$\mu_{\delta}$,$\varpi$) = ($-$7.477, $-$19.738, 3.337) $\pm$ (0.381, 0.393, 0.090) (mas, mas a^-1, mas a^-1). When updating the $Gaia$ parameters from DR2 to DR3 for the members reported in that paper we noticed a non negligible dispersion in the astrometric data, but not in the photometric values (see Figs. 12 and 13). For this reason we decided to recalculate the cluster centre adopting the average DR3 values of their bona fide members. Thus, we now found the cluster at ($\mu_{\alpha*}$,$\mu_{\delta}$,$\varpi$) = ($-$7.459, $-$19.839, 3.367) $\pm$ (0.345, 0.419, 0.088), where the errors correspond to the standard deviation. Once we determined the cluster centre in the astrometric space, we adopted a radius around it of 3$\sigma$, 4$\sigma$ and 5$\sigma$ as a membership criterion to define three classes of members: type I (high-probability), type II (medium-probability) and type III (marginal membership), respectively. In this way, we identified 196, 48 and 27 likely members within these three radii, finding 147, 5 and 0 bona fide members from Cantat-Gaudin et al. (2018), respectively.

In a second step we refined our selection of members by means of photometry. We plotted the ($G_{\textrm{BP}}-G_{\textrm{RP}}$)/$G$ diagram and rejected those objects (six in total) whose position on the diagram was discrepant, since they lied far away from the sequence of the cluster, that is clearly outlined. Finally, we also used the $Gaia$ RV to improve our member list. As a result, five other objects have been identified as outliers and, therefore, their membership has been ruled out (see Sect. 4.1). Our final member list is composed of 260 objects, out of which 195 are considered type I, 43 belong to type II and the remaining 22 are classified as type III members. The cluster probabilities for all our targets are listed in Table 9 and the colour-magnitude diagram is displayed in Fig. 1.

We point out that the cluster is well defined in the astrometric space and contamination from field stars is low. We have expanded the member list provided by Cantat-Gaudin et al. (2018) by about 30%, but nevertheless we have not found any giants. According to our criterion we observed 17 high-probability members among our targets, while we had to reject the stars 3004, 3423 and 4294 as cluster members. Therefore, their parameters were not used in calculating the average values of the cluster.

By examining the spatial distribution of the members in the RA$-$DEC plane (Fig.3) we found the cluster centre by fitting a Gaussian to the histogram showing the projected distribution on each coordinate. In this way we put the cluster at $\alpha$(J2000) = 21^h 31^m 23.4^s, $\delta$(J2000) = 48^∘ 11$\arcmin$ 35.3$\arcsec$, which is slightly displaced with respect to the estimate by Cantat-Gaudin et al. (2018), ours minus theirs, $\Delta(\alpha,\delta)$ = ($-$10.0^s,$-$3.2$\arcmin$). Likewise, from the parallaxes of the members we derived the distance to the cluster. Firstly, we corrected the parallax zero-point offset for each star following the recommendations outlined by Lindegren et al. (2021). As a guide, we found that, on average, this correction was of $-$0.032 mas with respect to the published value. We then calculated the individual distances by inverting the corrected parallaxes and finally, the cluster value was obtained after fitting a Gaussian to the distribution of the member distances. We place the cluster at 294$\pm$7 pc, where the error is the $\sigma$ of the Gaussian.

Before starting the spectral analysis, the telluric H₂O lines at the H$\alpha$ and Na i D₂ wavelengths, as well as those of O₂ at 6300 Å, were removed from the spectra. The same procedure described by Frasca et al. (2000) was followed. For this purpose we used telluric templates (spectra of hot, fast-rotating stars) acquired during the observing run.

In order to evaluate how our observations are affected from interstellar extinction ($A_{V}$), we resorted to the spectral energy distribution (SED) fitting tool, as previously done in other works (Frasca et al., 2019; Alonso-Santiago et al., 2021). For each of our targets we fitted the corresponding SED with BT-Settl synthetic spectra (Allard, 2014). To build the SED, on the one hand we made use of publicy available optical and near-infrared photometric data. Specifically, we used the $BVg^{\prime}r^{\prime}i^{\prime}$ (from the APASS catalogue, Henden et al., 2016), $JHK_{\textrm{S}}$ (2MASS, Skrutskie et al., 2006) and $W_{1}W_{2}W_{3}W_{4}$ bands (WISE, Cutri & et al., 2012). On the other hand, we adopted the atmospheric parameters ($T_{\textrm{eff}}$ and log $g$) derived in Sect. 4.2 as well as the (corrected) $Gaia$-DR3 parallax. The stellar radius ($R$) and the $A_{V}$ were set as free parameters. Their best values were obtained by $\chi^{2}$ minimisation. The errors on $A_{V}$ and $R$ are found by the minimisation procedure considering the 1$\sigma$ confidence level of the $\chi^{2}$ map, but we also took the error on the $T_{\textrm{eff}}$ into account. An example of this fitting is shown in Fig. 5.

The results of the SED fitting are reported in Table 5. Among the members analysed the $A_{V}$ ranges from 0.00 to 0.38 mag, with an average of $A_{V}$=0.06$\pm$0.10, where the error is the standard deviation. For most stars the reddening is on the cluster average with only star 7140 showing a slightly different value, 0.38 mag, the largest in the sample. If we recalculate the average without taking this star into account we obtain $A_{V}$=0.04$\pm$0.05 mag, where the dispersion has decreased. This result indicates that extinction is negligible in the cluster field, in good agreement with the literature (Cantat-Gaudin et al., 2018, $A_{V}=0.0$ mag).

To investigate the cluster age, we resorted to the isochrone-fitting method. In a first step we used archival photometry to plot the colour-magnitude diagram (CMD) of the cluster, which is a snapshot of its evolutionary stage. Then, we looked for the age-dependent model, the isochrone, that best reproduced the position of the cluster members on the CMD. To compute the different isochrones we took advantage of the results previously obtained from spectroscopy, that is metallicity (Sect. 4.3) and reddening (Sect. 5).

We used three different photometric systems (Johnson, 2MASS and $Gaia$) to construct the following diagrams: $M_{V}$/$(B-V)_{0}$, $M_{K_{\textrm{S}}}$/$(J-K_{\textrm{S}})_{0}$ and $M_{G}$/$(G_{\textrm{BP}}-G_{\textrm{RP}})_{0}$. In each of the diagrams the number of plotted members is different (111, 252 and 265, respectively) since photometric data is not always available for all of them. To convert apparent to absolute magnitudes we used the individual distance of each source. We calculated it as the inversion of the parallax, once it was corrected as explained above (Sect. 3).

In the two first CMDs the reddening was corrected individually for those stars observed in this work with the value previously derived, while for the remaining stars the mean cluster value was applied. In the $Gaia$ CMD we made use of the $A_{G}$ and $E(G_{\textrm{BP}}-G_{\textrm{RP}})$ published in the DR2. When comparing all the three CMDs we immediately realised that the brightest stars in the upper MS behaved anomalously in the $M_{V}$/$(B-V)_{0}$ diagram compared to the other two. This is due to the fact that these stars (with $V\approx$7 mag) are likely saturated in the APASS photometry. In order to fix this issue we resorted to the ASCC2.5 catalogue (Kharchenko & Roeser, 2009), from which we took $V$ and ($B-V$) for six stars brighter than $V$=8 after scaling both photometric datasets³³3We selected 25 members with good photometry in both catalogues in the range 7.5$\leq$ $V$ $\leq$10.5, finding average differences (ASCC2.5 minus APASS) of $\Delta\,V$=0.009 and $\Delta\,(B-V)$=$-$0.019.. The distribution of these stars on the CMD improved considerably after this correction, so that we decided to use these magnitudes (listed in Table 10 and displayed in Fig. 6) in this work.

Once the dereddened CMDs were built, we overplotted on them a set of isochrones of different ages to find the one that best reproduced the distribution of the cluster members and, thus, find the likely age for the cluster. We used PARSEC isochrones (Marigo et al., 2017) computed at the metallicity found in this work ([Fe/H]=$0.04$) in a wide range of ages (log $\tau$=8.2–8.9, which corresponds to ages from 160 to 780 Ma). We carefully examined by eye how well the isochrones (in steps $\Delta$ log$\tau$=0.01) fitted the CMDs. Since there are no giants we had to pay attention to the stars located at the upper part of the MS. We noted that the best-fitting isochrone is not exactly the same in all the diagrams, but perfectly compatible within the errors. From the optical CMD we find an age for the cluster of 460$\pm$100 Ma, similar to that obtained from the infrared one, 400$\pm$150 Ma. The errors represent the interval of isochrones that are capable of reproducing the CMD.

In the case of the $Gaia$ CMD, it is worth highlighting the role that reddening plays. As previously commented, we started taking the $Gaia$ data for the reddening and extinction. When available we used individual values, as we did in previous works. However, for the rest of members we estimated these two values by averaging those of the five nearest stars (using their distances as weights). Nevertheless, the average reddening that can be deduced for our members from the $Gaia$ DR3 catalogue, namely $A_{G}$=0.30$\pm$0.23 and $E(G_{\textrm{BP}}-G_{\textrm{RP}})$=0.20$\pm$0.14, is quite different from the one calculated by us and exhibits a greater dispersion. When adopting it, the age of cluster becomes $\approx$300 Ma, somewhat younger than the age derived from the other CMDs. If we instead redden the isochrone with the $A_{V}$ derived in this work, the dispersion of the stars along the MS decreases and the isochrones fit better the CMD. In this way, the cluster turns out to be of 420$\pm$80, a value comprised between those derived from the other two CMDs. This difference between our extinction and that derived from $Gaia$ data could be explained by the known fact that $Gaia$-DR3 tends to overestimate the extinction in areas of the sky where its value is very low (Andrae et al., 2023; Babusiaux et al., 2023; Fouesneau et al., 2023).

Finally, the definitive age of the cluster is obtained by averaging the estimates derived from each CMD, using the individual uncertainties as weights. The resulting value is 430$\pm$110 Ma, which corresponds to a MSTO mass of around 2.8 M_☉. As error we took the mean of each uncertainty. The CMDs, along with the best-fitting isochrones and their uncertainties, are shown in Fig. 6. This result agrees very well with that reported in Cantat-Gaudin et al. (2020).

We conducted this research with the idea of adding one more tile to the mosaic of Galactic Archaeology that will help us improve our knowledge about the formation and evolution of the Milky Way. We focused on M 39, a nearby young open cluster little-studied in recent years, whose chemical composition was so far unknown. We performed high-resolution spectroscopy with the HARPS-N and FIES spectrographs for 20 likely cluster members that were supplemented with archival photometry and $Gaia$-DR3 data.

According to the literature, M 39 is formed only by MS stars. In order to find any giant that belongs to the cluster we searched for new members. We recalculated the members identified by Cantat-Gaudin et al. (2018) taking advantage of the latest $Gaia$-DR3 astrometry. We expand their list, finding 260 likely members within a radius of 250$\arcmin$ around the nominal cluster centre. Among them, no evolved stars have been found. By examining the spatial distribution of the members we infer a distance to the cluster of 300 pc.

We find three SB2s in our sample. For the remaining stars, we derived their radial and projected rotational velocities and estimated the extinction and their atmospheric parameters. Additionally, for the coolest stars, nine in total, we carried out the chemical analysis, providing abundances for 21 elements. We do not find a significant reddening along the cluster field and estimate a mean RV of $-$5.5$\pm$0.5 km s^-1. By fitting isochrones on three different CMDs we find an age for the cluster of 430$\pm$110 Ma, which is consistent with the content of Li and chromospheric activity found among the members.

M 39 shows a solar-like metallicity, [Fe/H]=$+$0.04$\pm$0.08 dex, which is consistent with that expected for its Galactocentric distance. We performed the first detailed study of the chemical composition of the cluster to date. We determined the abundances of C, odd-Z elements (Na, Al), $\alpha$-elements (Mg, Si, S, Ca, Ti), Fe-peak elements (Sc, V, Cr, Mn, Co, Ni, Cu, Zn) and $s$-elements (Sr, Y, Zr, Ba). The cluster stars have a chemical composition similar to that of the Sun. Only Na shows a lower abundance, while S and the heaviest elements, specially Ba, display higher values. Its ratios [X/Fe] are on the Galactic trends displayed by the large sample of open clusters studied in surveys such as the $Gaia$-ESO, OCCAM or SPA. The cluster shows solar-like mean ratios for $\alpha$ ([$\alpha$/Fe]=$+$0.12$\pm$0.10 dex and Fe-peak elements ([X/Fe]=$+$0.05$\pm$0.10 dex), while for the neutron-capture elements the ratio is slightly overabundant ([$s$/Fe]=$+$0.28$\pm$0.10 dex). Finally, we conclude our research by claiming that the chemical composition of M 39 is fully compatible with that of the Galactic thin disc.

Based on observations made with the Italian Telescopio Nazionale Galileo (TNG) operated on the island of La Palma by the Fundación Galileo Galilei of the INAF (Istituto Nazionale di Astrofisica) at the Observatorio del Roque de los Muchachos.

This research is also partially based on observations made with the Nordic Optical Telescope, owned in collaboration by the University of Turku and Aarhus University, and operated jointly by Aarhus University, the University of Turku and the University of Oslo, representing Denmark, Finland and Norway, the University of Iceland and Stockholm University at the Observatorio del Roque de los Muchachos, La Palma, Spain, of the Instituto de Astrofísica de Canarias.

The research leading to these results has received funding from the European Union Seventh Framework Programme (F7/2007-2013) under grant agreement No. 312430 (OPTICON).