EDP Sciences logo
Subscriber Authentication Point
Search
Menu
Home All issues Volume 650 (June 2021) A&A, 650 (2021) A118 Full HTML
Free Access
Issue
A&A
Volume 650, June 2021
Article Number A118
Number of page(s) 21
Section Stellar structure and evolution
DOI https://doi.org/10.1051/0004-6361/202040207
Published online 18 June 2021

© ESO 2021

1. Introduction

The late-type giants known as S stars display, as the most characteristic feature of their optical spectra, ZrO bands (Merrill 1922), along with the usual TiO bands present in stars of similar temperatures (approximately 3000−4000 K, as in M-type stars). The spectra show overabundances of s-process elements (Smith & Lambert 1990) fd by the slow capture of neutrons on elements that are heavier than Fe (Burbidge et al. 1957; Käppeler et al. 2011) during the thermally pulsing AGB phase (TP-AGB). These elements are then brought to the surface of the AGB star through mixing processes called third dredge-ups (TDUs). The carbon over oxygen (C/O) ratio of S stars is in the range between 0.5 and 0.99, which makes them transition objects between M-type (C/O ∼ 0.4) and carbon (C/O > 1) stars (Iben & Renzini 1983).

Another important characteristic of the S star family is the technetium (Tc) dichotomy. The element Tc is produced by the s-process and has no stable isotope. Its isotope, 99Tc, has a half-life of 210 000 yr. The puzzling detection of Tc in some S stars but not in others was solved thanks to the discovery that S stars without Tc (i.e. Tc-poor S stars) belong to binary systems. They are called extrinsic S stars (Smith & Lambert 1988; Jorissen et al. 1993) because they owe their s-process abundances to mass transfer from a former AGB companion that has since evolved into a white dwarf. Therefore, the extrinsic S stars show s-process overabundances, except for Tc, which has decayed since the termination of the AGB phase of the companion. On the contrary, Tc-rich S stars, called intrinsic, are known to produce s-process elements, including Tc, and transporting them to their surface via ongoing and recurrent TDU episodes.

We recently discovered a third class of S stars, the bitrinsic S stars (Shetye et al. 2020, hereafter S20) that share properties with both the intrinsic and extrinsic classes. These stars are Tc-rich and as such are located on the TP-AGB, but they also show overabundances of Nb, a signature of extrinsically-enriched stars (Neyskens et al. 2015; Shetye et al. 2018; Karinkuzhi et al. 2018). Indeed, the unstable 93Zr isotope (produced by the s-process) decays into 93Nb, the only stable isotope of Nb, in 1.53 Myr. Therefore a niobium enrichment is expected in extrinsic stars, while it is not in TP-AGB stars.

The intrinsic S stars play an important role in our understanding of AGB nucleosynthesis and of the TDU in particular. They are the first objects on the AGB to show clear signatures (Tc and ZrO) of the occurrences of TDUs. They constrain the minimum luminosity of the TDU as well as its mass and metallicity dependence. The recent discovery of intrinsic, low-mass (initial mass < 1.5 M⊙), solar metallicity S stars (Shetye et al. 2019, hereafter S19) has provided observational evidence of the operation of TDUs even in low-mass stars.

In a previous paper (Shetye et al. 2018, hereafter S18) we introduced a new method for the parameter and abundance determination in S-type stars. However, the number of intrinsic S stars available at the time of that study was limited by the bias of the Tycho-Gaia astrometric solution against extremely red sources (Michalik et al. 2015). In the current work, we investigate the chemical composition and evolutionary status of 13 intrinsic S stars from Gaia DR2 for which we could obtain high-resolution spectra. This paper is a follow up on the investigation of the evolutionary status of S stars conducted in the pioneering work of Smith & Lambert (1990). We introduce the observational sample in Sect. 2 and discuss the Tc detection in Sect. 3. Sections 4 and 5 are dedicated to the description of the parameter and abundance determination. In Sect. 6, we present the Gaia EDR3 Hertzsprung-Russell (HR) diagram of intrinsic and extrinsic S stars. We discuss the derived elemental abundances of the sample stars and compare them with nucleosynthesis predictions in Sect. 7. Finally, we conclude with a summary of our most important results.

2. Observational sample

Based on the General Catalogue of Galactic S-Stars (Stephenson 1984, CGSS), we selected those with a Gaia Data Release 2 (DR2) parallax matching the condition σϖ/ϖ ≤ 0.3 and with available HERMES high-resolution spectra (Raskin et al. 2011). From this sample, we distinguished the intrinsic S stars from the extrinsic ones using Tc lines (Sect. 3) and we retained only the Tc-rich S stars. Among these, the Tc-rich S stars with initial masses smaller than 1.5 M⊙ were studied in S19. The cases of BD +79°156 and o1 Ori, two bitrinsic stars, were discussed in S20. In the current work, we derive parameters and abundances for the remaining Tc-rich S stars and we also compile the Tc-rich S star results from S18 (three stars), S19 (five stars), and S20 (two stars) to increase the size of our sample.

Even though the sample was designed using Gaia DR2, the Gaia Early Data Release 3 (EDR3) became public during the course of our analysis. Hence, the stellar parameters and luminosities were derived using Gaia EDR3. The relative parallax difference between the two releases, |ϖDR2 − ϖEDR3|/ϖEDR3, is usually smaller than 12% (except for V812 Oph, CSS 151, and CSS 454 which show the largest relative deviations of 22%, 33%, and 57.5%, respectively). The cause of the large parallax difference of some stars between DR2 and EDR3 is not yet clear. However, to make a consistent comparison of the present study with the previously studied samples (S18, S19, and S20), we re-computed the luminosities of all the extrinsic and intrinsic stars of S18, S19, and S20 using EDR3 (Table D.1). As a result, some stars that had previously been identified (using DR2 parallaxes) as low-mass S stars (BD +34°1698 = CSS 413 and HD 357941 = CSS 1190; see S19) turned out to have higher masses (Mini = 1.4 and 3.5 M⊙, respectively) according to their location in the HR diagram using EDR3 parallaxes (Sect. 6). This change in their masses required the adoption of a log g value that was different from the one used for their abundance analysis in S19 (both stars had log g ∼ 1 in S19; with EDR3 we find instead log g ∼ 0 for both of them). Hence, in the current work we also re-computed the stellar parameters and abundances of BD +34°1698 and HD 357941 along the guidelines described in Sects. 4 and 5 using the EDR3 parallaxes (Table D.2).

A last criterion to define the Tc-rich S-star sample relates to the photometric variability. Large convective cells or pulsations are responsible for the photometric variability of TP-AGB stars. The thermal structure of these pulsating atmospheres can be significantly different from the time-independent, hydrostatic atmospheres used to derive their stellar parameters, which can therefore prove quite unreliable. We checked the time series data obtained by the American Association of Variable Star Observers (AAVSO) and removed from our sample the Tc-rich S stars with photometric variability ΔV >  1 mag (see Table 1), apart for V812 Oph. The HERMES spectra of V812 Oph, despite its high variability (ΔV >  1.4 mag), match the atmospheric models all over the optical range well; thus, we included it in the current sample. The sample of newly analyzed stars is listed in Table 2.

Table 1.

Tc-rich S stars excluded from the current study because of their large photometric variability.

Table 2.

Basic data for our S star sample.

3. Tc detection

We used three Tc I resonance lines located at 4238.19, 4262.27 and 4297.06 Å. A signal-to-noise ratio (S/N) of at least 30 in the V-band is needed to detect the Tc lines in S-star spectra (Van Eck et al. 1998, S18, S19). The spectra of our sample stars all have a S/N (listed in Col. 11 of Table 2) equal to or larger than 40 in the V band and 30 in the B band. The Tc absorption features of our sample stars around the three Tc lines are presented in Fig. 1.

thumbnail Fig. 1.

Spectral region around the three (4238.19, 4262.27 and 4297.06 Å) Tc I lines in our sample of Tc-rich S stars. For comparison purposes, the spectrum of a Tc-poor S star (HD 233158, in blue in the top panels) from S18 is also plotted. The spectra have been arbitrarily normalized and binned by a factor of 1.5 to increase the S/N. The Tc I line at 4238.19 Å could not be used for CD −27°5131 as local normalization was hampered by a nearby cosmic ray hit.

We confirm all previous Tc-rich classifications, except for one target – HD 288833, classified as Tc-poor by Jorissen et al. (1993) despite the fact that its IRAS excess ([12]–[25]) is larger than −1.3, whereas extrinsic stars have [12]–[25] < −1.3 according to their classification criterion. This criterion was designed by Jorissen et al. (1993) to diagnose the expected lower infrared excess of the extrinsic stars with respect to that of (more evolved) intrinsic stars. The three Tc absorption features can be readily identified for this object in Fig. 1.

For all other stars, our classification is in agreement with previous findings when they exist. HR Peg, AA Cam, V1139 Tau, V679 Oph, and HD 64147 were indeed classified as intrinsic S stars in Smith & Lambert (1988, 1990) and Jorissen et al. (1993). We agree with Otto et al. (2011) classification of CSS 1152 as an intrinsic S star based on AKARI photometry and also confirm the results of Neyskens et al. (2015) that KR CMa and CSS 454 are intrinsic S stars. We could not use the Tc line at 4238.19 Å for CD −27°5131, as the region around this line is dominated by a cosmic ray hit. From its other two Tc lines we assess the Tc-rich nature of this object, in agreement with SVE17. For V812 Oph, CSS 151 and BD −18°2608, that we classify as Tc-rich, no literature classification is available.

4. Derivation of the atmospheric parameters

Disentangling the intricate parameter space of S stars has always been a challenging task because their spectra are dominated by molecules. The atmospheres of S stars are complex, as the thermal structure is dependent on their chemical composition (C/O and, to a lesser extent, heavy elements). Different methods have been used in the past for the stellar parameters determination. For instance, in their pioneering work, Smith & Lambert (1985) derived Teff from the (V − K) colors and log g from the standard log g – mass relationship where they estimated the masses by comparing the positions of the S stars in the HR diagram with the evolutionary tracks. However, the Teff − (V − K) relation valid for M-type giants has been shown to be inappropriate for S-type stars (SVE17) because of the spectral energy distribution (SED) alteration (mainly by ZrO, LaO and YO bands) induced by the non-standard chemical composition of S stars. New MARCS model atmospheres (Gustafsson et al. 2008) were designed to cover the full parameter range of S stars, with Teff from 2700 K to 4000 K, log g from 0 to 5, [Fe/H] of 0.0 and −0.5, C/O values of 0.500, 0.752, 0.899, 0.925, 0.951, 0.971, 0.991, and [s/Fe] of 0.0, 1.0 and 2.0 (SVE17).

In the current study, the stellar parameters are determined as in S18. In summary, the high-resolution HERMES spectra are compared with a grid of ∼3500 synthetic spectra computed from the S star MARCS models to obtain atmospheric parameter estimates Teff, log g and [Fe/H]. The fitting is undertaken over small spectral windows (listed in Table E.1) because HERMES spectra cannot be considered as flux-calibrated over their whole wavelength range. The fit with the lowest total χ2 value then identifies the best-fitting model, providing a first estimate of the atmospheric parameters.

The stellar luminosities were calculated using the distances derived from the Gaia EDR3 parallaxes after applying the zero-point correction from Lindegren et al. (2021), the reddening EB − V computed from Gontcharov (2017) and the bolometric correction in the K band as computed from the MARCS model atmospheres. Combining the Teff, metallicity and luminosity, we located the stars in the HR diagram and compared them with STAREVOL (Siess 2006) evolutionary tracks to estimate their current stellar masses.

A new surface gravity was then computed and the procedure iterated (as described in Fig. 5 of S18) until the log g retrieved from spectral fitting was consistent with the one obtained from the HR diagram. The uncertainties on the stellar parameters were obtained from the variations of the atmospheric parameters while iterating for log g. Our final set of parameters and the corresponding uncertainties are presented in Table 3. A further discussion of the reliability of the stellar masses found in this way, based on their correlation with the height above the galactic plane and with 2MASS photometry, may be found in Appendix A.

Table 3.

Atmospheric parameters for S stars.

5. Chemical abundance determination

The abundance determination methodology is the same as the one adopted in S18 and S19. We compared the observed spectra with synthetic spectra generated using Turbospectrum v15.1 (Plez 2012) and MARCS model atmospheres of S stars with the parameters derived in Sect. 4. We made use of the same input molecular line lists as in SVE17 and atomic line list as in the Gaia-ESO survey (Heiter et al. 2021) and we varied the abundances till a satisfactory agreement could be found.

5.1. Li

We used the Li I line at 6707.8 Å to derive the Li abundance. This line is known to be blended with a neighboring Ce II line at 6708.099 Å in the warmer post-AGB stars (Reyniers et al. 2002). However, we did not identify any dominating blend from this line to the Li I line in our sample stars. In Fig. 2, we present examples of stars with a good (bottom panel) as well as bad (top panel) spectral-synthesis fit of the considered Li line. We could derive the Li abundance for only three stars from our sample, namely HR Peg, V679 Oph and Vy 12. For these stars, the absorption features of the Li doublet located at 6707.8 Å are very clear. For the rest of the sample, there were severe blends in the Li line (top panel of Fig. 2), so only upper limits on the Li abundances could be derived. The results are listed in Table C.1.

thumbnail Fig. 2.

Illustration of the fit quality between the observed and synthetic spectra of BD +79°156 (top panel) and V679 Oph (bottom panel) around the Li line at 6707.8 Å.

The Li abundances of all the intrinsic S stars of our sample are generally low except for HR Peg. HR Peg was found to be a Li-rich star also by Vanture et al. (2007) who classified the star as a high-mass star (M >  3 M⊙) with hot bottom burning (HBB) as an explanation for the Li abundance. However, our mass estimate for HR Peg is Mcurr = 2.2 M⊙, in agreement with the 2.0 M⊙ value found by Neyskens et al. (2015). Therefore, HR Peg does not appear to be massive enough to be producing Li through HBB. Other plausible explanations involve some extra mixing in low-mass AGB stars (e.g. Charbonnel & Balachandran 2000; Uttenthaler et al. 2007; Uttenthaler & Lebzelter 2010) or the engulfment of planets or brown dwarfs (Siess & Livio 1999).

5.2. C, N, O

We used the CH bands around 4250 Å to determine the C abundance. The sensitivity of these bands to the carbon abundance is limited in S-star spectra because CH bands are blended with mainly TiO and ZrO (as can be seen from Fig. 16 of SVE17). It was not possible to derive the O abundance because the λ6300.3 Å O I line lies in a severely blended region. We used the solar oxygen abundance (Asplund et al. 2009), scaled with respect to the metallicity and included an alpha enhancement ([α/Fe] = −0.4 × [Fe/H]) for [Fe/H] ≥ −0.5. The uncertainty on C/O was estimated from the range of values of the C/O providing an acceptable fit to the CH G-band. The N abundance was then determined from the CN lines available in the region 8000−8100 Å. The line list for these CN lines was taken from Sneden et al. (2014).

5.3. [Fe/H]

We used the same line lists as S18 and S19 (see also Table F.1) for the Fe line synthesis. Between 9 and 15 Fe lines were used to derive the metallicity for all the sample stars, except for the star Vy 12, for which only six Fe lines could be used, as the other lines were strongly blended by molecules because of the high C/O of this object (0.97). The derived metallicity as well as the standard deviation due to line-to-line scatter are listed in Table 3.

5.4. Light s-process elements (Sr, Y, Zr, Nb)

The line list of S18 was used and complemented as documented in Table F.1. The strontium abundance could be derived only for HD 288833, BD −18°2608, and V679 Oph. The Y abundance was measured for all the sample stars. The Zr abundance was measured for all stars except KR CMa using the two Zr I lines at 7819.37 and 7849.37 Å with transition probabilities from laboratory measurements (Biémont et al. 1981). Some Nb I lines from Table F.1 could not be used in some stars. In fact, none of the Nb I lines could be used for CSS 151 because of the low S/N of its spectrum. The Nb abundance could therefore not be derived for this star.

5.5. Heavy s-process elements (Ba, Ce, Nd)

We derived the barium abundance using only one Ba I line located at 7488.07 Å. The cerium abundance was determined using the Ce II lines from S18 together with some new lines located at 7580.91, 8025.57, 8394.51, 8405.25, and 8769.91 Å. There is a huge scatter in the Ce abundances for HR Peg, o1 Ori, KR CMa, BD +79°156, and HD 64147 despite the use of Ce II lines displaying satisfactory fits. The bottom panel of Fig. 3 illustrates the fact that even neighboring Ce II lines (at 8769.91 and 8772.13 Å) can provide discrepant abundances (by ∼0.4 dex in this case). As explained in Karinkuzhi et al. (2018), cerium abundances derived using only Ce II lines above 7000 Å are ∼0.3 dex lower than the ones derived from Ce II lines in the range 4300−6500 Å. These authors mention that this difference might be attributed to non-LTE (non-local thermodynamic equilibrium) effects. This could be the reason for the sub-solar cerium abundances in some of our stars, as they were determined using only red Ce II lines (λ >  7000 Å). These uncertain abundances are marked with a colon in Tables C.2 and C.3.

thumbnail Fig. 3.

Illustration of the fit between the observed and synthetic spectra for the S star HR Peg. Top panel: presents 5 Å around the Zr I line at 7819.37 Å and bottom panel: around the Ce II line at 8772.135 Å.

5.6. Other heavy elements (Pr, Sm, Eu)

Good Eu II and Pr II lines are located in the bluer part of the spectrum and could be used in some stars where the blending was weak. The Sm abundance was derived when possible using the Sm II 7042.20 and 7051.55 Å lines. These lines are well fitted and yield consistent abundances.

5.7. Tc

All three Tc I resonance lines at 4238.19, 4262.27, and 4297.06 Å are heavily blended (Little-Marenin & Little 1979). We first derived the other s-process abundances in order to reproduce these blends as precisely as possible. The 4262.27 Å line is the best reproduced by our spectral synthesis. Its blends, consisting in two neighbouring lines of Nb I (at 4262.050 Å) and Gd II (at 4262.087 Å) were identified in Van Eck & Jorissen (1999). The Tc abundance was derived from the 4262.27 Å Tc I line in our sample stars as well as in all intrinsic S stars from S18 and S19 (Table C.1).

5.8. Uncertainties on the abundances

The uncertainties on the abundances were computed using the ones of the S star V915 Aql investigated in S18. The atmospheric parameters of V915 Aql are representative of those of most of our sample stars, hence we computed the elemental abundance error by quadratically adding the elemental standard deviation due to line-to-line scatter, the abundance errors due to parameter uncertainties of V915 Aql as derived in S18 (see also Appendix B), and a term of 0.1 dex to take into account continuum placement errors. For abundances that were derived using only one line, an arbitrary line-to-line scatter of 0.1 dex was assumed. Error bars on Tc abundances were estimated as the range of Tc abundances providing an acceptable fit of the Tc I 4262.27 Å line. The final elemental uncertainties on abundances are listed in Tables C.2–C.4.

6. HR diagram of S stars

In Fig. 4 we present an HR diagram collecting Tc-rich S stars studied in S18, S19, and S20 as well as those of the present work, together with Tc-poor stars from S18. We used the final stellar parameters presented in Table 2. The errors on Teff are taken from Table 2. For HD 288833, V812 Oph, AA Cam, and V679 Oph for which Teff did not change during the log g iterations, we imposed a standard symmetric error of 100 K on their Teff. The asymmetric error on the luminosity was derived after propagating the error on the parallax. The evolutionary tracks displayed in Fig. 4 were computed with the STAREVOL code (Siess et al. 2000) and are described in detail in Escorza et al. (2017). Briefly, we use standard input physics with a mixing-length parameter α = 1.75, grey surface boundary conditions and the solar mixture from Asplund et al. (2009). Opacity enhancement due to the formation of molecules in carbon-rich atmospheres is also accounted for following the formulation of Marigo (2002). The Schröder & Cuntz (2007) mass-loss prescription is activated up to the end of core helium burning followed by the Vassiliadis & Wood (1993) formulation during the AGB phase. We also consider overshooting below the convective envelope following the exponential decay expression of Herwig et al. (1997) with the parameter fover = 0.01. Models were computed for various metallicities, including [Fe/H] = 0, −0.25, and −0.5.

thumbnail Fig. 4.

HR diagram of intrinsic (filled triangles) S stars from our large sample and extrinsic (open squares) S stars from S18 along with the STAREVOL evolutionary tracks corresponding to the closest metallicities. The red giant branch is represented in blue, the core He-burning phase in green, whereas the red dashed lines correspond to the AGB tracks. The black dashed line represents the predicted onset of the third dredge-up, i.e., the lowest stellar luminosity following the first occurrence of a TDU episode (down to 1.5, 1.3, 1.0 M⊙ for [Fe/H] = 0.0, −0.25, −0.50 respectively).

There is a clear segregation between the location of intrinsic and extrinsic S stars in the HR diagram. Indeed, whatever the considered metallicity range, intrinsic S stars are always located above the black dotted line (see Fig. 4) marking the predicted onset of TDU for masses above 1.5 M⊙. On the other hand, it is obvious that Tc-rich stars must be TP-AGB stars (because of the Tc detection). Thus, such a consistency between the luminosity of the third dredge-up onset in theoretical models and the luminosity of observed intrinsic S stars, which are the least evolved objects identified as TP-AGB stars, has not been as clearly demonstrated thus far.

Table 4 summarizes the observational constraints on the first occurrence of TDUs in terms of luminosity, initial mass, and metallicity. For each mass and metallicity bin, we indicate the lowest luminosity of intrinsic S stars from our so-called “large sample”, collecting stars from S18, S19, S20, and from the current work. These luminosities represent an observational upper limit on the first occurrence of the TDU in AGB stars, in the sense that Tc-rich stars are observed at such luminosities, so that a TDU must have already occurred by the time the star reaches this position in the HR diagram. However, since S stars are the first objects on the AGB to show clear signatures of TDU events (Tc and ZrO), this upper limit must be quite close to the genuine TDU occurrence line. It can be used as an observational constraint to be satisfied by the stellar evolutionary models of the corresponding masses and metallicities.

Table 4.

Observed lowest luminosities of intrinsic S stars in different mass and metallicity bins (collected from this work, S18, S19, and S20).

7. Discussion on the abundances and comparison with STAREVOL nucleosynthesis predictions

The elemental abundances derived following the methodology presented in Sect. 5 allow us to investigate nucleosynthesis in intrinsic S stars. We compare the measured abundance profiles of our sample stars with the nucleosynthesis calculations of the STAREVOL code (Goriely & Siess 2018) which uses an extended network of 411 species.

The predicted and measured abundance distributions are presented in Fig. 5. The pulse number is chosen in order to optimally match both the overabundance level and the location in the HR diagram. We find a good overall agreement between the predicted and measured distributions of heavy elements. In particular, the peak of heavy s-elements is well reproduced in all stars. However, problems persist with the light elements such as carbon or oxygen, as discussed in Sect. 7.1, and with some heavy elements, for instance, Ce, as discussed in Sect. 5.4. The models do account for most of the derived Tc abundances; however, we note that abundance predictions for Tc are extremely sensitive to the pulse number and to the amount of dilution in the stellar envelope due to the initial absence of Tc in the star. Hence, the agreement between predicted and measured abundances may be poor in some cases, for instance, for CSS 151, BD −18°2608, and CSS 454. We now go on to investigate specific element ratios.

thumbnail Fig. 5.

Measured s-process element abundances compared with the nucleosynthesis predictions. The blue line represents the nucleosynthesis predictions compatible with the mass and metallicity of the individual stars. The number of pulses n required to best-match the measured abundances is mentioned in the label of every panel as “p.n”.

7.1. [C/Fe] and [s/Fe]

The surface composition of TP-AGB stars should reflect the addition of 12C originating from the He-burning shell and s-process material produced either radiatively in the interpulse or in the convective thermal pulses for more massive objects. The TDU is then responsible for transporting these products to the stellar surface. As the star evolves on the TP-AGB, its carbon abundance should thus increase along with its s-process over-abundances. Eventually, the star becomes carbon-rich when the C/O ratio exceeds unity. However, massive AGB stars with Mini ≳ 4 M⊙ experience HBB and can efficiently burn the dredged-up carbon to produce mainly 14N. In the top panel of Fig. 6, we present the carbon abundance as a function of s-process abundance for our intrinsic S star sample. The [s/Fe] index has been calculated using the Y, Zr, and Ba abundances and is listed in Table C.5. In order to compare this trend with that of the stars from the next evolutionary stage, we added in Fig. 6 the Tc-rich carbon stars from Abia et al. (2002). A larger marker size indicates a lower metallicity (stars are grouped in three [Fe/H] bins: [−∞, − 0.4], [−0.4, −0.2], and [−0.2, 0]).

thumbnail Fig. 6.

Distribution of the carbon abundance of our sample stars with respect to their [s/Fe] index. Top panel: measured [C/Fe] abundances as a function of [s/Fe] for our large sample of Tc-rich S stars (light green, purple and orange squares) where the size of the symbol increases with decreasing metallicity, considering three metallicity bins: [Fe/H] in [0.0; −0.2]: light green, [−0.2; −0.4]: purple, below −0.4: orange. Tc-rich carbon stars from Abia et al. (2002) are shown as green squares. Bottom panel: derived C/O ratio versus [s/Fe] for the same stars. Predictions for a 2 M⊙ STAREVOL model at metallicity [Fe/H] = 0.0 (light green), −0.3 (violet) and −0.5 (orange) are overplotted. The empty circles along the tracks indicate the successive TDUs, while the three filled circles mark the first TDU, which led to C/O > 1 in each model.

In the top panel of Fig. 6, the trend of increasing [C/Fe] with increasing [s/Fe], expected for TP-AGB stars, is not marked. Nevertheless, the S stars most enriched in s-process elements are the ones with the highest carbon abundance. The carbon stars have higher carbon abundances than most S stars but are characterized by [s/Fe] indices in the same range as those of the bulk of S stars. A group of intrinsic S stars have similar [s/Fe] and [C/Fe] ratios as carbon stars. This can be explained as follows: the carbon stars from Abia et al. (2002) have solar metallicities while the intrinsic S stars belonging to the group with similar [C/Fe] (in the range 0.2−0.35 dex) have lower metallicities (in the range −0.2 to −0.5 dex). Because [α/Fe] increases with decreasing metallicity in the range −0.5 < [Fe/H] < 0, this group of S stars has a higher initial [O/Fe] compared to carbon stars, enabling their O-rich classification.

The [C/Fe] abundances in this work were derived from the C/O ratios assuming solar oxygen abundance scaled with respect to metallicity and taking into account an α-element enhancement (Sect. 5.2). The determination of [C/Fe] is therefore quite indirect. We thus present in the bottom panel of Fig. 6 the C/O ratios. Their determination is very robust, given the high sensitivity of molecular bands to the C/O ratio. As expected, there is no overlap between C/O ratio of C stars and S-type stars. We note that three S-type stars have C/O ratios close to, but definitely lower than 1: indeed their C/O of 0.899, 0.971, and 0.998 produces an observed spectrum that is markedly different from the spectrum of an SC star (characterized by C/O = 1).

A common misunderstanding is based on the notion that S stars have C/O = 1. This statement is often encountered in the literature even as a definition of S-type stars. Once again (see also Van Eck et al. 2017), we stress here that only S stars with the highest [s/Fe] values have C/O ratios approaching (but not reaching) unity. Most S-type stars have intermediate C/O ratios (between 0.5 and 0.8). Stars with C/O = 1 are actually classified as SC or CS stars.

Figure 6 also compares the measured carbon and s-process abundances with the nucleosynthesis predictions at three different metallicities ([Fe/H] = 0.0, −0.3, and −0.5). The filled circles along the tracks mark the thermal pulses turning the C/O ratio to values above 1, thus changing the (model) stars into carbon stars. First, the different models predict a tight correlation between C/O and [s/Fe], whereas there is a large scatter in measured [s/Fe] at a given C/O. Second, theoretical calculations show a much faster increase of C/O with [s/Fe] than what is actually measured in stars. For example, all models predict that stars with [s/Fe] ≥ 0.55 should have C/O > 1 (and be carbon stars), whereas many S stars (which must have C/O < 1) are observed with such large s-process enrichments. In other words, in theory, high s-process enrichments go along with very high C/O ratios which are incompatible with S-star classification. The reason why carbon stars do not show such large s-process overabundances (stronger line blending? dust obscuration of the most evolved objects?) is yet unclear.

Finally, we remark that Fig. 6 reveals the existence of one S star, o1 Ori, with very mild – if any – enrichment in s-process elements (see [s/Fe] in Table C.5, Fig. 4 and Table A.1 from S20 for individual heavy-element abundances). This star is another example of the “Stephenson M-type stars” uncovered by Smith & Lambert (1990). Nevertheless, it shows clear Tc signatures (Fig. 1 of S20). The possible reasons for the low [s/Fe] index of o1 Ori are discussed in detail in S20 and Jorissen et al. (2019).

7.2. [hs/ls] and metallicity

We now go on to investigate the potential correlations between the heavy (hs) to light (ls) s-process element ratio [hs/ls], mass and metallicity. The neutron irradiation index [hs/ls] is actually expected to increase with decreasing metallicity (Goriely & Mowlavi 2000), as the number of neutrons per iron seed increases.

From our measured abundances, we find a large scatter in [hs/ls] with respect to metallicity (Fig. 7) within our limited range of metallicity. De Smedt et al. (2015, 2016) also reported such a scatter in the [hs/ls]−[Fe/H] plane from their study of s-process enriched post-AGB stars. However, the evolutionary link between post-AGB stars and intrinsic stars remains to be established with certitude. Figure 7 also shows the absence of a clear correlation between the [hs/ls] ratio and the initial stellar masses.

thumbnail Fig. 7.

Distribution of [hs/ls] as a function of metallicity for the intrinsic S stars of our study (filled squares). The hs-index has been calculated using Ba and Nd, and the ls-index using Y and Zr. The symbols are color-coded with respect to the initial mass in the bins Mini < 1.5 M⊙ (blue), 1.5 ≤ Mini < 2.5 M⊙ (red), and Mini ≥ 2.5 M⊙ (green). We note that these mass bins are designed to match the [hs/ls] STAREVOL predictions only available for Mini = 1.5, 2, 2.5 M⊙. Open circles denote the predictions from pulse to pulse for the different initial masses.

When compared with the theoretical predictions accounting for the pulse-to-pulse variation, plotted in the top panel of Fig. 7, we find that the overall range covered by our measured [hs/ls] indices is compatible with that of the predicted abundances of the theoretical models. However, the [hs/ls] model predictions are not available for the complete metallicity range covered by our measured [Fe/H].

7.3. Technetium abundances

7.3.1. [s/Fe], C/O and Tc

Figure 8 presents [s/Fe] (calculated as described in Sect. 7.1 using Y, Zr, and Ba abundances) as a function of the Tc abundance, for different C/O ratio bins. The stars with the highest [s/Fe] and Tc abundances are also the ones with the highest C/O ratios, which is consistent with expectations from TP-AGB evolution. However, it is worth mentioning that whereas [s/Fe] is predicted to monotonically increase as the star ascends its AGB (see Fig. 6, this is also true for C/O if there is no HBB, and for luminosity, set aside the luminosity variations during thermal pulses), the technetium abundance has a more complex behaviour, because the 99Tc half life (210 000 yr) is not totally negligible with respect to the TP-AGB duration. For example, the evolution of the technetium-to-zirconium abundance ratio is not flat but shows a non-trivial evolution displayed in Fig. 2 of Neyskens et al. (2015). The lack of a clear trend in Fig. 8 is therefore not surprising.

thumbnail Fig. 8.

[s/Fe] index versus Tc abundance. The different colors represent different C/O ratio bins as described in the figure.

7.3.2. Luminosity and Tc

In Fig. 9, we compare the Tc abundances and the luminosities of the sample stars. A large scatter in luminosity is present for each Tc abundance. If we focus on a specific mass bin, there is a loose trend of increasing Tc abundance with increasing luminosity, represented by the dashed least-square fit lines in Fig. 9. We also find that the most luminous and massive star in Fig. 9 has a relatively moderate Tc overabundance. A likely explanation is the larger dilution of s-elements in the bigger envelope of more massive AGB stars. In addition, with increasing stellar mass, the mass of the 13C pocket responsible for the s-process, as well as that of the thermal pulses are reduced because of the stronger compression of the shells induced by the larger core mass. Lower surface abundances of s-process elements are thus expected in higher-mass TP-AGB stars (García-Hernández et al. 2013).

thumbnail Fig. 9.

Stellar luminosity as a function of Tc abundance. The symbols are color coded according to their mass bin. The blue and red dashed lines represent the linear least-squares fit for the stars in the corresponding mass bin.

7.4. Segregation of the intrinsic and extrinsic stars in the Zr–Nb plane

Particular attention can also be paid to the Zr and Nb abundances that can be used as extrinsic or intrinsic star markers (Neyskens et al. 2015). As already mentioned in Sect. 1, niobium is mono-isotopic and can only be produced by the decay of 93Zr (with a half-life of 1.53 × 106 yr). Intrinsic S stars are freshly producing s-process elements, including 93Zr, and there is not enough time on the TP-AGB for a significant production of 93Nb. On the contrary, in extrinsic S stars, enough time has elapsed since the end of the nucleosynthesis in the companion for 93Zr to have totally decayed into 93Nb. As expected, intrinsic and extrinsic S stars follow different trends in the [Zr/Fe]−[Nb/Fe] plane, where intrinsic S stars have [Zr/Nb] < 1, while extrinsic stars have [Zr/Nb] ∼ 1 (Neyskens et al. 2015; S18). The Zr−Nb plane has been studied extensively for extrinsically-enriched objects: the high Nb abundance in extrinsic stars has been demonstrated in various extrinsic-star families among which CH and CEMP stars (Karinkuzhi et al. 2018, 2021). The apparent shift of extrinsic stars from the STAREVOL 2−3 M⊙ predictions by a few tenths of a dex is as well discussed in Karinkuzhi et al. (2018). These authors speculated that this shift could arise due to the oscillator strengths of the two Zr lines used to derive the Zr abundance of the extrinsic S stars, which have a tendency to yield abundances that are about 0.1 dex too low in the benchmark stars of their sample (Arcturus and V762 Cas). Here, we populate the high-Zr, low-Nb region of this plane while adding constraints from the Tc-rich S stars.

Figure 10 confirms, on the basis of our so-called “large sample” of intrinsic S stars, the apt segregation between extrinsic and intrinsic stars in the [Zr/Fe]−[Nb/Fe] plane (with one exception discussed below). We also compare the measured Zr and Nb abundances with nucleosynthesis predictions from STAREVOL for intrinsic and extrinsic stars of 2 and 3 M⊙, with metallicities [Fe/H] = 0.0 and −0.2. As shown in Fig. 10, the intrinsic S stars from our sample have [Zr/Fe] < 1.6 and follow the trend predicted for TP-AGB stars (open squares in red and green). It is worth noting that some 93Zr decay (inducing some 93Nb production) is both predicted and observed, as can be seen from the fact that highly-enriched (high-Zr) intrinsic stars, which are the most evolved S stars on the TP-AGB, tend to have the largest Nb enrichments, and nicely follow the STAREVOL inclined trend in the [Zr/Fe]−[Nb/Fe] plane.

thumbnail Fig. 10.

[Zr/Fe] as a function of [Nb/Fe] for the intrinsic S stars (black filled triangles) and extrinsic S stars (open squares) from S18 compared with the nucleosynthesis predictions for different initial masses and metallicities. The bitrinsic S stars from S20 are plotted with a blue star symbol. The “ext” and “int” labels next to the model parameters refer to extrinsic (respectively intrinsic) model abundances.

The two Tc-rich S stars o1 Ori and BD +79°156 are “bitrinsic” S stars as they are Tc-rich (intrinsic) but show signs of binarity together with a large Nb enrichment (for the corresponding [Zr/Fe] ratio), two bits of evidence to show that they are also extrinsic stars, as discussed in S20. In addition to these two stars, the use of Gaia EDR3 parallaxes revealed a potential new “bitrinsic” candidate, BD +34°1698. We re-computed the s-process abundances of this star as its stellar parameters (mainly mass and log g) changed when Gaia DR2 parallaxes were replaced by Gaia EDR3. The revised s-process abundances of BD +34°1698 are presented in Table D.2. It has a [Zr/Nb] ratio close to unity, along with clear signatures of Tc (see Fig. C.1 of S19), hence, it qualifies as a “bitrinsic” candidate. Wang & Chen (2002) classified it as a candidate extrinsic S star based on its IRAS photometry. Two spectra taken with the Hermes spectrograph revealed a clear radial-velocity variation: Vr = 18.59 ± 0.05 km s−1 on JD 2 457 502.44 and Vr = 22.59 ± 0.09 km s−1 on JD 2 459 289.51. Our detection of the binary motion associated to a clear Tc enrichment and a [Zr/Nb] ratio close to unity unambiguously classify BD +34°1698 as a member of the restricted family of bitrinsic stars (S20). Lastly, Fig. 10 confirms that the Zr−Nb analysis successfully serves as an additional test (apart from Tc) for the classification of S stars as intrinsic or extrinsic.

8. Conclusions

Thanks to the combination of Gaia EDR3 parallaxes and the high-resolution HERMES spectra, we were able to determine the stellar parameters of a sample of 13 intrinsic S stars with metallicities in the range of −0.7 < [Fe/H] < 0. We then derived their s-process element abundances. The heavy-element abundances of intrinsic S stars reveal their rich nucleosynthetic history. The main results from our study can be summarized as follows:

(i) The Gaia EDR3 HR diagram of S stars confirms that intrinsic S stars are more evolved than their extrinsic counterparts.

(ii) The luminosity lower limits for the occurrence of TDU in different mass and metallicity ranges provided in Table 4 can be used to constrain AGB evolutionary models.

(iii) The objects from our sample with the largest C/O ratios are also the ones with the largest [s/Fe], which is consistent with TP-AGB evolution predictions.

(iv) However, we clearly demonstrate the too rapid increase of the C/O ratio with respect to [s/Fe] in model predictions.

(v) The measured s-process abundances of intrinsic S stars are in good agreement with the AGB nucleosynthesis predictions for models of the corresponding mass and metallicity. In particular, the Zr and Nb abundances are matching very well the predicted trend for intrinsic S stars, confirming our previous finding (S18, S20) that the Nb abundance can be used as an intrinsic or extrinsic diagnostic as efficient as the Tc presence or absence.

(vi) We present Tc abundances for a large sample of intrinsic S stars (20 stars). We find that the stars with the highest C/O ratios tend to be the ones with the highest Tc abundances.

In conclusion, the current investigation of a sample of intrinsic S stars has extended our understanding of their properties, particularly with regard to their location in the HR diagram and their chemical characterization, in considering ∼10 chemical elements, including their C/O ratio and the radioactive element technetium.

Acknowledgments

The authors thank the anonymous referee for constructive comments. This research has been funded by the Belgian Science Policy Office under contract BR/143/A2/STARLAB. SS, SVE, SG, MG acknowledge support from the FWO and FNRS Excellence of Science Programme (EOS-O022818F). SVE thanks Fondation ULB for its support. GW’s research is supported by The Kennilworth Fund of the New York Community Trust. Based on observations obtained with the HERMES spectrograph, which is supported by the Research Foundation – Flanders (FWO), Belgium, the Research Council of KU Leuven, Belgium, the Fonds National de la Recherche Scientifique (F.R.S.-FNRS), Belgium, the Royal Observatory of Belgium, the Observatoire de Genève, Switzerland and the Thüringer Landessternwarte Tautenburg, Germany. 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 research has also made use of the SIMBAD database, operated at CDS, Strasbourg, France. LS and SG are senior FNRS research associates.

References

  1. Abia, C., Domínguez, I., Gallino, R., et al. 2002, ApJ, 579, 817 [NASA ADS] [CrossRef] [Google Scholar]
  2. Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481 [NASA ADS] [CrossRef] [Google Scholar]
  3. Biémont, E., Grevesse, N., Hannaford, P., & Lowe, R. M. 1981, ApJ, 248, 867 [NASA ADS] [CrossRef] [Google Scholar]
  4. Burbidge, E. M., Burbidge, G. R., Fowler, W. A., & Hoyle, F. 1957, Rev. Mod. Phys., 29, 547 [NASA ADS] [CrossRef] [Google Scholar]
  5. Charbonnel, C., & Balachandran, S. C. 2000, A&A, 359, 563 [Google Scholar]
  6. Corliss, C. H., & Bozman, W. R. 1962, in Experimental Transition Probabilities for Spectral Lines of Seventy Elements; Derived from the NBS Tables of Spectral-line Intensities (US Government Printing Office), NBS Monograph, 53 [CrossRef] [Google Scholar]
  7. De Smedt, K., Van Winckel, H., Kamath, D., & Wood, P. R. 2015, A&A, 583, A56 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  8. De Smedt, K., Van Winckel, H., Kamath, D., et al. 2016, A&A, 587, A6 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  9. Den Hartog, E. A., Lawler, J. E., Sneden, C., & Cowan, J. J. 2003, ApJS, 148, 543 [Google Scholar]
  10. Duquette, D. W., & Lawler, J. E. 1982, Phys. Rev. A, 26, 330 [NASA ADS] [CrossRef] [Google Scholar]
  11. Escorza, A., Boffin, H. M. J., Jorissen, A., et al. 2017, A&A, 608, A100 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  12. García, G., & Campos, J. 1988, J. Quant. Spec. Rad. Transf., 39, 477 [NASA ADS] [CrossRef] [Google Scholar]
  13. García-Hernández, D. A., Zamora, O., Yagüe, A., et al. 2013, A&A, 555, L3 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  14. Gontcharov, G. A. 2017, Astron. Lett., 43, 472 [NASA ADS] [CrossRef] [Google Scholar]
  15. Goriely, S., & Mowlavi, N. 2000, A&A, 362, 599 [NASA ADS] [Google Scholar]
  16. Goriely, S., & Siess, L. 2018, A&A, 609, A29 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  17. Gustafsson, B., Edvardsson, B., Eriksson, K., et al. 2008, A&A, 486, 951 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  18. Heiter, U., Lind, K., Bergemann, M., et al. 2021, A&A, 645, A106 [EDP Sciences] [Google Scholar]
  19. Herwig, F., Bloecker, T., Schoenberner, D., & El Eid, M. 1997, A&A, 324, L81 [NASA ADS] [Google Scholar]
  20. Iben, I., Jr., & Renzini, A. 1983, ARA&A, 21, 271 [NASA ADS] [CrossRef] [Google Scholar]
  21. Jorissen, A., Frayer, D. T., Johnson, H. R., Mayor, M., & Smith, V. V. 1993, A&A, 271, 463 [NASA ADS] [Google Scholar]
  22. Jorissen, A., Boffin, H. M. J., Karinkuzhi, D., et al. 2019, A&A, 626, A127 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  23. Käppeler, F., Gallino, R., Bisterzo, S., & Aoki, W. 2011, Rev. Mod. Phys., 83, 157 [Google Scholar]
  24. Karinkuzhi, D., Van Eck, S., Jorissen, A., et al. 2018, A&A, 618, A32 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  25. Karinkuzhi, D., Van Eck, S., Goriely, S., et al. 2021, A&A, 645, A61 [EDP Sciences] [Google Scholar]
  26. Kupka, F., Piskunov, N., Ryabchikova, T. A., Stempels, H. C., & Weiss, W. W. 1999, A&AS, 138, 119 [NASA ADS] [CrossRef] [EDP Sciences] [MathSciNet] [PubMed] [Google Scholar]
  27. Kurucz, R. L. 2007, Robert L. Kurucz On-line Database of Observed and Predicted Atomic Transitions [Google Scholar]
  28. Lebzelter, T., Mowlavi, N., Marigo, P., et al. 2018, A&A, 616, L13 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  29. Lindegren, L., Bastian, U., Biermann, M., et al. 2021, A&A, 649, A4 [EDP Sciences] [Google Scholar]
  30. Little-Marenin, I. R., & Little, S. J. 1979, AJ, 84, 1374 [NASA ADS] [CrossRef] [Google Scholar]
  31. Marigo, P. 2002, A&A, 387, 507 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  32. Martin, G., Fuhr, J., & Wiese, W. 1988, J. Phys. Chem. Ref. Data Suppl., 17 [Google Scholar]
  33. Meggers, W. F., Corliss, C. H., & Scribner, B. F. 1975, Tables of Spectral-line Intensities. Part I, II – Arranged by Elements (US Government Printing Office) [Google Scholar]
  34. Merrill, P. W. 1922, ApJ, 56, 457 [NASA ADS] [CrossRef] [Google Scholar]
  35. Michalik, D., Lindegren, L., & Hobbs, D. 2015, A&A, 574, A115 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  36. Miles, B. M., & Wiese, W. L. 1969, At. Data, 1, 1 [Google Scholar]
  37. Nave, G., Johansson, S., Learner, R. C. M., Thorne, A. P., & Brault, J. W. 1994, ApJS, 94, 221 [NASA ADS] [CrossRef] [Google Scholar]
  38. Neyskens, P., Van Eck, S., Jorissen, A., et al. 2015, Nature, 517, 174 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  39. Nilsson, A. E., Johansson, S., & Kurucz, R. L. 1991, Phys. Scr., 44, 226 [NASA ADS] [CrossRef] [Google Scholar]
  40. O’Brian, T. R., Wickliffe, M. E., Lawler, J. E., Whaling, W., & Brault, J. W. 1991, J. Opt. Soc. Am. B Opt. Phys., 8, 1185 [NASA ADS] [CrossRef] [Google Scholar]
  41. Otto, E., Green, P. J., & Gray, R. O. 2011, ApJS, 196, 5 [NASA ADS] [CrossRef] [Google Scholar]
  42. Palmeri, P., Quinet, P., Wyart, J., & Biémont, E. 2000, Phys. Scr., 61, 323 [NASA ADS] [CrossRef] [Google Scholar]
  43. Palmeri, P., Fischer, C. F., Wyart, J. F., & Godefroid, M. R. 2005, MNRAS, 363, 452 [NASA ADS] [CrossRef] [Google Scholar]
  44. Plez, B. 2012, Astrophysics Source Code Library [record ascl:1205.004] [Google Scholar]
  45. Raskin, G., van Winckel, H., Hensberge, H., et al. 2011, A&A, 526, A69 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  46. Reyniers, M., Winckel, H., Biemont, E., & Quinet, P. 2002, A&A, 395, L35 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  47. Schröder, K.-P., & Cuntz, M. 2007, A&A, 465, 593 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  48. Shetye, S., Van Eck, S., Jorissen, A., et al. 2018, A&A, 620, A148 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  49. Shetye, S., Goriely, S., Siess, L., et al. 2019, A&A, 625, L1 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  50. Shetye, S., Van Eck, S., Goriely, S., et al. 2020, A&A, 635, L6 [CrossRef] [EDP Sciences] [Google Scholar]
  51. Siess, L. 2006, A&A, 448, 717 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  52. Siess, L., & Livio, M. 1999, MNRAS, 304, 925 [NASA ADS] [CrossRef] [Google Scholar]
  53. Siess, L., Dufour, E., & Forestini, M. 2000, A&A, 358, 593 [Google Scholar]
  54. Smith, V. V., & Lambert, D. L. 1985, ApJ, 294, 326 [NASA ADS] [CrossRef] [Google Scholar]
  55. Smith, V. V., & Lambert, D. L. 1988, ApJ, 333, 219 [NASA ADS] [CrossRef] [Google Scholar]
  56. Smith, V. V., & Lambert, D. L. 1990, ApJS, 72, 387 [NASA ADS] [CrossRef] [Google Scholar]
  57. Sneden, C., Lucatello, S., Ram, R. S., Brooke, J. S. A., & Bernath, P. 2014, ApJS, 214, 26 [Google Scholar]
  58. Stephenson, C. B. 1984, Publ. Warner Swasey Obs., 3, 1 [Google Scholar]
  59. Uttenthaler, S., & Lebzelter, T. 2010, A&A, 510, A62 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  60. Uttenthaler, S., Lebzelter, T., Palmerini, S., et al. 2007, A&A, 471, L41 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  61. Van Eck, S., & Jorissen, A. 1999, A&A, 345, 127 [NASA ADS] [Google Scholar]
  62. Van Eck, S., Jorissen, A., Udry, S., Mayor, M., & Pernier, B. 1998, A&A, 329, 971 [NASA ADS] [Google Scholar]
  63. Van Eck, S., Neyskens, P., Jorissen, A., et al. 2017, A&A, 601, A10 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  64. Vanture, A. D., Smith, V. V., Lutz, J., et al. 2007, PASP, 119, 147 [NASA ADS] [CrossRef] [Google Scholar]
  65. Vassiliadis, E., & Wood, P. R. 1993, ApJ, 413, 641 [Google Scholar]
  66. Wang, X. H., & Chen, P. S. 2002, A&A, 387, 129 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  67. Wenger, M., Ochsenbein, F., Egret, D., et al. 2000, A&AS, 143, 9 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  68. Xu, H. L., Svanberg, S., Quinet, P., Garnir, H. P., & Biémont, E. 2003, J. Phys. B At. Mol. Phys., 36, 4773 [NASA ADS] [CrossRef] [Google Scholar]

Appendix A: Reliability of the S-star masses

A.1. Height above the galactic plane

Figure A.1 presents the height above the galactic plane (|Z|) of our sample stars. Although a large Z scatter is present at low masses, the maximum |Z| value decreases with increasing mass, as expected. This trend somehow validates the masses estimated in the present paper. Nevertheless, the limited size of the present sample, and the absence of bias control in this sample, precludes us from drawing any further conclusion.

A.2. Gaia-2MASS photometry

Lebzelter et al. (2018) present a classification of AGB stars using Gaia and 2MASS photometry, in the plane WRP, BP − RP − WK, J − K vs MK, where MK is the absolute KS magnitude from 2MASS and Gaia EDR3 parallaxes. The Wesenheit functions WRP, BP − RP and WK, J − K are calculated using the definitions from Lebzelter et al. (2018), that is, WK, J − K = KS − 0.686 (J − K) and WRP = GRP − 1.3 (GBP − GRP), with J and KS from 2MASS, and GBP and GRP are the magnitudes in the GaiaBP and RP bands. In Fig. A.2, we plot our sample stars in this classification scheme. Most of our sample stars with masses smaller than 2.5 M⊙ lie, as expected, in the “oxygen-rich low-mass AGB” zone. While an oxygen-rich chemistry is indeed expected for S-type stars (which have C/O smaller than unity), their location in the low-mass zone constitutes a fine confirmation of the masses derived in the present work. The case of HD 357941, which was flagged as a low-mass star (Mini ∼ 1 M⊙) based on DR2 parallaxes in S19, is noteworthy. We re-evaluated its mass with Gaia EDR3 parallaxes and found it to be Mini = 3.5 M⊙. This revised mass is now consistent with its location in Fig. A.2 (MK = −7.94, WRP, BP − RP − WK, J − K = −0.03), where it is located at the border between the low-mass and intermediate-mass O-rich AGB stars.

thumbnail Fig. A.1.

Height above the galactic plane for the sample intrinsic stars and for the intrinsic stars from S18, S19 and S20, as a function of their initial mass (left panel) or luminosity (right panel).

thumbnail Fig. A.2.

The WRP, BP − RP − WK, J − K vs MK diagram of our sample of Tc-rich S stars. The definitions of the boundaries (black dashed lines) are from Lebzelter et al. (2018).

Appendix B: Error analysis of the Tc and Li abundances

Uncertainties on the Li and Tc abundances of V915 Aql are listed in the bottom panel of Table B.1. In the upper panel, model A designates the adopted model for V915 Aql from S18, whereas models B-G correspond to models differing by one grid step from model A, each parameter varied at a time. The abundances resulting from each of these models is then compared with the abundance from model A and these differences are listed as columns ΔB − A, ..., ΔG − A in the bottom panels of Table B.1. Model H is the one used to compute the error on the abundances of V915 Aql, as described in Sect. 5.8 (or Sect. 4.4 of S18). The contribution from the atmospheric parameter uncertainties on the error on the Li and Tc abundances is given by ΔH − A in Table B.1 as described in Sect. 5.8.

Table B.1.

Sensitivity of the Li and Tc elemental abundances of V915 Aql upon variations of its atmospheric parameters.

Appendix C: Elemental abundances of sample stars

In this section, we present the tables listing the elemental abundances for the stars of our current study. Table C.1 lists the Tc and Li abundances in our sample stars and also in the Tc-rich stars taken from S18, S19, and S20 that we computed during the current study. Tables C.2–C.4 provide the full list of elemental abundances in our sample stars. Table C.5 lists the different elemental indices [ls/Fe], [hs/Fe], [hs/ls], and [s/Fe] for our sample stars as well as for Tc-rich stars from S18, S19, and S20.

Table C.1.

Tc and Li abundances of the sample stars as well as of the stars from S18, S19, and S20.

Table C.2.

Elemental abundances of sample stars, along with the standard deviation due to line-to-line scatter.

Table C.3.

Elemental abundances of sample stars, along with the standard deviation due to line-to-line scatter (cont’d).

Table C.4.

Elemental abundances of sample stars, along with the standard deviation due to line-to-line scatter (cont’d).

Table C.5.

The heavy (hs) and light (ls) s-process indices of the sample stars.

Appendix D: Revised stellar parameters and abundances of stars from S18, S19, S20

As explained in Sect. 2, the Gaia EDR3 parallaxes became available in the course of our current study. Hence, we revised the stellar parameters of Tc-rich and Tc-poor stars from S18, S19, and S20 using the Gaia EDR3 parallaxes. In Table D.1 we list these revised stellar parameters, which were derived using the same method as described in Sect. 4. In Table D.2 we present the re-computed elemental abundances of BD +34°1698 and HD 357941, which are the only two stars for which the EDR3 parallaxes imposed a revision of log g (see Table D.1).

Table D.1.

Atmospheric parameters for S stars from S18, S19, and S20.

Table D.2.

Elemental abundances for the sample stars, along with the standard deviation due to line-to-line scatter.

Appendix E: Spectral windows used for spectral fitting in Sect. 4

Table E.1 lists the different spectral windows used in the spectral fitting routine to derive an initial estimate of the stellar parameters.

Table E.1.

Spectral windows used to compare observed spectra with MARCS synthetic spectra of S stars.

Appendix F: Atomic line list

Table F.1 lists the lines used for the abundance analysis.

Table F.1.

Atomic lines used in this study.

All Tables

Table 1.

Tc-rich S stars excluded from the current study because of their large photometric variability.

In the text
Table 2.

Basic data for our S star sample.

In the text
Table 3.

Atmospheric parameters for S stars.

In the text
Table 4.

Observed lowest luminosities of intrinsic S stars in different mass and metallicity bins (collected from this work, S18, S19, and S20).

In the text
Table B.1.

Sensitivity of the Li and Tc elemental abundances of V915 Aql upon variations of its atmospheric parameters.

In the text
Table C.1.

Tc and Li abundances of the sample stars as well as of the stars from S18, S19, and S20.

In the text
Table C.2.

Elemental abundances of sample stars, along with the standard deviation due to line-to-line scatter.

In the text
Table C.3.

Elemental abundances of sample stars, along with the standard deviation due to line-to-line scatter (cont’d).

In the text
Table C.4.

Elemental abundances of sample stars, along with the standard deviation due to line-to-line scatter (cont’d).

In the text
Table C.5.

The heavy (hs) and light (ls) s-process indices of the sample stars.

In the text
Table D.1.

Atmospheric parameters for S stars from S18, S19, and S20.

In the text
Table D.2.

Elemental abundances for the sample stars, along with the standard deviation due to line-to-line scatter.

In the text
Table E.1.

Spectral windows used to compare observed spectra with MARCS synthetic spectra of S stars.

In the text
Table F.1.

Atomic lines used in this study.

In the text

All Figures

thumbnail Fig. 1.

Spectral region around the three (4238.19, 4262.27 and 4297.06 Å) Tc I lines in our sample of Tc-rich S stars. For comparison purposes, the spectrum of a Tc-poor S star (HD 233158, in blue in the top panels) from S18 is also plotted. The spectra have been arbitrarily normalized and binned by a factor of 1.5 to increase the S/N. The Tc I line at 4238.19 Å could not be used for CD −27°5131 as local normalization was hampered by a nearby cosmic ray hit.
In the text
thumbnail Fig. 2.

Illustration of the fit quality between the observed and synthetic spectra of BD +79°156 (top panel) and V679 Oph (bottom panel) around the Li line at 6707.8 Å.
In the text
thumbnail Fig. 3.

Illustration of the fit between the observed and synthetic spectra for the S star HR Peg. Top panel: presents 5 Å around the Zr I line at 7819.37 Å and bottom panel: around the Ce II line at 8772.135 Å.
In the text
thumbnail Fig. 4.

HR diagram of intrinsic (filled triangles) S stars from our large sample and extrinsic (open squares) S stars from S18 along with the STAREVOL evolutionary tracks corresponding to the closest metallicities. The red giant branch is represented in blue, the core He-burning phase in green, whereas the red dashed lines correspond to the AGB tracks. The black dashed line represents the predicted onset of the third dredge-up, i.e., the lowest stellar luminosity following the first occurrence of a TDU episode (down to 1.5, 1.3, 1.0 M⊙ for [Fe/H] = 0.0, −0.25, −0.50 respectively).
In the text
thumbnail Fig. 5.

Measured s-process element abundances compared with the nucleosynthesis predictions. The blue line represents the nucleosynthesis predictions compatible with the mass and metallicity of the individual stars. The number of pulses n required to best-match the measured abundances is mentioned in the label of every panel as “p.n”.
In the text
thumbnail Fig. 6.

Distribution of the carbon abundance of our sample stars with respect to their [s/Fe] index. Top panel: measured [C/Fe] abundances as a function of [s/Fe] for our large sample of Tc-rich S stars (light green, purple and orange squares) where the size of the symbol increases with decreasing metallicity, considering three metallicity bins: [Fe/H] in [0.0; −0.2]: light green, [−0.2; −0.4]: purple, below −0.4: orange. Tc-rich carbon stars from Abia et al. (2002) are shown as green squares. Bottom panel: derived C/O ratio versus [s/Fe] for the same stars. Predictions for a 2 M⊙ STAREVOL model at metallicity [Fe/H] = 0.0 (light green), −0.3 (violet) and −0.5 (orange) are overplotted. The empty circles along the tracks indicate the successive TDUs, while the three filled circles mark the first TDU, which led to C/O > 1 in each model.
In the text
thumbnail Fig. 7.

Distribution of [hs/ls] as a function of metallicity for the intrinsic S stars of our study (filled squares). The hs-index has been calculated using Ba and Nd, and the ls-index using Y and Zr. The symbols are color-coded with respect to the initial mass in the bins Mini < 1.5 M⊙ (blue), 1.5 ≤ Mini < 2.5 M⊙ (red), and Mini ≥ 2.5 M⊙ (green). We note that these mass bins are designed to match the [hs/ls] STAREVOL predictions only available for Mini = 1.5, 2, 2.5 M⊙. Open circles denote the predictions from pulse to pulse for the different initial masses.
In the text
thumbnail Fig. 8.

[s/Fe] index versus Tc abundance. The different colors represent different C/O ratio bins as described in the figure.
In the text
thumbnail Fig. 9.

Stellar luminosity as a function of Tc abundance. The symbols are color coded according to their mass bin. The blue and red dashed lines represent the linear least-squares fit for the stars in the corresponding mass bin.
In the text
thumbnail Fig. 10.

[Zr/Fe] as a function of [Nb/Fe] for the intrinsic S stars (black filled triangles) and extrinsic S stars (open squares) from S18 compared with the nucleosynthesis predictions for different initial masses and metallicities. The bitrinsic S stars from S20 are plotted with a blue star symbol. The “ext” and “int” labels next to the model parameters refer to extrinsic (respectively intrinsic) model abundances.
In the text
thumbnail Fig. A.1.

Height above the galactic plane for the sample intrinsic stars and for the intrinsic stars from S18, S19 and S20, as a function of their initial mass (left panel) or luminosity (right panel).
In the text
thumbnail Fig. A.2.

The WRP, BP − RP − WK, J − K vs MK diagram of our sample of Tc-rich S stars. The definitions of the boundaries (black dashed lines) are from Lebzelter et al. (2018).
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.