s-PROCESSING IN THE GALACTIC DISK. I. SUPER-SOLAR ABUNDANCES OF Y, Zr, La, AND Ce IN YOUNG OPEN CLUSTERS

The sample includes 19 OCs with ages in the range of ∼0.1–8.4 Gyr and Galactocentric distances (R_GC) between 7.05 and 22 kpc. This sample overlaps considerably with the one from D'Orazi et al. (2009); therefore, we refer to this paper and to the references therein for additional information on the clusters in common and on their spectroscopy. In particular, we adopt here the same cluster ages used by D'Orazi et al. (2009); these are on a homogeneous scale and were taken from Magrini et al. (2009).

A few differences are, however, present in our sample; namely, we do not include the young clusters IC 2602, IC 2391, and NGC 6475, since the available spectral range did not cover useful lines for the abundance analysis. On the other hand, we added three new clusters: IC 4756, NGC 5822 (Pace et al. 2010), and NGC 6192 (Magrini et al. 2010). The spectra of the stars in these clusters were all collected with UVES consistently with the rest of the sample. Details on the observations and instrument configuration can be found in Pace et al. (2010) and Magrini et al. (2010). Also, and most important, the adopted ages for the clusters we study are on the same scale as those of D'Orazi et al. (2009).

The full list of sample clusters is given in Table 1, as are their ages (Column 2), Galactocentric distances (Column 3), metallicities (Column 4), number of stars analyzed, and spectral regions considered. We analyzed spectra of main-sequence (MS) stars for the first eight OCs in the table, and spectra of red giant branch (RGB) or clump stars for the remaining ones, where MS stars are too faint to be observed with UVES.

In Table 2 we list the lines of Y, Zr, Ce, and La used in the present analysis, along with the atomic parameters and their references. All the lines were inspected very carefully on the UVES spectrum of the Sun, and, in particular, the presence of possible blends was checked. Note that all lines belong to ionized species. Following previous works in the literature, we did not consider hyperfine structure (HFS) for Ce, Y, and Zr (Lawler et al. 2009; Hannaford et al. 1982; Sneden 1973). On the other hand, La needs HFS calculations, and we adopted those used in Lawler et al. (2001). None of the lines is known to be affected by non-LTE (NLTE) effects.

As anticipated above, for all the elements used in this analysis, we selected only the spectral features that can be detected in the UVES spectrum of the Sun, in order to be able to derive differential abundances line-by-line with respect to the Sun itself.

Equivalent widths (EWs) of the lines listed in Table 2 were measured using the splot routine within the IRAF package.⁴ Whenever possible, measurements were performed by means of direct integration of lines instead of fitting with Gauss' or Voigt's profiles.

Also for the Sun, abundances were derived from EW measurements using the MOOG code (Sneden 1973; we adopted the 2009 version) and the grid of one-dimensional local thermal equilibrium model atmospheres by Kurucz (1993). Stark and radiative broadening were treated in the standard way, and for collisional damping we adopted the classical Unsöld approximation. In Columns 4 and 5 of Table 2 we show our EW measurements for the Sun along with abundances in the usual form (A(X) = log[N(X)/N(H)]+12). EWs were measured on the solar spectrum obtained with UVES. As in Randich et al. (2006), we adopted the following Solar parameters: T_eff = 5770 K, log g = 4.44, and ξ = 1.1 km s⁻¹.

Initially, we had also derived abundances using lines in the wavelength interval 300–390 nm, but the resulting values are systematically lower (∼0.3 dex) than those from lines with λ  >  390 nm. Since the same behavior is also obtained by performing spectral synthesis (C. Sneden 2010, private communication), we believe that some problems in the opacity calculation arise in this UV region. Therefore, we decided to discard these lines, limiting our analysis to those with wavelengths larger than 390 nm.

The mean solar abundances for the four elements derived from our analysis are shown in Table 3, together with the values derived by Asplund et al. (2009, hereafter AS09) using three-dimensional models. The meteoritic abundances from Lodders et al. (2009) are also indicated.

Our abundances for Y, Zr, and Ce are in very good agreement with both the meteoritic values and AS09 determinations. On the other hand, the abundance of La derived in this study agrees with the value of AS09, but both are lower than in the meteoritic data. Although the difference is within the uncertainties (see Table 3), it is larger than the differences we found for other elements and indicates that additional studies on the solar lanthanum are required.

We applied the same method used for the Sun to measure abundances in OC stars. Stellar parameters for the sample stars were taken from the reference papers given in Table 1, while in Table 4 we show their values. We underline that all the stellar parameters of this study were determined spectroscopically. As mentioned, our analysis of OC stars is strictly differential with respect to the Sun; thus, for each line we subtracted from the derived abundance the corresponding solar abundance from the same line, obtaining [X/H]_line. [X/H] values for each star were then obtained as the average of [X/H]_line values. In this way, the uncertainties of the adopted set of atomic parameters, in particular log gf values, cancel out. The [X/Fe] estimate for each star was finally derived by subtracting the [Fe/H] value of the star itself (rather than the mean metallicity of the cluster) from its [X/H] value.

Random errors include the contributions from errors of EW measurements and from uncertainties of stellar parameters.

Errors due to EW measurements in general correspond to the line-to-line abundance scatter for a given element, since the possible scatter due to uncertainties in log gf values is minimized by the use of differential analysis with respect to the Sun. However, when fewer than three lines were available, we obtained five EW measurements and computed the corresponding average [El/H] values; the standard deviation of the measurements was then adopted as the error in [El/H].

As is usually done, errors due to uncertainties in stellar parameters were estimated by varying each parameter separately while leaving the others unchanged. The results are shown in Table 5; we performed this analysis for four stars that fairly well cover the parameter space of the whole sample. Uncertainties in effective temperature, surface gravity, and microturbulence were estimated following the discussions in Sestito et al. (2006) and Randich et al. (2006). We found that variations of 100 K in T_eff neither produce significant changes in the abundances, nor in red giants, nor in MS stars. On the other hand, and as is expected from an analysis based on lines from ionized elements, a 0.25 dex difference in log  g leads to a variation of ±0.1 dex in the [X/H] value. Finally, changes of ±0.15 km s⁻¹ in the microturbulence parameter ξ of red giants have a rather large effect (about 0.15 dex) for Y, while they are less significant (∼0.05 dex) for Ce (note that we do not have measurements for Zr and La in red giants). The effect of a different ξ is smaller for dwarf stars.

Systematic errors might be due to different effects such as the spectrocode, model atmospheres, or estimates of the continuum in the spectra, and of log gf values. While, as mentioned, the effect of the latter should be minimal due to the use of the differential analysis, the other three effects are more difficult to estimate. For three of the sample clusters, measurements of one or more of the elements analyzed in this paper are available in the literature, namely M67 (Tautvaisiene et al. 2000; Yong et al. 2005; Pancino et al. 2010; Önehag et al. 2011), Be 29 (Yong et al. 2005), the Hyades and Collinder 261 (De Silva et al. 2007), and IC 4651 (Mikolaitis et al. 2011). For one cluster only, M67, more than one s-process determination is available in the literature. In Figure 1 we compare our abundances for this cluster with three literature values. The figure indicates that our estimates generally agree within the errors and within ±0.05 dex with previous determinations; however, the two Zr measurements from the literature are considerably lower than our own value. We believe that the latter is likely more correct, since the significantly undersolar [Zr/Fe] derived by Tautvaisiene et al. (2000) and Yong et al. (2005) would not be consistent with the close-to-solar abundance ratios for Y, La, and Ce found by us and by other authors, and, in the more general case, with the solar composition of all the other elements. In any case, possible systematic errors would not affect our results, which are based on the comparison of clusters analyzed with the same method, and thus on the same abundance scale.

Zoom In Zoom Out Reset image size

2.5.1. Gravity Uncertainty

In Figure 2 we plot [Ce/H] versus [Y/H] for all the stars for which both elements were measured. Filled and open circles indicate dwarf and giant stars, respectively. The figure shows that the data points for dwarfs are clustered around the 1:1 relationship and are thus consistent with a similar average enrichment in Y and Ce, within uncertainties; on the other hand, the points of the giants have systematically lower Ce abundances than Y. Interestingly, the same also holds for the Hyades where dwarf stars were analyzed and where indeed Y appears to be more abundant than Ce. In these stars, the 508.743 nm yttrium line gives [Y/H] values consistent with [Ce/H] and systematically smaller than those from the other yttrium lines. The discrepancy is larger than 0.3 dex in some stars, characterized by low temperature (below 4500 K), namely, 2, 6, 7 in Cr 261 and 1240 in Be 20, and again the [Y/H] from the 508.743 nm is consistent with [Ce/H] (see Table 4). Moreover, in all these stars the yttrium EWs are larger than 60 mÅ. In order to find some explanations for these observations, we considered the values [Y/Fe] derived from one yttrium line only (508.743 nm) for four OCs: M67, Be 32, Mel 66, and NGC 6253. These clusters have ages between 4 and 6 Gyr. Figure 3(a) shows our EW measurements of the men- tioned line versus the corresponding [Y/Fe] values. The points scatter around the center at [Y/Fe] ≈ 0. Performing the same exercise with the 520.041 nm yttrium line, a very different behavior appears. We show this in Figure 3(b). The M67 stars are centered on the [Y/Fe] ≈ 0 value, and indeed the EWs are smaller than 60 mÅ. The other points have larger EWs and are considerably shifted to the right side of the plot, hence to higher abundances, forming a plateau. For the 520.041 nm, the same happens for the remaining yttrium lines. The strict and systematic correlation between this shift to higher abundances and the EW values for which it happens (EWs larger than about 60 mÅ), together with the shape of the trend of Figure 3(b), are typical effects of saturation. Therefore, in the following we shall use Y abundances from the 508.743 nm line only for the evolved stars and for the Hyades members, because this line does not suffer from saturation for the EW values we measured.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 4, where we plot [X/Fe] ratios versus cluster age, summarizes our results for the evolution with age of the four elements analyzed in this study. The figure clearly shows that all four elements, and specifically their logarithmic ratios to Fe, increase with decreasing cluster age. All [El/Fe] ratios are consistent with the solar value, or are slightly below it, in old clusters. A steep growth is then seen between the ages of ∼1.5 and 0.5 Gyr, after which the [El/Fe] values remain constant around a plateau value of +0.2 dex. The distribution and growth of the first-peak elements are very similar to those of the second-peak ones. Also, although lanthanum and zirconium have a smaller number of points than yttrium and cerium because of the available wavelength range of the star spectra of each cluster, the measured [Zr/Fe] ratios are in very good agreement with [Y/Fe] in the clusters where both elements could be measured; the same happens for lanthanum and cerium. Our results confirm and strengthen those previously obtained by D'Orazi et al. (2009) for barium; indeed, our data are based on much higher statistics, as shown in Figure 5. This, in turn, makes clear that there is a need for an s-element production from very low mass AGB stars (with large ages, hence capable of affecting young clusters, not old ones) more effective than previously imagined. Together with the general trend of increasing s-element abundances in young clusters, some other features appear in the figure. First, we notice that at all ages a dispersion in s-process abundance is present. In particular, the Praesepe cluster lies significantly below the main trend and is less enriched than, e.g., the coeval Hyades cluster. On the other hand, three clusters with ages around 0.8–1 Gyr, NGC 2660, NGC 5822, and NGC 3680, are more enriched than other members of the sample with similar ages. We have no secure explanation for this dispersion. We suggest that it might not be intrinsic in part. Rather, it may be due to incorrect [Fe/H] determinations for these clusters. One example of this fact is probably shown by Praesepe, for which a high metallicity ([Fe/H] = 0.27) was measured (Pace et al. 2008). Its [El/H] values are in general agreement with those of other coeval clusters, but due to the high [Fe/H] estimate, the [El/Fe] ratios become much smaller. The high [Fe/H] value of Praesepe is actually rather strange for its Galactocentric radius, R_GC. This peculiarity makes Praesepe a very interesting target for new measurements of the metallicity. On the other hand, part of the dispersion in s-process elements might instead be intrinsic and be related to the different positions in the disk of the sample clusters (see the labeled OCs of Figure 4 NGC 6192, IC 4651, NGC 6253, Cr 261 and the following paragraph).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We finally note the very high [Y/Fe] and [Ce/Fe] ratios of the old cluster Berkeley 29, which are both extremely above the trend. We will discuss the case of this cluster in the following section.

4.1.1. [El/H] versus R_GC

The s-element enhancement can also be put in evidence by looking at the trend of [El/H] with R_GC for clusters in different age intervals. This is shown in Figure 6, where the trends of [Y/H] and [Ce/H] with the Galactocentric radius are plotted. The figure suffers from low number statistics, in particular at large Galactocentric distances. Nevertheless it clearly shows that s-elements are characterized by a gradient along the Galactocentric radius similar to the one displayed by iron, iron-peak, and α elements (see Magrini et al. 2009). Most noticeably, two different gradients actually seem to be present. One is traced by older OCs (open circles of the figure), and one characterizes the younger objects (filled circles). The latter lies above the previous one, along the entire range of R_GC (see Figure 7), suggesting that the increase in the s-element enrichment occurs across the whole Galactic disk, once very low mass stars with long evolutionary timescales have started to contribute.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Moreover, looking at Figure 7, the four labeled clusters with R_GC smaller than ∼7.7 kpc show a different behavior with respect to the other OCs. These clusters, Cr 261, NGC 6253, IC 4651, and NGC 6192, are also labeled in Figure 4, where the trend they trace remains slightly below the main one. A possible explanation could be related to the infall process that occurred during the disk formation. Indeed, a rapid enrichment of iron in the inner zone of the disk, favored by mass infall from the halo, could shift the peak of star formation there toward higher [Fe/H] and [O/H]; the zones with larger R_GC would be instead characterized by a star formation rate peaked at lower [Fe/H] values. Since a larger amount of iron and oxygen makes the occurrence of the third dredge-up in AGB stars more difficult and reduces neutron exposure (see Straniero et al. 2003 and references therein), it is reasonable that in the inner zone of the disk the s-element production was more limited. Further studies are required to extend the statistics of s-element measurements in the inner OCs (R_GC < 7.5 kpc), while chemical evolutionary calculations at different R_GC could verify the reliability of the hypothesis outlined above.

4.1.2. Berkeley 29

As mentioned above, both yttrium and cerium in this cluster show an anomalous enrichment. This result is not in agreement with the trend shown by other clusters, as is evident in Figure 4. Indeed, the age (4.3 Gyr) of this cluster should imply a solar [El/Fe] value. Members of this cluster are characterized by a dispersion in both yttrium and cerium, ranging from moderate to extreme enrichment.

In order to further check the high abundances by means of a differential comparison, we superimposed the spectra of two stars with similar parameters, namely, star 933 of Be 29 and star 938 of Be 32.

Figure 8 shows an example of this comparison for the yttrium line λ = 508.742 nm and supports the anomalous s-element enrichment found in Be 29. The signal-to-noise ratios of the two spectra (40–50 for 933, 50–60 for 938) do not allow us to be confident of the reliability of our EW measurements. Iron and α-element abundances of this cluster (see Sestito et al. 2008) have values that are compatible with the solar ones, so that our s-element abundances look strange. On the other hand, in favor of the enrichment there is the fact that all the analyzed lines of the four stars in Be 29 indicate an overproduction of yttrium and cerium. We note that Be 29 is also enriched in Ba (see Figure 5). Further investigations on s-element abundances in outer OCs are required and could become relevant clues for the knowledge of these external zones of the disk and their evolution.

Zoom In Zoom Out Reset image size

4.1.3. Clusters Studied in the Literature

Several recent studies focusing on abundance determinations in OCs include the measurement of one or more s-process elements. In Figure 9 we show the evolution of Y and Zr (upper panel) and Ce and La (bottom panel) obtained by using the most recent abundance determinations. Note that cluster ages are (or have been put by us) on the same scale as that used for the OCs discussed in this paper.

Zoom In Zoom Out Reset image size

Measurements of s-process elements from the literature are highly non-homogeneous and difficult to compare with the present study because different lines, stars, and methods have been used. Nevertheless, the figure shows an overall fair agreement, as data from the literature follow the same trend displayed by our own measurements. The agreement is particularly good for Y, La, and Ce, while Zr is more scattered; specifically, a couple of relatively young clusters (NGC 1883 and NGC 7789) have a low [Zr/Fe] ratio, while a few very old ones (NGC 7142, NGC 1193, NGC 188, Be 31) are characterized by enhanced ratios. We note that for NGC 7789, enhanced [Y/Fe], [La/Fe], and [Ce/Fe] have been reported, so we believe that Zr was underestimated by Tautvaisiene et al. (2005). Zr determinations for NGC 1193, NGC 188, NGC 7142, and Be 31 come from Friel et al. (2010) and Jacobson et al. (2008). In both papers the abundances of Zr were based on at most two Zr i lines; large variations are seen within each cluster and, as stressed by Friel et al. (2010), abundances are highly uncertain. While new measurements of Zr and of the other s-process elements for all these clusters are needed, we conclude that the data from the literature generally confirm the main results of our paper.

Table 4 shows the presence of one anomalous star, which is significantly more s-element enriched than the other members of the same cluster. This star is 941 in Be 32. The discrepancy between its [El/H] values with respect to the mean of the cluster to which it belongs is evident. This enhancement is present both for Y and Ce, thus indicating that it is most likely real and not due to problems with the analysis.

In order to further investigate this point, we superimposed the spectra of this anomalous star to that of another star with similar stellar parameters and belonging to the same cluster, namely, star 19; this allows us to check if the discrepancy occurs because of possible errors in the determination of parameters like the star gravity or temperature. In Figure 10 we show an example of this exercise. The temperatures of the two objects are equal, while their log  g differs by 0.1 dex and their metallicities differ by 0.06 dex. The two spectra are perfectly in agreement with each other, with the exception of the Y ii line λ = 508.7 nm, which is stronger in the 941 spectrum. The agreement in the other regions of the spectra indicates that errors in stellar parameter determinations cannot be responsible for the anomalies in the s-element abundances found in 941. We performed this test for the other cerium lines, finding the same results as for yttrium. We thus conclude that the enrichment of this star is real.

Zoom In Zoom Out Reset image size

The above anomaly might be understood attributing the s-element enhancement to a mass transfer phenomenon in a binary system. If the anomalous star has a white dwarf companion, we might suppose that this was once the primary component of the system, with an intermediate mass (e.g., 3–4 M_☉ or more), and that it underwent s-processing during its AGB phase. Such a star would not have activated efficiently the ¹³C neutron source, but it would have nevertheless produced neutrons through the ²²Ne(α, n) ²⁵Mg reaction in the thermal pulses. This neutron source is actually the dominant one for all intermediate-mass (Lugaro et al. 2007) and high-mass stars (Busso & Gallino 1985; Raiteri et al. 1992). In this scenario, the secondary component we see today would be the equivalent of a Ba-star, i.e., it would have accreted s-process-rich material from the AGB primary.