Extreme r-process Enhanced Stars at High Metallicity in Fornax^*

We present four Fornax stars, including three with anomalous high Eu (1.25 ≤ [Eu/Fe] ≤ 1.45). The quality of the available observations allows for an accurate and precise determination of abundances (typically to within ±0.2 dex). The stars were observed at high and low resolution with FLAMES/GIRAFFE (Pasquini et al. 2002) with the setups HR10, HR13, HR14, and LR8 (see Table 1). The LR8 setup (Battaglia et al. 2006) is used here only for radial velocity determinations. We use reduced spectra from the ESO archive ⁶ and perform the data processing (sky correction, radial velocity shifts, co-adding, and normalization of the spectra) as in Reichert et al. (2020).

Table 1. Different FLAMES/GIRAFFE Setups Showing Minimum and Maximum Wavelength Coverage and Resolving Power (R)

Grating	${\lambda }_{\min }$ (Å)	${\lambda }_{\max }$ (Å)	R
HR10	5339	5619	19,800
HR13	6120	6405	22,500
HR14	6308	6701	17,740
LR8	8206	9400	6500

Download table as:  ASCIITypeset image

Our sample consists of Fornax stars is from Reichert et al. (2020), who analyzed 380 stars in dSph galaxies. The stars have been extracted from the reduced ESO archive and the Keck Observatory Archive. More specifically, we assigned a star membership if its coordinates agree within a circle of three arcseconds with the coordinates in the SIMBAD (Wenger et al. 2000), NASA/IPAC Extragalactic Database (NED; NED Team 2007), or the second Gaia data release (DR2; Gaia Collaboration 2018a, 2018b).

Therefore, the three Eu-enhanced stars are clearly members of Fornax as previous studies have already assigned them Fornax membership based on radial velocities, and distance to the center of Fornax (see Figure 1 and Battaglia et al. 2006; Lemasle et al. 2014). Also the proper motions of Gaia DR2 (Gaia Collaboration 2018a, 2018b) indicate their membership (see Table 2).

Zoom In Zoom Out Reset image size

Table 2. Proper Motions in the R.A. Direction μ_δ , and in the Decl. Direction ${\mu }_{\alpha }\cos \delta$ Taken from Gaia DR2 (Gaia Collaboration 2018a, 2018b)

Object	μδ (mas yr−1)	${\mu }_{\alpha }\cos \delta$ (mas yr−1)
Fnx-mem0546	0.088 ± 0.145	−0.199 ± 0.210
Fnx-mem0556	0.400 ± 0.148	−0.775 ± 0.203
Fnx-mem0595	0.344 ± 0.120	−0.237 ± 0.167
Fornax (mean)	0.038 ± 0.003	−0.416 ± 0.004

Note. Furthermore, we list the mean values of Fornax (McConnachie & Venn 2020).

Download table as:  ASCIITypeset image

The [α/Fe] ratio of the stars also fits with the general trend of Fornax (Figure 2) and similar to many other r-process enriched metal-rich stars, a clear contribution from Type Ia SNe is visible.

Zoom In Zoom Out Reset image size

The stars were previously analyzed by Lemasle et al. (2014), Reichert et al. (2020), and Battaglia et al. (2006). Table 3 shows their determined effective temperature, surface gravity, and metallicity (for Battaglia et al. 2006 the metallicities are from the calcium triplet). We note that the star Fnx-mem0546 was excluded from Lemasle et al. (2014) due to convergence problems when determining the stellar parameters. Here, we adopt stellar parameters from Reichert et al. (2020). The effective temperatures and metallicities were derived using the automatic code SP_Ace (Boeche & Grebel 2016), which fits a theoretically calculated spectrum to the observed spectrum of the star until convergence is reached. The theoretical spectrum is constructed by using a previously calculated curve of a growth library. We note that the metallicity is therefore also biased by the local thermodynamic equilibrium (LTE) assumptions.

The surface gravities were derived using

$$\begin{eqnarray}\begin{array}{rcl}\mathrm{log}{g}_{* } & = & \mathrm{log}{g}_{\odot }+\mathrm{log}\displaystyle \frac{{M}_{* }}{{M}_{\odot }}+4\,\mathrm{log}\displaystyle \frac{{T}_{\mathrm{eff},* }}{{T}_{\mathrm{eff},\odot }}\\ & & +0.4\left({M}_{\mathrm{bol},* }-{M}_{\mathrm{bol},\odot }\right),\end{array}\end{eqnarray} \tag{ 1 }$$

with $\mathrm{log}{g}_{\odot }=4.44$, T_eff,⊙ = 5780 K, M_* = 0.8 ± 0.2M_⊙, and M_bol,⊙ = 4.72 mag, and the distance to Fornax of 147 ± 8 kpc taken from Karczmarek et al. 2017).

The microturbulence was calculated with an empirical relation from Kirby et al. (2010):

$$\begin{eqnarray}&&{\xi }_{{\rm{t}}}=((2.13\pm 0.05)-(0.23\pm 0.03)\cdot \mathrm{log}g)\,\mathrm{km}\,{{\rm{s}}}^{-1}.\end{eqnarray} \tag{ 2 }$$

In general, effective temperatures in Reichert et al. (2020) are hotter than in Lemasle et al. (2014) because of the different methods in deriving them (spectroscopic versus photometric, see Appendix B of Reichert et al. (2020) for a more detailed discussion). However, we stress that the strong neutron-capture element enhancement is also present when using the stellar parameters of Lemasle et al. (2014), hence, the choice of stellar parameters does not change our conclusions.

We have derived (neutral) Zr abundances for all Fornax stars presented in Reichert et al. (2020). For many stars it is the first Zr determination, making it the largest sample of Zr abundances in a dwarf galaxy to date. Figure 4 shows the [Zr/Fe] in Fornax, including the neutron-capture enhanced stars. The average Zr abundance for all Fornax stars is slightly subsolar and the reference star (Fnx-mem0607) is located slightly below this general trend. There may be corrections to the abundances due to the LTE assumption, which tends to affect neutral atoms more when they are the minority species as it is the case for Zr (see Andrievsky et al. 2017, for a discussion). The exact size of the nonlocal thermodynamic equilibrium (NLTE) correction depends on several factors such as the stellar parameters and the abundance value. Currently, there is no large, available Zr NLTE grid covering our parameter space. However, based on NLTE computations in Velichko et al. (2010) and more recent ones in Hansen et al. (2020) we would estimate corrections of the order of ∼0.0 to >+0.3 dex, depending on the Zr abundances and stellar parameters. For our three enhanced stars the corrections would likely amount to ≳+0.3 dex.

Zoom In Zoom Out Reset image size

The three Eu-enhanced Fornax stars (Fnx-mem0546, Fnx-mem0556, and Fnx-mem0595) are clearly enhanced also in Zr. This agrees with these stars having an r-process pattern typical of other r-I or r-II stars; see, for example, the Reticulum II stars (teal diamonds) in Figure 4 and the discussion in Section 4.1. Despite the high metallicities and Zr being dominated by the s-process in the solar system (∼66%, Bisterzo et al. 2014), at low metallicities Zr is formed by the r-process as shown by most r-II stars. For the Eu-enhanced stars in Fornax, the enhancement of Zr is probably produced by an r-process event and not by the s-process, see Section 4.1.

The three Eu-enhanced Fornax stars are enriched in Eu with ∼1 dex more than the average Fornax stars (see Figure 5). With $\mathrm{log}\epsilon (\mathrm{Eu})=0.98$, Fnx-mem0595 and Fnx-mem0556 are the most Eu-enriched stars observed to date, together with SCMS 982 that is enriched by an s- or i-process (see Skúladóttir et al. 2020 for a discussion). Even though rare, there are several stars with enhanced Eu at low metallicities. However, such stars are even more rare at high metallicities. In Figure 5, we show some of those Eu-stars: Cos 82 (Shetrone et al. 2001; Sadakane et al. 2004; Aoki et al. 2007), J1124+4535 (Xing et al. 2019), HD 222925 (Roederer et al. 2018b), and SCMS 982 (Geisler et al. 2005; Skúladóttir et al. 2020). With the exception of SCMS 982, these stars show a dominant contribution from the r-process as indicated by their [Ba/Eu] ratio and discussed below (Figure 7).

Zoom In Zoom Out Reset image size

The combination of metal-rich stars together with the extreme enhancements of neutron-capture elements gives us the unique possibility to derive abundances of neutron-capture elements that are otherwise challenging to detect. Therefore, we are able to derive Lu abundances (Figure 6).

Zoom In Zoom Out Reset image size

Due to the heavy blending of the absorption line, we claim a Lu detection if more than three points deviate by more than 3σ (dashed line in Figure 6) from the flux when assuming basically no Lu (i.e., [Lu/H] = −9). Even though uncertain due to irregular noise next to the absorption line, we clearly detect Lu in Fnx-mem0546, but also in Fnx-mem0556 and Fnx-mem0595. Because the continuum shows some artifacts next to the absorption line in Fnx-mem0595, we assign an additional error of 0.1 dex to the lutetium abundance. This adds all three stars to the rare stars with lutetium detections (nine in JinaBase; Abohalima & Frebel 2018; 13 in SAGA; Suda et al. 2017) and makes it the first detections in a dwarf spheroidal galaxy.

At high metallicity, the neutron-capture elements stem mainly from the s- and r-processes, possibly also from the i-process. The i-process operates at intermediate n-densities and exposures and may be able to take place in rapidly accreting white dwarfs (e.g., Denissenkov et al. 2017), super asymptotic giant branch stars (e.g., Doherty et al. 2015), or low-mass AGB (e.g., Stancliffe et al. 2011). In order to check whether the Eu-stars are enriched by an r-process event or whether they have a strong contribution from the s-process, we first check the [Ba/Eu] ratio as shown in Figure 7 and Table 5. For the three stars, the ratio is well below zero and this is often used to define neutron-capture-rich, r-process enhanced stars ([Ba/Eu] < 0, Beers & Christlieb 2005). A stronger constraint on the heavy-element enhancement is the estimate of the pure process trace, which is typically assessed through the [Ba/Eu] being less than ∼ −0.7 dex (pure r) or larger than ∼1.1 dex (pure s) (Arlandini et al. 1999; Hansen et al. 2018). The [Ba/Eu] ratio points toward a pure r-ratio in the three Fornax stars, while the reference star, Fnx-mem0607, is clearly a mixture of r and s.

Zoom In Zoom Out Reset image size

We also probe if there is an i-process contamination. A typical fingerprint of the i-process is a positive [Ba/La] ratio (Hampel et al. 2016; Denissenkov et al. 2019; Koch et al. 2019), but this ratio is negative for the four Fornax stars. In this connection, we note that Fnx-mem0546 has a relatively low Ba abundance with respect to La and also compared to the Ba in the other two stars. The La abundance is more than 1 dex higher than the Ba abundance in Fnx-mem0546. This leads to a strong odd–even abundance difference in this region. In Figure 7, it can also be clearly seen that SCMS 982 is not r-process dominated with [Ba/Eu] close to 0.5. Whether the peculiar pattern of SCMS 982 stems from an s- or i-process is still under debate (Geisler et al. 2005; Skúladóttir et al. 2020). Based on these abundance ratios, we focus our further comparison on r-process and s-process yields.

We now compare the observationally derived abundances to theoretical predictions. For the s-process, we take yields for different AGB star masses and metallicities from the F.R.U.I.T.Y. database (Cristallo et al. 2011, 2015). In Figure 7, this contribution is shown by horizontal dashed and dotted lines covering the metallicity range where they are predicted and appropriate. For the r-process, we compare to yields from different scenarios (for details see Section 4.3). The variation in any abundance ratio within a given scenario is due to the nuclear physics uncertainties (see Eichler et al. 2019; Côté et al. 2021 for more details about the models and nuclear physics included in the nucleosynthesis). As can be seen from Figure 7, the [Ba/Eu] ratios of the three Eu-stars agree with the theoretical predictions of the r-process within uncertainties. However, those uncertainties are rather large and thus it is not possible to conclude which scenario contributed most to the observed abundances.

Now we explore another possibility to check whether the three Eu-star's abundances are mainly due to the r-process. We compare their abundance patterns to that of HD 222925 (Roederer et al. 2018b), a star that has been identified as r-process enhanced. Figure 8 shows this comparison and demonstrates that two of the Eu-stars (Fnx-mem0556 and Fnx-mem0595) fit with an excellent reduced χ² < 1 the r-process pattern of HD 222925. Fnx-mem0546 deviates slightly from this pattern with an enhancement in lanthanum, and praseodymium relative to barium and Eu. Lu on the other hand agrees well with the r-process pattern for Fnx-mem0546 and Fnx-mem0556. For Fnx-mem0595, Lu is underabundant compared to the r-process ratio, however, it also has a larger error (see Section 3.2).

Zoom In Zoom Out Reset image size

In summary, all stars show a striking similarity to the r-process pattern without too many contributions from other processes. We can therefore safely assume that the abundances of the three Eu-stars are mainly enriched by the r-process.

After showing that the Eu-stars are r-process enhanced, we estimate how many r-process events are necessary to enrich these stars and calculate the ejected Eu per event. The three Eu-stars were born in a gas cloud/region with a total mass M_gas. Magg et al. (2020) provided an approximation for the dilution mass where the new elements from an r-process event(s) are mixed:

$$\begin{eqnarray}&&{M}_{\mathrm{gas}}=1.9\cdot {10}^{4}{M}_{\odot }{E}_{51}^{0.96}{n}_{0}^{-0.11},\end{eqnarray} \tag{ 3 }$$

where n₀ is the ambient density and E₅₁ is the explosion energy of the event. Magg et al. (2020) used n₀ = 1 corresponding to the environment where Population III stars formed, however, this has an small impact on the final dilution mass due to the small exponent. The explosion energy (E₅₁) of the event can be assumed to be between 10⁵¹ and 10⁵² erg (see 10⁵¹ erg for GW170817; Kathirgamaraju et al. 2019 and the explosion energy of 10⁵² erg of hypernovae; Nomoto et al. 2006). The dilution mass (M_gas) therefore lies between 10⁴ and 10⁵ M_⊙, which is at least three orders of magnitudes smaller than the total stellar mass of Fornax M_Fornax ∼10⁸ M_⊙ (Łokas 2009).

Before the r-process event(s), the mass of Eu in the gas was ${M}_{\mathrm{Eu}}^{\mathrm{pre} \mbox{-} \mathrm{event}}$ and the mass injected by the event was ${M}_{\mathrm{Eu}}^{{\rm{r}} \mbox{-} \mathrm{event}}$. We can estimate ${M}_{\mathrm{Eu}}^{\mathrm{pre} \mbox{-} \mathrm{event}}$ using the abundance of our reference star, Fnx-mem0607, in the following definition:

$$\begin{eqnarray}&&\mathrm{log}\epsilon (\mathrm{Eu})=\mathrm{log}\displaystyle \frac{{M}_{\mathrm{Eu}}^{\mathrm{pre} \mbox{-} \mathrm{event}}}{{A}_{\mathrm{Eu}}\cdot {M}_{\mathrm{gas}}}+12,\end{eqnarray} \tag{ 4 }$$

where A_Eu = 152 is the average mass number of Eu. Therefore, the Eu in the r-process stars is

$$\begin{eqnarray}&&\mathrm{log}\epsilon (\mathrm{Eu})=\mathrm{log}\displaystyle \frac{{M}_{\mathrm{Eu}}^{\mathrm{pre} \mbox{-} \mathrm{event}}+{M}_{\mathrm{Eu}}^{{\rm{r}} \mbox{-} \mathrm{event}}}{{A}_{\mathrm{Eu}}\cdot {M}_{\mathrm{gas}}}+12,\end{eqnarray} \tag{ 5 }$$

and from this expression we obtain a Eu mass per event ${M}_{\mathrm{Eu}}^{{\rm{r}} \mbox{-} \mathrm{event}}$ between ∼1.5 · 10⁻⁵ and ∼3 · 10⁻⁴ M_⊙. This is similar to the values reported by Ji et al. (2016a) to explain the r-process enhanced stars in Reticulum II (∼2.5 · 10⁻⁵−5 · 10⁻⁵ M_⊙). Therefore, it is likely that Fnx-mem0556, Fnx-mem0595, and Fnx-mem0546 also got enriched by a single r-process event. Moreover, the stars probably formed only a few megayears after the event. Otherwise, the r-process material would have been mixed into a larger amount of gas (van de Voort et al. 2020).

Our simple estimate has several input parameters (e.g., n₀, E₅₁), therefore, we show in Figure 9 an overview covering different possible values. In this figure, the absolute Eu abundances of Fnx-mem0556 are indicated by horizontal orange lines, the estimated dilution mass by vertical orange lines, and the colors correspond to the Eu mass ejected by the r-process event using Equation (5). The latter changes smoothly over several orders. However, the amount of Eu needed to explain the Eu-stars (M_Eu ∼ 10⁻⁵–10⁻⁴ M_⊙, Figure 9) agrees with having only one r-process event (see also Beniamini et al. 2016 for estimates of the ejected Eu mass per r-event in dSph galaxies).

Zoom In Zoom Out Reset image size

Which r-process event is responsible for the enhanced Eu in the three Fornax stars? The possible candidates are compact binary mergers (two neutron stars or a neutron star and a black hole) and MR-SNe including the early explosion phase and the late evolution as collapsar.

In Figure 7, we have compared the [Ba/Eu] of the Eu-stars to various theoretical predictions from compact binary mergers (dynamical and disk ejecta) and from MR-SNe using different nuclear physics input (Kodama & Takahashi 1975; Duflo & Zuker 1995; Panov et al. 2001; Möller et al. 2003; Kelic et al. 2008; Marketin et al. 2016). We test the dynamical ejecta of mergers of two neutron stars (NSNS (B), Bovard et al. 2017 and NSNS (R), Korobkin et al. 2012), and an NSBH (R) (Korobkin et al. 2012); ejecta from the accretion disk formed after merger surrounding the compact object (Disk 1, Disk 2, and Disk 3;, Wu et al. 2016); ejecta from magneto-rotational SNe MR-SN (W) (Winteler et al. 2012) and MR-SN (O) (Reichert et al. 2021). A similar set of models was used also by Côté et al. (2021) and Eichler & Arcones (2021) to compare to meteorite ratios of r-process isotopes. However, there are too large theoretical uncertainties in the nuclear physics input as well as the variation in the astrophysical conditions and this makes it impossible to conclude from the [Ba/Eu] ratio alone which scenario has contributed to the enhanced Eu.

The mass of Eu estimated above (M_Eu ∼ 10⁻⁵–3 · 10⁻⁴ M_⊙) could be also used as a constraint on the r-process site. We compare this amount of Eu to the mass that is ejected in NSMs and MR-SNe, as obtained in hydrodynamic simulations. However, these simulations are still uncertain due to various aspects (e.g., resolution, magnetic fields, neutrino matter interactions, high-density equation of state) and predictions about the amount of Eu ejected are only approximate.

In compact binary mergers, the masses of ejected Eu in the dynamic ejecta depends on the simulation and can differ by orders of magnitudes (see, e.g., Côté et al. 2018 for an overview of different ejecta masses of the dynamical ejecta). Just to name a few examples, the Eu mass based on the simulation of Bovard et al. (2017) is M_Eu ∼ 10⁻⁶–10⁻⁵ M_⊙. Yields of other simulations are slightly higher with M_Eu ∼ 10⁻⁵–8 · 10⁻⁵ M_⊙ (Korobkin et al. 2012) and similar M_Eu ∼ 3 · 10⁻⁵–5 · 10⁻⁵ M_⊙ (Goriely et al. 2011). In the case of an NSBH merger, the ejected Eu mass of the dynamic ejecta can reach M_Eu ∼ 3 · 10⁻⁴ M_⊙ (Korobkin et al. 2012). In addition, the disk ejecta will also contribute to Eu (M_Eu ∼ 3 · 10⁻⁶ − 10⁻⁵ M_⊙, Wu et al. 2016). However, it is still under discussion as to how much each ejected component contributes and how neutron-rich the components are (see, e.g., Shibata & Hotokezaka 2019, for a recent review).

In GW170817, Watson et al. (2019) directly observed Sr and predicted a mass M(Sr) ≳ 5 · 10⁻⁵ M_⊙ . When assuming that the ejecta is scaled like the solar r-process residual (Sneden et al. 2008) the ejected Eu mass is M_Eu ∼ 3 · 10⁻⁶ M_⊙. We note that the lighter heavy elements like strontium usually scatter with respect to the heavier elements. Because of this and other involved uncertainties, this value should be treated as a rough estimate only. Nevertheless, it agrees well with the masses of M_Eu ∼ 3 · 10⁻⁶–1.5 · 10⁻⁵ M_⊙ obtained in Côté et al. (2018) for GW170817.

The amount of Eu produced in MR-SNe based on recent simulations is ≲ 1 · 10⁻⁵ M_⊙ (Winteler et al. 2012), ≲ 1.5 · 10⁻⁵ M_⊙ (Nishimura et al. 2015), ≲ 8 · 10⁻⁶ M_⊙ (Nishimura et al. 2017), and ≲ 5 · 10⁻⁶ M_⊙ (Reichert et al. 2021). These values are on the lower end of our assumed uncertainties (Figure 9), but they could still be responsible for the enrichment. After some MR-SNe, a black hole forms and is surrounded by an accretion disk. The disk conditions may be similar to those found in accretion disks after NSMs and may favor an r-process. Siegel et al. (2019) have found that Eu may be produced in collapsars, however, more detailed investigations are necessary to understand whether the conditions are appropriate for an r-process, see Miller et al. (2020) and Just et al. (2021).

In Figure 9, we show the maximum Eu mass ejected by NSMs (black line, 8 · 10⁻⁵ M_⊙) and by MR-SNe (red line, 1.5 · 10⁻⁵ M_⊙). Moreover, we include the estimated Eu mass necessary to explain r-II stars in Reticulum II (Ji et al. 2016a). All of these estimates are close to the mass that is necessary to enhance the Eu-stars, as derived from the observed stellar abundances.

In addition to Eu, one can use α-elements, e.g., Mg, to check whether MR-SNe are the r-process site responsible for more element abundances in the Fornax Eu-stars. Since the amount of ejected α-elements depends on the progenitor mass (e.g., Kobayashi et al. 2006), and the progenitors of MR-SNe may differ from the average core-collapse supernovae, the amount of ejected α-elements may also vary between the two types of SNe. If an abnormal supernova produces r-process material and a huge amount of Mg, one would possibly see a signature of this in the Mg abundances in the Fornax Eu-stars. However, in Figure 2, the three stars present normal Mg abundances (see also Table 5). This may be an indication of an NSM producing the observed Eu or of MR-SNe ejecting normal amount of Mg as any other supernova in Fornax. Therefore, we cannot use Mg as an indicator for rare SNe and their r-process yields.

In summary, based on the abundance ratio of [Ba/Eu], on the estimated amount of Eu, and on the Mg abundance, we cannot determine the r-process site enriching the Eu-stars in Fnx. Based on Figure 9, NSMs seem a promising site to explain the r-material in the Eu-stars, however, their time delay coupled with recent star formation in Fnx may pose an issue that is easier to overcome if MR-SNe would be the r-site. On the other hand, recent studies also show that a substantial fraction of the NSM population may occur on short timescales (<100 Myr, see e.g., Beniamini & Piran 2019).

The strong r-process enhancement, despite the remaining unknown formation site, combined with the typical low α-content indicates, together with the Gaia DR2 proper motions (Gaia Collaboration 2018a, 2018b), that these rare Eu-stars are indeed Fornax members (see Section 2.2).

From a timescale perspective, NSMs remain viable sources and could explain the observed abundances. NSMs are expected to contribute with some delay and thus such an event could account for producing r-process material late or at high metallicities. However, this would imply that there is some star formation just after the neutron stars merge and eject r-process material. Similarly linked to the late star formation, massive stars could also have formed and exploded as MR-SNe producing the r-process observed in the Eu-stars. This would require more gas at a later stage, which may be feasible in Fornax. In any case, the three r-rich stars must have formed shortly after the r-process event, i.e., there must have been an active star-forming environment. According to Lemasle et al. (2014), Fnx-mem0556 has an age of 4.36 ± 0.86 Gyr and Fnx-mem0595 an age of 5.75 ± 1.78 Gyr. Fornax shows a complex star formation history with several outbreaks. A significant number of stars were formed at early times, i.e., more than 10 Gyr ago. Moreover, there was a very recent period of star formation ∼4 Gyr ago (e.g., Coleman & de Jong 2008; de Boer et al. 2012; Hendricks et al. 2014; Weisz et al. 2014; Rusakov et al. 2021). This late sudden burst of star formation agrees well with the age of the r-process enhanced stars. Although the origin of this burst is not known, there are speculations that Fornax has undergone some galaxy merger events in its history, indicated by over-dense features in the spatial distribution of stars in Fornax (e.g., Coleman et al. 2004).

Another explanation for such a peculiar enhancement of neutron-capture elements was investigated by Tsujimoto & Shigeyama (2002). They proposed that inhomogeneous mixing, caused by a high-velocity dispersion in dSph galaxies may be responsible for the existence of stars such as Cos 82, and possibly also Fnx-mem0546, Fnx-mem0556, and Fnx-mem0595 (see also Sadakane et al. 2004; Aoki et al. 2007 for a discussion).