S⁵: The Destruction of a Bright Dwarf Galaxy as Revealed by the Chemistry of the Indus Stellar Stream

Figure 2 displays the derived neutron-capture element abundances for Indus_13 (blue dots) along with the scaled solar system r-process (black line) and s-process (gray line) abundance patterns (Sneden et al. 2008). An average of the abundances from La to Hf has been used for the scaling (Ba was excluded due to possible NLTE effects as described above). It is clear that, for the heavy elements (La−Hf), the abundance pattern of Indus_13 displays a good match to the scaled solar system r-process abundance pattern. Thus Indus_13 was enriched by an r-process nucleosynthesis event and, with an Eu abundance of [Eu/Fe] = +1.81, can be classified as an r-ii star according to the classification scheme of Holmbeck et al. (2020). The other six Indus stream stars exhibit milder enhancements in Eu (+0.40 < [Eu/Fe] < +0.67) with all having [Ba/Eu] < +0.0 (Ji et al. 2020; see also Figure 3) consistent with enrichment from an r-process event. They can thus be labeled r-i stars.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As can be seen from Figure 2, the Th abundance of Indus_13 falls +0.33 dex below what is predicted from the Sneden et al. (2008) solar system r-process abundance pattern. We find $\mathrm{log}\epsilon (\mathrm{Th}/\mathrm{Eu})=-0.66$, while the Sneden et al. (2008) pattern predicts $\mathrm{log}\epsilon (\mathrm{Th}/\mathrm{Eu})=-0.33$. This means that Indus_13 joins the sub-group of actinide-poor r-process-enhanced stars. This group also includes the two other r-ii stars in dwarf galaxies for which a Th abundance has been derived: COS 82 in UMi for which Aoki et al. (2007) found $\mathrm{log}\epsilon (\mathrm{Th}/\mathrm{Eu})\,=-0.59$ and DES J033523−54040 in Reticulum ii (Ret ii) for which Ji & Frebel (2018) found the extremely low ratio $\mathrm{log}\epsilon (\mathrm{Th}/\mathrm{Eu})=-0.84$. The other extreme is the so-called actinide-boost stars, where the actinides are overproduced relative to the rare-earth elements. Examples include CS 31082−001 with $\mathrm{log}\epsilon (\mathrm{Th}/\mathrm{Eu})=-0.22$ (Hill et al. 2002) and 2MASS J09544277 + 5246414 with $\mathrm{log}\epsilon (\mathrm{Th}/\mathrm{Eu})=-0.12$ (Holmbeck et al. 2018). This abundance signature has been known to exist for some time in MW halo stars, and has also recently been detected in stars associated with the Helmi debris stream (Gull et al. 2021), but actinide-boost stars have yet to be discovered in any present-day surviving dwarf galaxy.

The actinide elements, Th and U, are exclusively produced via the r-process, and modelers have struggled to explain the variations seen in these elements in r-process-enhanced metal-poor stars (Mashonkina et al. 2014). However, recent results from r-process nucleosynthesis modeling suggest that the abundance patterns seen in actinide-poor and actinide-boost stars can be produced by varying the electron fraction (Y_e ) distribution within the same r-process site (Holmbeck et al. 2019). The Holmbeck et al. analysis was performed for the r-process nucleosynthesis in neutron-star mergers, which is currently the only r-process element production site that has been directly observed (Drout et al. 2017). A similar analysis of actinide element production in other proposed r-process element production sites such as magneto-rationally driven supernovae and collapsars (Winteler et al. 2012; Siegel et al. 2019) may help to better identify the source of r-process elements in systems like Indus.

Another way to investigate the source of r-process elements in Indus is to calculate the fraction of lanthanide elements to lighter neutron-capture elements—the so-called lanthanide fraction (X_LA)—for Indus_13, which can be directly compared to the lanthanide fraction found for the 2017 neutron-star merger GW170817 (Ji et al. 2019). We find $\mathrm{log}{X}_{\mathrm{LA}}\,=-1.33\pm 0.2$ for Indus_13 (the error is dominated by systematics). This is somewhat higher than the $\mathrm{log}{X}_{\mathrm{LA}}\,=-2.2\pm 0.5$ found for GW170817 but similar to the $\mathrm{log}{X}_{\mathrm{LA}}=-1.44$ found for the highly r-process-enhanced stars in the MW halo. This discrepancy between the $\mathrm{log}{X}_{\mathrm{LA}}$ for GW170817 and the highly r-process-enhanced MW halo stars led Ji et al. (2019) to conclude that neutron-star mergers with more lanthanide-rich ejecta are needed for neutron-star mergers to remain viable as the dominant r-process element production site. It is likely that many of the stars included in the highly r-process-enhanced MW halo sample used for the calculation in Ji et al. (2019) are accreted dwarf galaxy stars similar to Indus_13. With the separation of the halo r-process-enhanced stars into dynamically tied groups, a comparison X_LA can be done on a system-by-system basis, allowing us to better characterize the range of lanthanide fractions captured in metal-poor stars.

A further interesting feature of Indus_13, in connection to its r-process enrichment, is its relatively high metallicity. The stars in the Indus stream span a range in metallicity from [Fe/H] = −2.44 to −1.69 (Ji et al. 2020), where Indus_13, with a metallicity of [Fe/H] = −1.92, joins the metal-rich end of this range. With its relatively high metallicity, Indus_13 adds to the sample of r-ii stars in the MW with [Fe/H] > −2 that has emerged in recent years (e.g., Hansen et al. 2018; Roederer et al. 2018a; Sakari et al. 2018). Early exploration and surveys for r-II stars in the MW halo found r-ii stars only in a fairly narrow metallicity range from [Fe/H] = −3.19 to −2.58, suggesting that these stars are preferentially found at the lowest metallicities. On the other hand, the first r-i stars were discovered over a wider metallicity range, extending to metallicities of [Fe/H] ∼ −1.5 (Barklem et al. 2005). However, the recent extensive search for r-process-enhanced stars initiated by the R-Process Alliance (RPA) (Hansen et al. 2018) has found a number of r-ii stars with [Fe/H] > −2. In fact, of the 74 r-II stars discovered in the four RPA data release papers, 21 have [Fe/H] > −2, corresponding to 28%. This percentage is very similar to that found for the 242 r-i stars discovered, where 70 have [Fe/H] > −2, corresponding to 29% (Hansen et al. 2018; Sakari et al. 2018; Ezzeddine et al. 2020; Holmbeck et al. 2020). These results strongly suggest that the metallicity distribution function is the same for r-i and and r-ii stars in the halo. Thus the high metallicity of Indus_13 is not unexpected or unusual since r-ii stars, though rare, are equally likely to be found at high metallicities as r-i stars.

Apart from the MW, r-process-enhanced stars have also been discovered in several classical dwarf galaxies (e.g., Cohen & Huang 2009, 2010) as well as in ultrafaint dwarf (UFD) galaxies. The discovery of r-process-enhanced stars in the UFD galaxies Ret II (Ji et al. 2016; Roederer et al. 2016), Tucana III (Hansen et al. 2017; Marshall et al. 2019), and Grus II (Hansen et al. 2020) supports the idea that r-process-enhanced stars in the MW halo likely found their way into the halo through the accretion and disruption of such smaller systems (Brauer et al. 2019). With the release of kinematic data from the Gaia satellite, this theory has been tested: first on the smaller sample in Roederer et al. (2018b), and most recently by Gudin et al. (2021) on the full literature sample that includes the large number of r-process-enhanced stars discovered as part of the RPA. Both of these find that a number of r-process stars cluster in groups sharing kinematic features, suggesting that the stars were likely stripped from the same parent object. Indus_13, as the first highly r-process-enhanced star to be discovered in a disrupted dwarf galaxy stream, along with the recent discovery of r-process-enhanced stars in the Helmi debris, Helmi tail, and ω Centauri progenitor streams (Gull et al. 2021), thus provides a vital observational link supporting the accretion origin for r-process-enhanced stars. In addition, as is pointed out by Gull et al. (2021), many r-process-enhanced stars in the MW halo likely come from relatively massive galaxies that have experienced multiple r-process events and not just from UFD galaxies. As a consequence, we should expect to find stronger kinematic clustering among these stars from larger galaxies, compared with the smaller number of stars accreted from small galaxies.

The Indus stream contains a second chemically distinct star, namely Indus_0, for which Ji et al. (2020) found [N/Fe] = +1.16, [Na/Fe] = +0.94, and [Al/Fe] = +0.32, clearly separating it from the other Indus stars for these elements, as can be seen in Figure 3. The abundances for other elements in this star, however, are very similar to those for the rest of Indus sample. Since both Na and Al are known to be affected by NLTE effects we have also plotted the NLTE-corrected abundances for Na and Al for the Indus stars in Figure 3 using corrections from Lind et al. (2011) and Nordlander & Lind (2017), respectively. These corrections mainly result in an overall shift of the Na and Al abundances of the Indus stars and do not change the status of Indus_0 as an outlier.

With high N, Na and Al abundances, Indus_0 is similar to the abundance pattern found in so-called "second-generation" stars in globular clusters. Such stars show distinct enhancements in N, Na and Al abundances coupled with depletions in C, O, and Mg relative to the abundances of "first-generation" or "primordial" stars in the cluster. While the enrichment process is not understood, it is clear that nuclear-burning cycles, like the CNO, Ne−Na, and Mg−Al cycles, are involved in generating the variations in the light element abundance pattern of the stars (e.g., Gratton et al. 2019). The presence of these light element abundance variations is a characteristic of globular clusters, and there is no evidence of such correlated light-element abundance variations occurring in dSph or UFD field stars (e.g., Shetrone et al. 2003; Norris et al. 2017; Salgado et al. 2019). It should be noted that, with a metallicity of [FeII/H] = −2.44, a globular cluster progenitor of Indus_0 would be among the most metal-poor globular clusters known (Beasley et al. 2019). However, Letarte et al. (2006) have shown from high-dispersion analyses of a small number of stars that similar abundance anomalies are found in the Fornax globular clusters that are only slightly more metal rich (see the discussion below).

The spread in metallicity and the velocity dispersion of the Indus stream firmly point to it being a disrupted dwarf galaxy (Li et al. 2019; Ji et al. 2020). However, given that Indus_0 is in the observed sample, and given the apparent lack of globular cluster second-generation abundance pattern stars in the field populations of other dSph and UFD systems, the inevitable conclusion from our data is that the Indus progenitor possessed a globular cluster in which Indus_0 formed.

Excluding the most massive MW satellites (Magellanic Clouds and Sagittarius) globular clusters have been detected in association with only two MW dwarf galaxies; Fornax and Eridanus 2 (Huang & Koposov 2021). Specifically, one faint (M_V ≈ −3.5) cluster has been detected in the ultrafaint system Eridanus 2 (Koposov et al. 2015; Crnojević et al. 2016), while Fornax has six globular clusters (Wang et al. 2019). The Fornax globular clusters have long been known to be old and metal-poor (e.g., Zinn & Persson 1981; Buonanno et al. 1998, 1999; Strader et al. 2003), and Larsen et al. (2012) found that the four most metal-poor clusters, which have −2.5 < [Fe/H] < −2, match the metal-poor tail of the Fornax field star abundance distribution. Further, Larsen et al. (2014) have estimated the nitrogen abundances of red giants in these four clusters using multi-band HST imaging. They find a range in N abundance of Δ[N/Fe] ∼ +2 dex, with roughly equal numbers of N-normal and N-enhanced stars in each cluster. Together with the small sample of Fornax globular cluster red giants analyzed at high dispersion discussed in Letarte et al. (2006), these results point to the existence of multiple stellar populations in the Fornax globular clusters, with properties very similar to those seen in MW globular clusters. The inference that can be drawn from the Indus stars is then a similar picture to Fornax, with Indus_0 originating in a metal-poor globular cluster in which multiple stellar populations existed, and which was associated with the Indus progenitor.

Aside from the ongoing accretion and disruption of the Sgr dwarf, which possesses at least six associated globular clusters (e.g., Da Costa & Amandroff 1995; Law & Majewski 2010; Massari et al. 2019; Vasiliev 2019), our abundance results for the Indus stream members provide the first evidence for the accretion of a dwarf galaxy that contained at least one globular cluster. Other kinematic studies have found evidence for potentially similar events. For example, Malhan et al. (2019) found a cocoon of stellar material associated with the globular cluster stream GD-1, which they interpret as the remnants of the dwarf galaxy with which GD-1 was accreted. Similar analysis of the Jhelum stream has revealed a two-dimensional structure for this stream with a thin dense component and a more diffuse component (Bonaca et al. 2019). These components may represent distinct but related progenitors (Bonaca et al. 2019). In the context of the Indus stream results, this scenario is intriguing as it has been suggested that the Indus and Jhelum streams are two wraps of the same stream (Shipp et al. 2018; Bonaca et al. 2019). None of the Jhelum stars analyzed, however, displays abundance patterns compatible with a globular cluster origin (Ji et al. 2020). We note for completeness that no current Galactic globular cluster has been found to have orbital characteristics similar to those of the Indus stream in the study of Riley & Strigari (2020).

Apart from clues from various nucleosynthesis channels, the chemical signature and specific abundance trends of a stream also provide some information on the nature of the stream progenitor. For example, streams originating solely from globular clusters can be separated from dwarf galaxy streams using properties such as the presence or absence of iron abundance spreads, stream width, and velocity dispersion (e.g., Figure 4 of Ji et al. 2020). For dwarf galaxy streams, the mean metallicity and the iron abundance of the "knee" in the ([α/Fe], [Fe/H]) relation can also yield information on the stellar mass or luminosity of the progenitor (Kirby et al. 2011). We now use our information and that in Ji et al. (2020) to place constraints on the Indus stream progenitor.

Of the present-day intact dwarf spheroidal galaxies, a mixture of r-i and r-ii stars has been discovered in three systems: Ursa Minor (UMi), Draco, and Fornax (Shetrone et al. 2001; Cohen & Huang 2009; Letarte et al. 2010; Lemasle et al. 2014). r-i stars are also known to be present in Carina (Shetrone et al. 2003; Venn et al. 2012) and Sculptor (Kirby & Cohen 2012). Figure 4 shows the [Eu/Fe] abundances as a function of metallicity for the r-i and r-ii stars in Draco, UMi, Fornax, and Indus. For the dwarf galaxies, the stellar sample is that compiled in Hansen et al. (2017), where r-process-enhanced stars in dwarf galaxies were selected to have [Eu/Fe] > +0.3 and [Ba/Eu] < −0.4. The latter criterion is to ensure the stars were predominantly enriched by an r-process event as opposed to the slow neutron-capture process (s-process), which also produces both Ba and Eu but in a different ratio (Sneden et al. 2008). In Figure 4 we also include three of the recently re-analyzed Fornax stars from Reichert et al. (2021) satisfying the above criteria. In metallicity, the Indus stream stars mostly overlap with the r-process-enhanced stars found in Draco and UMi, whereas the r-i and r-ii stars in the more massive Fornax galaxy generally have higher metallicities, and those in the small ultrafaint, Ret II galaxy are predominantly found at lower metallicities. This suggests that the Indus's progenitor was a relatively massive galaxy with efficient chemical enrichment, similar to UMi. It should be noted, however, that the Indus stars selected for high-resolution follow-up, and thus included in this study, were preferentially selected to be metal-poor. Ongoing analysis of medium-resolution spectra for a larger sample of Indus stream stars finds a mean metallicity for the stream of [Fe/H] = −1.97 (A. B. Pace et al. 2021, in preparation). Consequently, a similar analysis of a larger sample of Indus stars would likely increase the overlap with the Fornax sample.

Zoom In Zoom Out Reset image size

The chemical enrichment efficiency of a galaxy, and by proxy its mass, can also be estimated from the evolution of the α-element abundances with metallicity. The galaxy will initially be enriched by a population of massive stars. These have short lifetimes that end in Type II supernovae explosions resulting in an α-enhanced enrichment. After this initial enrichment, lower mass and more slowly evolving stars are formed, and eventually Type Ia supernovae will contribute to the chemical signature of the system, resulting in a decrease in the α to Fe ratio of the system, also known as the α-knee (Tinsley 1979). The metallicity at which this transition in [α/Fe] occurs reflects the chemical enrichment efficiency of the galaxy: in a low-mass galaxy, the α-knee will appear at lower metallicities than in a high-mass galaxy.

All the Indus stream stars for which detailed abundances have been derived are enhanced in the α-elements (〈[Mg/Fe]〉 = +0.40, 〈[Si/Fe]〉 = +0.49, 〈[Ca/Fe]〉 = +0.43), suggesting that the location of the α-knee of the Indus stream progenitor is at higher metallicity than the most metal-rich star included in Ji et al. (2020), which has [FeII/H] = −1.69. A recent abundance reanalysis of high-resolution data for stars in Fornax and UMi has found the α-knee in these two galaxies to be located at [Fe/H] = −1.6 and [Fe/H] = −2.2, respectively, and predicted the location in Draco to be [Fe/H] = −2.1 (Reichert et al. 2020). Comparing these numbers to the metallicity of the Indus stars supports the hypothesis from the Eu abundances that the Indus progenitor was likely a dwarf galaxy whose mass exceeded that of UMi but which may not have been as massive as Fornax.

We can also use the fact that the Indus progenitor likely possessed at least one globular cluster to estimate its luminosity by calculating the normalized globular cluster specific frequency S_N (see Harris & van den Bergh 1981 and Huang & Koposov 2021 for the definition of S_N ). For a system such as Fornax with M_V = −13.4 (McConnachie 2012) one globular cluster results in S_N = 4.5 while the same number for UMi with M_V = −8.8 (McConnachie 2012) is ≈300. An S_N of 300 is high compared to what is generally found for the Local Group (Mackey et al. 2019) and seems unlikely given the results of Huang & Koposov (2021) on the MW satellites, thus supporting our contention that the luminosity of the Indus progenitor likely exceeded that of systems such as UMi. This is also in agreement with the initial estimate from Shipp et al. (2018) of the Indus progenitor's dynamical mass when it was stripping stars to the stream of 6.5 × 10⁶ M_⊙ based on the width of the stream. In this context we also note that the postulated Indus globular cluster was unlikely to have been of low luminosity or mass. As shown by Lagioia et al. (2019) and Milone et al. (2020) for example, globular clusters that possess multiple populations tend to have relatively large initial masses: only a few clusters with initial masses above 10⁵ M_⊙ lack multiple populations, while those that do not show multiple populations are generally of low mass (Lagioia et al. 2019; Milone et al. 2020).