The R-process Alliance: First Release from the Southern Search for R-process-enhanced Stars in the Galactic Halo^*

From the survey carried out by Barklem et al. (2005), the expected frequencies of r-II and r-I stars in the halo system are ∼3% and ∼15%, respectively. Among the 1658 metal-poor stars in JINAbase, 42 stars have abundances that match our limited-r definition, resulting in a frequency of ∼3% for this subclass (Abohalima & Frebel 2017). This frequency of limited-r stars is a lower limit, as many candidates with high upper limits on their Eu abundance exist, and these could belong to the limited-r (or r-I) subclasses. However, it is clear that only a few stars belonging to this subclass have data available to facilitate abundance derivation for a large number of heavy neutron-capture elements. In order to significantly increase the number of known r-II and limited-r stars in the halo we need to obtain snapshot spectra for a large number of bright stars, which facilitate abundance determination for a large number of neutron-capture elements.

To achieve the first goal of the RPA, we plan to obtain snapshot spectra of ∼2500 stars with V ≲ 13.5 and [Fe/H] ≤ −2.0 over both the southern and northern hemispheres using a variety of moderate- to large-aperture telescopes. Here we present the first sample, taken with the Echelle spectrograph at the Las Campanas Observatory du Pont 2.5 m telescope. In a second paper, we report on a similar effort carried out with the Apache Point Observatory 3.5 m telescope (C. M. Sakari et al. 2018, in preparation), with additional papers to follow.

Although several tens of thousands of candidate stars with [Fe/H] < −2 are now known (primarily from SDSS/SEGUE; Abazajian et al. 2009; Yanny et al. 2009), the great majority are too faint to efficiently obtain high-resolution data for a large number of stars in order to enable the determination of accurate abundances. Fortunately, a number of recent large surveys have provided suitably bright stars for consideration. Unfortunately, not all of these stars were known a priori to be VMP or EMP stars, thus Phase I of the RPA search for r-process-enhanced stars was to obtain medium-resolution validation data for those targets. These include stars from the RAVE survey (DR4; Kordopatis et al. 2013, DR5; Kunder et al. 2017) and the Best & Brightest Survey (B&B; Schlaufman & Casey 2014). Note that, although the RAVE data releases provided estimates of metallicity for many bright stars, we have found that only about 60% of the stars from this survey with quoted [Fe/H] < −2 are in fact this metal-poor (Placco et al. 2018). We therefore decided to validate as many of these stars as possible before proceeding to the high-resolution observations. We have also included stars satisfying our selection criteria that already have had medium-resolution validation completed. These include stars from the HK Survey of Beers and collaborators (Beers et al. 1985, 1992; T. C. Beers et al. 2018, in preparation) or the Hamburg/ESO Survey of Christlieb and colleagues (Frebel et al. 2006; Christlieb et al. 2008; Beers et al. 2017; T. C. Beers et al. 2018, in preparation), from the SkyMapper Survey (Wolf et al. 2018), and (in particular in the northern hemisphere) G-K-giants selected from the LAMOST survey (T. C. Beers et al. 2018, in preparation). Stars from smaller lists of candidates from various programs carried out by members of our team were also included.

For Phase II, we selected targets with V < 13.5, [Fe/H] < −2, 4000 K < T_eff < 5500 K, and not strongly enhanced in carbon, as validated by either extant medium-resolution spectra in our possession or spectra that we gathered as part of Phase I of the search (see Placco et al. 2018 for a more detailed description of our target selection and determination of atmospheric parameters and [C/Fe] using the n-SSPP, described by Beers et al. 2014). We then obtained snapshot high-resolution spectra with an S/N of ∼30 at 4100 Å and a resolving power R ∼ 25,000. This can be achieved in a 1200 s exposure for a V = 11.5 star with a 2.5 m class telescope. Christlieb et al. (2004) and Barklem et al. (2005) established that such spectra are sufficient to detect (or derive meaningful upper limits on) an enhanced Eu line at 4129 Å and an Sr line at 4077 Å, for cool, VMP stars. We then derive C, Sr, Ba, and Eu abundances for these stars, which allow us to identify likely r-II, r-I, limited-r, and CEMP-r stars, following the selection criteria summarized in Table 1. When identifying r-II, r-I, and CEMP-r stars from spectra taken in Phase II, we also require that our selected stars satisfy [Ba/Eu] < 0 to ensure that the chemical composition of the stars is dominated by the r-process and not the slow neutron-capture process (s-process; Beers & Christlieb 2005).

Phase III of this effort, which is already underway, makes use of generally large-aperture telescopes in order to obtain much higher-quality "portrait" spectra (S/N ∼100 at 3850 Å; R ∼ 50,000 or higher) for all of the r-II and limited-r stars (and a subset of the r-I stars) identified during Phase II. Placco et al. (2017) reports on the first portrait spectrum taken during Phase III, for RAVE J203843.2–002333, a bright (V = 12.7), VMP ([Fe/H] = −2.91), r-II ([Eu/Fe] = +1.64, and [Ba/Eu] =−0.81) star originally identified during the RAVE survey; it is also only the fourth VMP/EMP star reported with a measured uranium abundance. In addition, RAVE J153830.9–180424, one of the most metal-rich r-II stars known in the Milky Way halo ([Fe/H] = −2.09), was recently identified in the northern hemisphere Phase-II search. The portrait spectrum of this star, and its elemental-abundance analysis, is presented in Sakari et al. (2018). The full set of portrait spectra will be analyzed and published in due course.

The first dataset on this program in the southern hemisphere was obtained during six nights in 2016 August, with the Echelle spectrograph on the du Pont 2.5 m telescope at the Las Campanas Observatory. During this run, we obtained spectra for 107 stars in the magnitude range of 10 < V < 13.5 and with metallicities of −3.13 < [Fe/H] < −0.79 (note that these metallicities reflect the high-resolution abundance results, not those from Phase I). The spectra were obtained using the 1'' × 4'' slit and 2 × 2 on-chip binning, yielding a resolving power of R ∼ 25,000, covering the wavelength range from 3860 to 9000 Å. The data were reduced using the Carnegie Python Distribution⁸ (Kelson 1998, 2003; Illingworth et al. 2000). Radial velocities were measured from order-by-order cross-correlation of the target spectra with a spectrum of HD 213575, a radial-velocity standard star obtained during the run, using the fxcor task in IRAF.⁹ On average 17 orders in each spectrum were used for the correlation. We also estimated the S/N of each spectrum in the region of the 4129 Å Eu line by taking the square root of the total counts in the continuum, converted to photons by multiplying with the CCD gain = 1.05 e⁻/DN. Table 2 lists the observed targets with R.A., decl., V magnitude, MJD, exposure time, S/N at 4129 Å, heliocentric radial velocity (RV_helio) and its associated error, the source catalog from which we originally selected each target, and additional common identifiers for the stars.

We derive 1D LTE stellar parameters for our program stars from equivalent-width measurements of a large number of Fe i and Fe ii lines, using the 2017 version of MOOG (Sneden 1973), including Rayleigh scattering treatment as described by Sobeck et al. (2011).¹⁰ The number of Fe lines used for the analysis ranges from 28 to 140 for Fe i and 6 to 27 for Fe ii, with a mean of 86 and 15 lines used per spectrum, respectively. Equivalent widths of Fe i and Fe ii lines measured for the individual stars, along with the wavelength (λ), excitation potential (χ), oscillator strength (log gf), and derived abundances (log ) for the lines, are listed in Table 3. Effective temperatures (T_eff) were derived by ensuring excitation equilibrium for the Fe i-line abundances. Spectroscopic derivation of stellar parameters using non-local thermodynamic equilibrium (NLTE) corrected Fe abundances has been found to agree better with photometric temperatures than what is found from LTE spectroscopic temperatures (Ezzeddine et al. 2017). We therefore place these initial temperatures on a photometric scale, correcting for the offset between spectroscopic and photometric temperature scales using the following relation from Frebel et al. (2013):

$$\begin{eqnarray*}&&{T}_{\mathrm{eff},\mathrm{corrected}}={T}_{\mathrm{eff},\mathrm{initial}}-0.1\times {T}_{\mathrm{eff},\mathrm{initial}}+670.\end{eqnarray*}$$

The shift in temperature varies from 25 K for the hottest stars to 265 K for the coolest stars, with a mean shift of 208 K, owing to the predominantly low-temperature range of the sample. Surface gravities (log g) were derived through ionization equilibration by ensuring agreement between abundances derived from Fe i and Fe ii lines. Microturbulent velocities (ξ) were determined by removing any trend inline abundances with reduced equivalent widths for both Fe i and Fe ii lines. Final parameters are listed in Table 4. We adopt an uncertainty of 150 K on T_eff, which reflects the uncertainty in the spectroscopically determined T_eff and a possible scale error from the application of the above relation. The standard deviation of the derived Fe i abundances varies from 0.09 dex to 0.28 dex, with a mean of 0.16 dex. Based on this value and the uncertainty in T_eff, we estimate uncertainties of 0.3 dex for log g and 0.3 km s⁻¹ for ξ. Figure 1 shows the derived gravities as a function of effective temperature for the sample stars. Over-plotted are four Y² isochrones with [α/Fe] =+0.4, age 12 Gyr, and [Fe/H] = −2.0, −2.5, −3.0, and −3.5 (Demarque et al. 2004). As the stars are selected to be bright and cool they mostly occupy the upper portion of the giant branch in this diagram. Four stars that are somewhat hotter than the rest of the sample were also observed, mainly because they were bright and/or due to the lack of other available targets in the appropriate R.A. range. Some of our stars also turned out to be more metal-rich than our initial selection criteria. This is mainly because not quite all of the targets initially chosen passed through the Phase I medium-resolution vetting process described in Section 2.1. A separate paper will address the differences in stellar parameters determined from the medium- and high-resolution spectra, respectively.

Zoom In Zoom Out Reset image size

Table 3.  Equivalent Widths of Fe i and Fe ii Lines in Our Sample

Stellar ID	Species	λ	χ	log gf	EW	log
		(Å)	(eV)		(mÅ)
J00002259−1302275	Fe i	4143.87	1.56	−0.51	133.30	5.04
J00002259−1302275	Fe i	4191.43	2.47	−0.67	94.10	5.28
J00002259−1302275	Fe i	4415.12	1.61	−0.62	117.10	4.78
J00002259−1302275	Fe i	4447.72	2.22	−1.34	94.40	5.59
J00002259−1302275	Fe i	4461.65	0.09	−3.19	122.20	5.73
⋮	⋮	⋮	⋮	⋮	⋮	⋮
J00002259−1302275	Fe i	6593.87	2.43	−2.42	30.50	5.44
J00002259−1302275	Fe i	6750.15	2.42	−2.62	43.30	5.87
J00002259−1302275	Fe ii	4522.63	2.84	−1.99	83.00	5.35
J00002259−1302275	Fe ii	4620.52	2.83	−3.19	74.20	6.33
J00002259−1302275	Fe ii	4923.93	2.89	−1.21	87.50	4.68
J00002259−1302275	Fe ii	5234.63	3.22	−2.18	37.00	4.97
J00002259−1302275	Fe ii	5316.62	3.15	−1.78	94.10	5.65
J00002259−1302275	Fe ii	6432.68	2.89	−3.71	34.70	6.04
J00021222−2241388	Fe i	4021.87	2.67	−0.73	84.24	5.43
J00021222−2241388	Fe i	4127.78	3.28	−1.61	18.42	5.55
J00021222−2241388	Fe i	4132.90	2.84	−0.92	58.74	5.22
J00021222−2241388	Fe i	4154.50	2.83	−0.69	76.67	5.41
⋮	⋮	⋮	⋮	⋮	⋮	⋮

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

As described in Section 2.1, we derive abundances for C, Sr, Ba, and Eu for the stars in our sample, which allows us to classify the stars as either r-I, r-II, limited-r, CEMP-s, CEMP-r, or non-r-process-enhanced. As the stars were preferentially selected not to exhibit carbon enhancement, we did not expect to find a large number of CEMP-s or CEMP-r stars, but nevertheless included carbon measurements in our analysis. A full abundance analysis of the stars will follow in future papers. All abundances are derived from spectral syntheses using MOOG2017 and α-enhanced ([α/Fe] = +0.4) 1D LTE ATLAS9 model atmospheres (Castelli & Kurucz 2003). We used solar photosphere abundances from Asplund et al. (2009) and line lists generated using the linemake package¹¹ (C. Sneden 2018, private communication), including molecular lines from CH, C₂, and CN (Brooke et al. 2013; Masseron et al. 2014; Ram et al. 2014; Sneden et al. 2014) and isotopic shift and hyperfine structure information for Ba and Eu (Lawler et al. 2001; Gallagher et al. 2010).

Carbon abundances are derived by fitting the CH G-band at 4313 Å and/or the C₂ Swan band at 5161 Å. Three Sr lines are available in our wavelength range at 4077, 4161, and 4215 Å. The 4077 Å line is very strong and is often saturated in our spectra; on the other hand, the 4161 Å is weak and severely blended. Hence, our Sr abundances are primarily derived from the line at 4215 Å. This line can be blended with CN features in cool metal-poor stars. However, as our stars are chosen not to be enhanced in C, this was only a complication for a few spectra. Barium abundances are derived from three of the five Ba lines in our wavelength range at 5853, 6141, and 6496 Å. There are also Ba lines at 4554 and 4934 Å, but they were often saturated in these cool stars. Furthermore, the 4934 Å line is also blended with an Fe i feature for which the log gf value is poorly constrained; these two lines were disregarded for the final Ba abundance. For Eu, eight absorption lines are present in our spectral range at 3819, 3907, 4129, 4205, 4435, 4522, 6437, and 6645 Å. The 3819 and 3907 Å lines are in a region of our spectra with poor S/N, and the 4205, 4435, and 4522 Å features are heavily blended. Hence, our Eu abundances are mainly derived from the 4129 Å feature, and those at 6437 and 6645 Å, when detectable. Examples of the synthesis of Sr, Ba, and Eu lines used for the classification of a limited-r star, an r-I star, and an r-II star are shown in Figure 2.

Zoom In Zoom Out Reset image size

For nine of the stars in our sample, we find velocities differing from those reported by the RAVE survey DR5 (Kunder et al. 2017); these are listed in Table 7. The typical uncertainty of the radial velocities listed by RAVE DR5 is < 2 km s⁻¹ (Kunder et al. 2017); the uncertainties of our radial velocities are typically < 1 km s⁻¹ (see Table 2). We have therefore chosen to highlight stars for which the two measurements differ by more than 5 km s⁻¹. One of these, J225319.9–224856, is a CEMP-s star, a class of stars enhanced in both C and s-process elements, and primarily found in binary systems (Hansen et al. 2016). Four of these stars are r-II stars, while two are r-I stars, resulting in six possible binaries among the 50 new r-I and r-II stars discovered in our sample. Additionally, one of the r-II star rediscoveries, HE 1523–0901, is also known to be in a binary system, while the other, CS 31082–001, shows no sign of radial-velocity variation (Hansen et al. 2015). No binary information is available for the r-I rediscoveries. In total, this results in a binary frequency of 11.1 ± 4.3% (assuming binomial errors), which agrees with the binary frequency of 18 ± 6% previously found for r-process-enhanced stars (Hansen et al. 2011, 2015). This is a lower limit, given that, e.g., HE 1523–0901 has radial-velocity variations well below 1 km s⁻¹, a level that is not accounted for here. However, it is striking that four of our ten newly detected r-II stars are apparent binaries, whereas Hansen et al. (2015) only found one binary system among the nine r-II stars included in their sample. We note that none of the possible binary systems identified in Table 7 are double-lined spectroscopic binaries.

Following the abundance-selection criteria listed in Table 1, we identify 12 r-II, 39 r-I, 20 limited-r, and 3 CEMP-r stars among the 107 stars observed. The three CEMP-r stars all have an Eu enhancement in the r-I star Eu range, and therefore we include them in this group in the following figures and discussion. Two of the identified r-II stars are rediscoveries: J15260106–0911388 = HE 1523–0901 (Frebel et al. 2007) and J01293113–1600454 = CS 31082–001 (Hill et al. 2002). Two of the r-I stars are rediscoveries: J21063474–4957500 =HD 200654 (Roederer et al. 2014a) and J23265258–0159248 (Thanathibodee 2016). Thus, we have identified 10 new r-II stars and 37 new r-I stars. The frequencies of new r-process-enhanced stars are somewhat higher than was found by Barklem et al. (2005), who identified 8 r-II and 35 r-I stars in their sample of 253 stars. However, the stars in the sample of Barklem et al. (2005) are generally warmer and fainter than our targets, often impeding the detection of Eu and other neutron-capture elements if the target S/N is not reached for the data. Hence, a number of their non-detections may not be real, but rather, a result of higher stellar temperature, low S/N, or a combination of both.

Specific searches for stars that exhibit a limited-r-process signature have not been carried out in the past, although multiple works have recognized the existence of these stars (e.g., Travaglio et al. 2004; Honda et al. 2006; Jacobson et al. 2015). We find 20 stars with abundances satisfying our limited-r criteria (see Table 1), which is about half of the stars with low Eu abundances in our sample.

We note that four CEMP-s stars were also found in our sample, again because not all targets for this run had been through our eventual Phase-I vetting process.

Figure 3 shows the number of identified r-I, r-II, and limited-r stars as a function of metallicity. We detect both r-I and r-II stars over a wide metallicity range (−3.0 ≲ [Fe/H] ≲ −1.5). The limited-r stars are found primarily at the low-metallicity end; from [Fe/H] ∼ −2 and down to [Fe/H] ∼ −3. The observed distribution for the limited-r stars indicates that this signature is more clear in metal-poor environments, where chemical evolution has not yet erased signatures of individual processes, as opposed to more chemically evolved systems, in which only processes leaving a strong chemical signature can be identified anymore, like the r-II stars. However, the currently available sample remains small. With future, larger samples of r-I, r-II, and limited-r stars collected by the RPA, we will soon be able to better determine the metallicity distributions of these stars.

Zoom In Zoom Out Reset image size

The metallicity distribution of the our sample also differs from that of Barklem et al. (2005). The Barklem et al. (2005) stars have a median [Fe/H] ∼ −2.8, after selecting candidates with [Fe/H] ≤ −2.5. In contrast, we chose [Fe/H] < −2.0 for our initial candidate selection, and the median of our sample is [Fe/H] ∼ −2.4. All of the r-II stars detected by Barklem et al. (2005) were found in a narrow metallicity range around [Fe/H] ∼ −3, as do many of the r-II stars discovered since. However, we detect r-II stars in a wider metallicity range, from [Fe/H] = −3 to [Fe/H] = −1.5, suggesting that the previous metallicity preference of these stars is likely an artifact or selection biases toward the overall lower metallicity in previous samples.

Interestingly, the metallicity distribution of the r-I and r-II stars found in the Milky Way satellites is more similar to what we find (Shetrone et al. 2001; Letarte et al. 2010). The one r-II star found in the bulge also has [Fe/H] ∼ −2 (Johnson et al. 2013). Furthermore, we recently detected an r-II star with [Fe/H] = −2 in our northern hemisphere sample (Sakari et al. 2018).

Our sample also includes the most metal-rich r-II star detected in the Milky Way halo to date, J18024226–4404426, with [Fe/H] = −1.55, [Eu/Fe] = +1.05, and [Ba/Eu] = −0.10. The relatively high [Ba/Eu] ratio found in J18024226–4404426 clearly shows that a substantial contribution from the s-process to the neutron-capture element abundances is present at this metallicity. Nevertheless, the neutron-capture-element abundance pattern of this star is dominated by the r-process, and, together with the other r-II stars discovered in this search, it widens the otherwise narrow metallicity distribution of the known r-II stars in the halo.

To demonstrate the difference in V magnitudes for our 10 newly discovered r-II stars compared to the literature sample r-II stars in the halo, we have plotted the number of r-II stars as a function of V magnitude in Figure 4 for our stars and the literature stars for which we could find V magnitudes. All of our new r-II stars have V < 13. Only four r-II stars this bright were known previously (J203843.2–002333, V = 12.7, Placco et al. 2017; CS 31082–001, V = 11.6, Hill et al. 2002; HE 1523–0901, V = 11.1, Frebel et al. 2006; and J153830.9–180424, V = 10.86, Sakari et al. 2018).

Zoom In Zoom Out Reset image size

This sample also includes the brightest r-II star detected in the Milky Way's halo to date, J21091825–1310062, with V =10.7, [Fe/H] = −2.40, [Eu/Fe] = +1.25, and [Ba/Eu] =−1.13. This star has one of the lowest [Ba/Eu] ratios found in our sample, indicating a pure r-process origin for the neutron-capture elements seen in this star. Note that the second brightest r-II star was discovered in the northern search for r-process-enhanced stars conducted by the RPA (Sakari et al. 2018). As the present manuscript was being prepared, yet another r-II star, even brighter than either of the above two stars (V = 10.1), was identified from northern hemisphere snapshot spectroscopy with the McDonald 2.7 m telescope. A portrait spectrum was also obtained, and is reported on by E. Holmbeck et al. (2018, in preparation). The brightness of the r-II stars in our sample means that we will be able to derive very complete neutron-capture-element abundance patterns for them, including thorium and possibly uranium, which is essential to distinguish between different r-process models. A number are sufficiently bright to be studied in the near-UV at high spectral resolution with the Hubble Space Telescope.

With the stellar-parameter range and S/N of this sample, we are able to detect Eu abundances in our stars down to [Eu/Fe] ∼−0.2. The distribution of Eu abundances for our sample stars as a function of metallicity is shown in Figure 5. For comparison, we have also included Eu measurements from the large sample of Roederer et al. (2014a). Similar to Roederer et al. (2014a), we find a large spread (> 2 dex) in the Eu abundances derived for our sample. This spread has been interpreted as a possible sign of multiple production sites for r-process elements (Sneden et al. 2000; Travaglio et al. 2004). In particular, we are able to detect Eu in stars over the full metallicity range of our sample, indicating the need for a source of r-process elements early in the universe.

Zoom In Zoom Out Reset image size

From Figure 5, it can also be seen that two of our identified limited-r stars have subsolar [Eu/Fe] abundances, a number of them cluster around around the solar value, and some are mildly enhanced in Eu. Our template star for the limited-r stars, HD 122563, has subsolar [Sr/Fe], [Ba/Fe], and [Eu/Fe], and is thus r-element poor. We have chosen to include more stars in this group, requiring that they show lower Eu abundances than those of the r-I (and r-II) stars and larger abundances of the light r-process elements, compared to the heavy r-process elements ([Sr/Ba] > +0.5), as we wish to investigate the production of the light r-process elements with this group of stars.

The [Ba/Eu] ratio is often used as a diagnostic for the level of r- versus s-process origin of the neutron-capture elements measured in stars. Figure 6 shows the [Ba/Eu] ratios for our sample stars, as a function of metallicity, along with the solar system r- and s-process ratios from Simmerer et al. (2004). The majority of our newly discovered r-II stars have [Ba/Eu] < −0.8, indicating a pure r-process origin for the neutron-capture-element content of these stars. The r-I and limited-r stars exhibit a larger variation in their [Ba/Eu] values, slightly increasing with metallicity.

Zoom In Zoom Out Reset image size

We also plot the absolute Eu abundances derived for our stars, as a function of metallicity, in Figure 7. A clear increase of Eu with Fe is seen for both the r-process-enhanced and non-r-process-enhanced stars; a similar trend is also seen for r-process-enhanced stars detected in dwarf galaxies (Hansen et al. 2017). The slopes for the r-I and r-II stars appear to differ in this plot, indicating different enrichment of these stars. It is likely that the sampling of r-I stars at the low-metallicity end is incomplete, due to the difficulty in detecting the lines for low Eu abundances in these stars. Also, the sampling of r-II stars at higher metallicity is sparse. Hence, the difference in slopes may just be a result of the metallicity sampling difference. This will be easier to determine with future larger samples of r-I and r-II stars gathered by the RPA.

Zoom In Zoom Out Reset image size

Finally, we plot the [Sr/Ba] abundances, as a function of Eu abundance, in Figure 8. The limited-r stars cluster in the upper-left corner of this plot, showing higher abundances of the light r-process element Sr than the heavy r-process elements Ba and Eu. A few of the r-II and r-I also exhibit [Sr/Ba] > +0.5, but generally these stars have lower [Sr/Ba] ratios than those of the limited-r stars.

Zoom In Zoom Out Reset image size