ABSTRACT
A few rare halo giants in the range [Fe/H] ≃ −2.9 ± 0.3 exhibit r-process element abundances that vary as a group by factors up to [r/Fe] ∼80, relative to those of the iron peak and below. Yet, the astrophysical production site of these r-process elements remains unclear. We report initial results from four years of monitoring the radial velocities of 17 r-process-enhanced metal-poor giants to detect and characterize binaries in this sample. We find three (possibly four) spectroscopic binaries with orbital periods and eccentricities that are indistinguishable from those of Population I binaries with giant primaries, and which exhibit no signs that the secondary components have passed through the asymptotic giant branch stage of evolution or exploded as supernovae. The other 14 stars in our sample appear to be single—including the prototypical r-process-element-enhanced star CS 22892-052, which is also enhanced in carbon, but not in s-process elements. We conclude that the r-process (and potentially carbon) enhancement of these stars was not a local event due to mass transfer or winds from a binary companion, but was imprinted on the natal molecular clouds of these (single and binary) stars by an external source. These stars are thus spectacular chemical tracers of the inhomogeneous nature of the early Galactic halo system.
Export citation and abstract BibTeX RIS
1. INTRODUCTION
The chemical composition of very metal-poor (VMP; [Fe/H] <−2.0) and extremely metal-poor (EMP; [Fe/H] <−3.0) stars provides a fossil record of the star formation and nucleosynthesis history of the early Galaxy. Carbon (and often N and O) enhancement appears to be common for stars with [Fe/H] ≲ −2.5 (Beers & Christlieb 2005; Carollo et al. 2011), but the overall abundance ratios of elements up to the iron peak are well established by [Fe/H] ≃ −2.5, with very small scatter (Cayrel et al. 2004; Arnone et al. 2005). The full range of r-process elements was also in place at this stage in the chemical evolution of the Galaxy (see review by Sneden et al. 2008). Other neutron-capture processes, such as the s-process, began to contribute significant amounts of heavy elements from [Fe/H] ≲ −2.5 (e.g., Simmerer et al. 2004).5 Hence, the processes by which the chemical elements were produced and recycled into the Galactic halo system are, at least in a broad-brush sense, well understood.
However, in a small fraction of stars in the abundance range −3.2 < [Fe/H] <−2.6, the uniform abundance pattern of the r-process elements as a group is enhanced by factors up to ∼80 relative to that of the iron-peak and lighter elements. Spectroscopic analyses of such stars with 8 m class telescopes have provided precise and detailed abundances of many r-process elements as a key to understanding their origin (Sneden et al. 2008; Cowan et al. 2011). Yet, the likely production site(s) of the r-process elements, as well as the mechanisms by which their abundances relative to the "standard" halo composition could vary so strongly from star to star in the early Galaxy, remain essentially unknown.
Explanations of this diversity fall into two main classes: inhomogeneous enrichment and incomplete mixing of the interstellar medium (ISM) by the first generation(s) of stars, or later, local pollution of neutron-capture elements by a binary companion of the presently observed star (Qian & Wasserburg 2001). In the first case, the very uniform abundance pattern of the α- and iron-peak elements is difficult to reconcile with the predictions of the models of stochastic star formation and enrichment in the early Galaxy (Arnone et al. 2005). This conflict would be resolved in the second case, but unlike the s-process-element-enhanced Ba and CH giants, which are known to all be long-period binaries (McClure & Woodsworth 1990; Jorissen et al. 1998), this conjecture is so far without observational foundation.
Here, we report the first results from four years of precise radial velocity monitoring, performed in order to assess the binary frequency of a sample of 17 r-process-element-enhanced VMP and EMP giants. Our results provide strong new constraints on the nature of the r-process production site(s) and on the use of these stars as tracers of the star formation and/or merger history of the early Galaxy.
2. SAMPLE DEFINITION AND OBSERVATIONS
Our program stars were drawn from the HERES search for r-process-element-enhanced stars (Christlieb et al. 2004; Barklem et al. 2005), supplemented by earlier and later discoveries as summarized by Hayek et al. (2009). Only stars north of declination ∼ − 25° and brighter than V ∼ 16.0 are accessible for study with the Nordic Optical Telescope (NOT) on La Palma, resulting in a total sample of 17 stars. Eight of these are in the r-I class (+0.3 < [Eu/Fe] <+1.0), as defined by Beers & Christlieb (2005), and nine are in the r-II class ([Eu/Fe] >+1.0). Table 1 lists the program stars, in order of increasing [r/Fe] ratio, in order to highlight the continuum of r-process enhancements that exists; the division into r-I and r-II classes appears to be merely one of convenience.6 Stars with measured U abundances, and those found here to be spectroscopic binaries, are indicated in the table.
Table 1. Stars Monitored for Radial Velocity Variation
| Star | R.A. (J2000) | Decl. (J2000) | V | B−V | [Fe/H] | [r/Fe] | Nobs | Remarks |
|---|---|---|---|---|---|---|---|---|
| HE 0524-2055 | 05:27:04 | −20:52:42 | 14.01 | 0.87 | −2.58 | +0.49 | 8 | r-I |
| HE 0442-1234 | 04:44:52 | −12:28:46 | 12.91 | 1.07 | −2.41 | +0.52 | 23 | r-I, SB |
| HE 1430+0053 | 14:33:17 | +00:40:49 | 13.69 | 0.58 | −3.03 | +0.72 | 19 | r-I |
| CS 30315-029 | 23:34:27 | −26:42:19 | 13.66 | 0.91 | −3.33 | +0.72 | 9 | r-I |
| HD 20 | 00:05:15 | −27:16:18 | 9.07 | 0.54 | −1.58 | +0.80 | 9 | r-I |
| HD 221170 | 23:29:29 | +30:25:57 | 7.71 | 1.02 | −2.14 | +0.85 | 23 | r-I |
| HE 1044-2509 | 10:47:16 | −25:25:17 | 14.34 | 0.66 | −2.89 | +0.94 | 13 | r-I, SB |
| HE 2244-1503 | 22:47:26 | −14:47:30 | 15.30 | 0.60 | −2.88 | +0.95 | 12 | r-I |
| HE 2224+0143 | 22:27:23 | +01:58:33 | 13.68 | 0.71 | −2.58 | +1.05 | 18 | r-II |
| HE 1127-1143 | 11:29:51 | −12:00:13 | 15.88 | ... | −2.73 | +1.08 | 12 | r-II |
| HE 0432-0923 | 04:34:26 | −09:16:50 | 15.16 | 0.73 | −3.19 | +1.25 | 14 | r-II |
| HE 1219-0312 | 12:21:34 | −03:28:40 | 15.94 | 0.64 | −2.81 | +1.41 | 6 | r-II |
| CS 22892-052 | 22:17:01 | −16:39:26 | 13.21 | 0.80 | −2.95 | +1.54 | 17 | r-II |
| CS 29497-004 | 00:28:07 | −26:03:03 | 14.03 | 0.70 | −2.81 | +1.62 | 8 | r-II |
| CS 31082-001 | 01:29:31 | −16:00:48 | 11.66 | 0.76 | −2.78 | +1.66 | 17 | r-II, U |
| HE 1523-0901 | 15:26:01 | −09:11:38 | 11.10 | 1.10 | −2.95 | +1.80 | 17 | r-II, U, SB |
| HE 1105+0027 | 11:07:49 | +00:11:38 | 15.64 | 0.39 | −2.42 | +1.81 | 9 | r-II |
Note. U indicates that uranium has been detected; SB indicates a confirmed spectroscopic binary.
Download table as: ASCIITypeset image
As the chemical compositions of the targets are already known—some in great detail—our observations were designed to just yield precise radial velocities as efficiently as possible, over a time span of several years. To this end, we obtained high-resolution spectra (R ∼ 45,000) with a signal-to-noise ratio (S/N) ∼10, using the bench-mounted, fiber-fed échelle spectrograph FIES at the NOT (Djupvik & Andersen 2010), which is installed in a separate, temperature-controlled underground enclosure. The useful wavelength range covered by these spectra is 4000–7000 Å.
Our goal was to reach an accuracy of 100–200 m s−1 per observation, except perhaps for the faintest program stars. The radial velocity zero point was checked with standard stars on every night of observation, and found to be reproducible to better than 45 m s−1 per observation over a five-year period. Thus, the accuracy of the radial velocities of the program stars is not limited by the instrument.
Our initial assumption, by analogy with the Ba and CH giants (Jorissen et al. 1998), was that any spectroscopic binaries in the sample would likely have orbits of long period, low eccentricity, and small amplitude—i.e., small, slow velocity variations. Thus, our strategy was to observe these stars at roughly monthly intervals, weather permitting, and then adapt the frequency of the observations to follow any objects with radial velocity variations detected in the initial data. Observations began in 2007 April and were continued on 51 nights through 2011 September, for a total of ∼234 spectra, an average of 14 per star. Faint stars at far southern declinations require ideal conditions and were observed less frequently than average.
3. DATA REDUCTION AND ANALYSIS
The entire set of reductions of the raw spectra (bias subtraction, division by a flat-field exposure, cosmic-ray removal, two-dimensional order extraction, and wavelength calibration) was performed with a program developed and extensively tested on exoplanet hosts by Buchhave (2010). For the fainter stars, it was found preferable to divide the long exposures into three pieces and remove cosmic-ray hits by median filtering.
Radial velocities were then derived from the reduced spectra by a multi-order cross-correlation procedure. This operation is the most difficult step in the analysis because (1) these stars are EMP and chemically peculiar and (2) the individual spectra have low S/N ratios. Thus, selecting an optimum template spectrum for each star is no trivial task. Noting that the primary objective of the analysis is to measure small velocity variations rather than absolute values, we have experimented with three types of template spectra: (1) the highest S/N spectrum of each star, (2) the velocity-shifted and co-added mean spectrum of each star, and (3) a synthetic spectrum consisting of δ functions at the (solar) wavelengths of the strongest visible lines. The choice of template for each star was then guided by the consistency of the resulting velocities. Templates (2) and (3) were generally found to give the most consistent results, the latter also yielding velocities on a reliable absolute scale.
In summary, our final procedure yielded radial velocities with a standard deviation of ≲ 100 m s−1 for the brighter and more metal-rich stars, rising to ∼300–1000 m s−1 for the faintest stars with the weakest spectral features.
4. BINARY DETECTION AND ORBIT DETERMINATION
Fourteen of our stars exhibit no significant variation in radial velocity over the period covered by our observations, including any earlier velocities reported from HERES (Barklem et al. 2005). Linear and parabolic fits of the run of velocities versus time were made to check for any long-term trends, but they were generally negligible within the uncertainties; a few of the stars will be kept under continued surveillance.
The star HE 0442-1234 was shown by P. Bonifacio et al. (2010, private communication) to be a probable long-period spectroscopic binary. Our new data enabled us to complete the orbit of this star, as well as for the newly discovered binaries HE 1044-2509 and HE 1523-0901. The orbital elements for these stars are listed in Table 2, and the observed and computed velocity curves are shown in Figure 1.
Figure 1. Observed and computed spectroscopic orbits for HE 0442-1234 (top), HE 1044-2509 (middle), and HE 1523-0901 (bottom). Orbital elements are listed in Table 2. Blue dots: FIES velocities; red dots: earlier data.
Download figure:
Standard image High-resolution image
Table 2. Orbital Elements for the Detected Binary Stars
| Element | HE 0442-1234 | HE 1523-0901 | HE 1044-2509 |
|---|---|---|---|
| P (d) | 2513.38 ± 4.46 | 302.78 ± 0.78 | 36.57 ± 0.20 |
| K (km s−1) | 12.50 ± 0.20 | 0.399 ± 0.008 | 27.48 ± 0.22 |
| e | 0.760 ± 0.001 | 0.23 ± 0.129 | 0.000 ± 0.000 |
| γ (km s−1) | 236.16 ± 0.20 | 162.50 ± 0.10 | 359.28 ± 0.55 |
| f(m) (M☉) | 0.1396 ± 0.0012 | 1.59 × 10−6 ± 0.54 × 10−6 | 0.0823 ± 0.0027 |
| asin i (R☉) | 403.2 ± 1.2 | 2.21 ± 0.17 | 20.24 ± 0.24 |
| i (for M2 ∼ 0.6 M☉) | 88 | 1.5: | 65 |
| RRoche (R☉) | 65a | 48 | 13 |
| σ (1 obs., km s−1) | 0.28 | 0.11 | 0.98 |
Download table as: ASCIITypeset image
In general, the individual component masses cannot be derived directly, but assuming a standard value of 0.8 M☉ for the mass of a halo giant allows us to estimate a minimum mass for the unseen companion. For HE 0442-1234, this is 0.67 M☉ if the inclination i ∼ 90°, and it cannot be much larger for the secondary star to remain invisible in the spectrum. In the other two systems, a secondary star on the main sequence could be of similar or lower mass. Assuming M2 = 0.6 M☉ leads to the estimates of i given in Table 2; note that the orbit of HE 1523-0901 is seen nearly face-on. These, in turn, lead to estimates of the size (volume equivalent radius) of the Roche lobes of the unseen stars, which are not very sensitive to the adopted geometry.
5. DISCUSSION
Our results enable us to address several issues of importance for understanding the likely astrophysical site(s) of the r-process, and also shed some light on the early Galactic production of carbon, as described below.
5.1. Binary Frequency
With three binaries detected in a sample of 17 stars, the binary frequency of the r-process-enhanced stars is 18%. The star HE 2327-5642, an r-II star showing a possibly variable velocity (Mashonkina et al. 2010), is below the southern limit of our sample, but is another candidate. This is fully consistent with the ∼20% binary frequency determined for normal cluster giants by Mermilliod (1996). An additional one to two future discoveries in a larger sample might boost the frequency to perhaps as much as 25%, but there is clearly no evidence that all r-process-element-enhanced stars are binaries, as speculated by Qian & Wasserburg (2001) and others.
Within the limitations imposed by the small sample, there also seems to be no difference in the binary population of the r-I and r-II classes. Similarly, of the two stars with measured U abundances, CS 31082-001 exhibits the so-called actinide boost (an overabundance of Th and U relative to third-peak r-process elements such as Eu) and is a single star, while HE 1523-0901 is a binary and shows no actinide boost. Remarkably, the C-enhanced prototypical r-II star CS 22892-052 also seems not to be a binary, despite the earlier suggestion by Preston & Sneden (2001), indicating that its C content was not produced in a former asymptotic giant branch (AGB) companion. Thus, membership in a binary system appears to be decoupled from details associated with these particular abundance variations.
5.2. Orbital Properties
The periods and eccentricities of our three confirmed binaries are entirely consistent with those of the sample of chemically normal Population I giant binaries by Mermilliod (1996): the longest-period orbit is highly eccentric and the shortest-period orbit is well below the limit of ∼100 days for tidal circularization. Note that the old, metal-poor CH stars typically have circular orbits for periods up to ∼1000 days. Moreover, the secondary Roche lobes are too small to accommodate typical AGB stars of ∼200 R☉, if the secondary stars passed through this phase of their evolution. This is again consistent with the lack of s-process-element enhancements in these stars.
5.3. The Astrophysical Site of the r-process
Proving that the r-process-element production did not occur in binaries provides no direct evidence of the nuclear physics process at the production site above that afforded by the detailed abundance analyses of these stars. Models attempting to explain the observed r-process abundance patterns fall into two classes, core-collapse supernovae (SNe) and merging binary neutron stars (see, e.g., Argast et al. 2004; Goriely et al. 2011), also discussed comprehensively by Sneden et al. (2008), and in the context of CS 31082-001 by Barbuy et al. (2011).
The key feature to be highlighted here is that the newly formed r-process elements were added in variable, but internally consistent, proportions, in the otherwise constant chemical composition of the next generations of EMP and VMP stars. That the r-process-element abundances varied so strongly in the clouds from which these stars formed indicates that the r-process elements were not simply uniformly dispersed, together with all lighter elements, in the SN explosions that are a common feature of all astrophysical models for the r-process. Ejection in a jet or beam directly from the nascent neutron star(s) seems the most natural scenario for achieving this. The chemical composition of the progenitor, and the varying distance and direction of the jet from the next cloud, would then explain, in a natural way, the continuously varying proportions of r-process elements to the bulk composition of the following generation of EMP stars.
5.4. r-process-enhanced Stars as Chemical Tags of the Early Galactic ISM
In this scenario, most stars would receive a "standard" dose of r-process elements; stars with r-process-element abundances exceeding the average by a factor of ≳ 2–3 would be seen as r-process-rich, while those below the average (exemplified by HD 122653; Westin et al. 2000; Honda et al. 2006) would have been bypassed by the r-process ejection and appear as r-process-poor. This latter group may be as numerous as the former, but without spectacularly strong spectral lines to call attention to them. The recent discovery by Aoki et al. (2010) of a cool, EMP main-sequence dwarf with highly r-process-enhanced elemental abundance ratios, consistent with classification as an r-II star, would obviate any model to explain r-process enhancements as only due to some atmospheric anomaly confined to red giants.
The existence of r-process-element-enhanced (or depleted) stars in a narrow range of metallicity near [Fe/H] ∼−3 would imply that such anisotropic SN explosions only appeared at a certain "chemical time," and that the ISM was quickly fully mixed soon thereafter. The spectacular abundance anomalies of these stars can thus be used as extreme examples of "chemical tags" of the sites and times of their formation. The r-I/r-II classification is just a coarse tool to indicate the degree of enhancement. However, as r-I stars are also found at higher metallicity than the r-II stars, they may have formed from clouds that were further enriched by material of "normal" halo composition.
The different processes responsible for the light and heavy r-process elements (see, e.g., François et al. 2007; Montes et al. 2007), as well as the existence of stars with and without an actinide boost, remain to be explained by further modeling, but the binary properties of EMP stars apparently played no role in this context: binaries formed with similar properties as in chemically normal stars (e.g., Gonzalez Hernandez et al. 2008).
Finally, it is remarkable that the prototypical r-II star CS 22892-052 is single and significantly enhanced in carbon, which has been assumed to originate in long-period binary stars together with the s-process elements, which are not observed in CS 22892-052. This casts doubt on the accepted explanation for the synthesis of C in the early Galaxy as due primarily to pollution by former AGB binary companions, and suggests the synthesis of C, N, and O in earlier, rapidly rotating massive stars as one attractive alternative (see, e.g., Meynet et al. 2006, 2010).
6. CONCLUSIONS
Eighty percent of our program r-I and r-II stars exhibit no detectable radial velocity variations, while three stars are binaries with well-determined orbits (Table 2), typical of systems with giant primaries, but no AGB secondary stars. Thus, the binary population among these stars is normal, and binary stars play no special role in producing the r-process elements and injecting them into the early ISM. The case of CS 22892-052 suggests that this may be true for the early synthesis of carbon as well.
We conclude that whatever progenitors produced the r-process elements (and carbon) were extrinsic to the EMP and VMP stars we observe today. These elements were likely ejected in a collimated manner, and make these stars archetypical chemical indicators of their formation sites in the early Galaxy.
This paper is based on observations made with the Nordic Optical Telescope, operated on the island of La Palma jointly by Denmark, Finland, Iceland, Norway, and Sweden, in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias. We thank Drs. Piercarlo Bonifacio, Luca Sbordone, Lorenzo Monaco, and Jonay Gonzalez Hernandez for alerting us to the binary nature of HE 0442-1234, and for allowing us to include their velocities in our orbital solution; Dr. G. Torres for computing the orbit of HE 1523-091; and Dr. Paul Barklem for providing the observing dates of the HERES spectra. We also thank several NOT students for obtaining most of the NOT observations for us in service mode. J.A. and B.N. acknowledge support from the Danish Natural Science Research Council, and L.A.B. from the Carlsberg Foundations. T.C.B. acknowledges partial funding of this work from grants PHY 02-16783 and PHY 08-22648: Physics Frontier Center/Joint Institute for Nuclear Astrophysics (JINA), awarded by the U.S. National Science Foundation.
Footnotes
- 5
- 6
Note that the r-II and r-I stars do occupy different, but overlapping, regions of metallicity space. The r-II stars are found only at very low metallicity, while the r-I stars extend over the range −3.0 ⩽ [Fe/H] ⩽−0.5; see the bottom panel of Figure 3 in Aoki et al. (2010).