ar
X
iv
:1
60
5.
05
73
2v
1 
 [a
str
o-
ph
.G
A]
  1
8 M
ay
 20
16
Accepted for publication in ApJ
Preprint typeset using LATEX style emulateapj v. 11/26/04
THE CHEMICAL EVOLUTION OF THE BOO¨TESI ULTRA-FAINT DWARF GALAXY*
Anna Frebel1,2, John E. Norris3, Gerard Gilmore4, Rosemary F. G. Wyse5
Accepted for publication in ApJ
ABSTRACT
We present chemical abundance measurements of two metal-poor red giant stars in the ultra-faint
dwarf galaxy Boo¨tes I, based on Magellan/MIKE high-resolution spectra. For Boo I-980, with [Fe/H] =
−3.1, we present the first elemental abundance measurements while Boo I-127, with [Fe/H] = −2.0,
shows abundances in good agreement with previous measurements. Light and iron-peak element
abundance ratios in the two Boo¨tes I stars, as well as those of most other Boo¨tes I members, collected
from the literature, closely resemble those of regular metal-poor halo stars. Neutron-capture element
abundances Sr and Ba are systematically lower than the main halo trend, and also show a significant
abundance spread. Overall, this is similar to what has been found for other ultra-faint dwarf galaxies.
We apply corrections to the carbon abundances (commensurate with stellar evolutionary status) of
the entire sample and find 21% of stars to be carbon-enhanced metal-poor (CEMP) stars, compared
to 13% without using the carbon correction. We reassess the metallicity distribution functions (MDF)
for the CEMP stars and non-CEMP stars, and confirm earlier claims that CEMP stars might belong
to a different, earlier population. Applying a set of abundance criteria to test to what extent Boo¨tes I
could be a surviving first galaxy suggests that it is one of the earliest assembled systems that perhaps
received gas from accretion from other clouds in the system, or from swallowing a first galaxy or
building block type object. This resulted in the two stellar populations observable today.
Subject headings: early universe — galaxies: dwarf — Galaxy: halo — Local Group — stars: abun-
dances — stars: Population II
1. INTRODUCTION
Dwarf satellite galaxies are versatile probes of chemical
evolution, galactic halo assembly, and early galaxy for-
mation processes. Ultra-faint dwarf galaxies (L< 105 L⊙)
are particularly well-suited for testing these processes
since they appear to have had limited star formation and
chemical evolution, thus rendering them relatively simple
and unevolved systems. What has made ultra-faints par-
ticularly interesting is their overall low-metallicity cou-
pled with large metallicity spreads of several dex (e.g.,
(Kirby et al. 2008)). While stars with metallicities be-
low [Fe/H] = −3.0 have been found in nearly all of these
systems, the ultra-faint dwarfs completely lack higher-
metallicity stars [Fe/H] > −1.0 showing that chemical
enrichment did not proceed long enough to reach even
close to the solar level.
Based on the chemical abundances of the seven bright-
est stars, the faintest of the ultra-faint dwarfs, Segue 1,
was found to be a promising candidate for a surviv-
ing first galaxy (Frebel et al. 2014). Criteria for such
a survivor include large spreads in [Fe/H], halo-like
α-abundances at higher metallicity and low neutron-
capture element abundances, all of which are found
* This paper includes data gathered with the 6.5 meter Magellan
Telescopes located at Las Campanas Observatory, Chile.
1 Department of Physics and Kavli Institute for Astrophysics
and Space Research, Massachusetts Institute of Technology, Cam-
bridge, MA 02139, USA
2 Joint Institute for Nuclear Astrophysics - Center for Evolution
of the Elements, East Lansing, MI 48824
3 Research School of Astronomy and Astrophysics, Australian
National University, Canberra, ACT, 2611, Australia
4 Institute of Astronomy, University of Cambridge, Cambridge
CB3 0HA, UK
5 The Johns Hopkins University, Department of Physics & As-
tronomy, Baltimore, MD 21218, USA
among Segue 1 stars. Being the faintest of all ultra-faint
dwarf galaxies, Segue 1 may be regarded the best can-
didate for one of the most primitive dwarf galaxies still
observable (see also Ji et al. 2015 and Webster et al.
2016). But how common could surviving first, or very
early, galaxies be today? This can only be answered with
detailed inspections of the stellar chemical abundances in
additional dwarf systems. This is, however, challenging
given the faint nature of even the brightest available stars
near the tip of the red giant branch. High-resolution
spectra are thus only available of few stars per galaxy
(e.g., three each in Ursa Major II and Coma Berenices:
Frebel 2010; one star in Leo IV: Simon et al. 2010; twelve
stars in Hercules although most of them have a very lim-
ited red wavelength coverage: Koch et al. 2008; Ade´n
et al. 2011; seven stars in Segue 1: Norris et al. 2010a;
Frebel et al. 2014; four stars on Boo¨tes II Koch & Rich
2014; Ji et al. 2016b).
Fortunately, Boo¨tes I has been studied rather exten-
sively since its discovery in 2006 by Belokurov et al.
(2006). Beginning with medium-resolution spectroscopic
surveys (Mun˜oz et al. 2006; Norris et al. 2008, 2010b),
membership was established and then [Fe/H] distribu-
tions. Hughes et al. (2008, 2014) also find a large abun-
dance spread of at least from −3.7 < [Fe/H] < −1.9
based on photometric metallicities. [C/Fe] spreads and
α-abundance were also spectroscopically measured for
some subsamples (Lai et al. 2011). From these samples,
a total of 11 stars have been observed at high-resolution
to produce chemical abundance patterns of these stars
(Feltzing et al. 2009; Norris et al. 2010c; Gilmore et al.
2013; Ishigaki et al. 2014), of which most have been ob-
served by more than one group. This has led to well-
established abundances for these stars, which is of impor-
tance for the overall interpretation of the chemical inven-
2 Frebel et al.
tory of Boo¨tes I. Additionally, a kinematic study Koposov
et al. (2011) has been carried out, age estimates Boo¨tes I
were obtained (Okamoto et al. 2012; Hughes et al. 2014;
Brown et al. 2014) and its chemical evolution modeled
(Romano et al. 2015).
In this paper we report observations for one star in
Boo¨tes I with no previous high-resolution spectroscopy,
and a second that has been observed before. Our goal is
to investigate the chemical enrichment history of Boo¨tes I
to find out to what extent this galaxy shows signs of a
surviving first galaxy. To test such a hypothesis, we use
the observational criteria suggested by Frebel & Bromm
(2012) about the metallicity spread and different abun-
dance levels in a given system. As part of this we also
aim at quantifying whether there were multiple stellar
generations present and to what extent chemical evolu-
tion occurred in this system. This follows earlier hints
that some element contributions by supernovae type Ia
have occurred in the system (Gilmore et al. 2013). Since
the luminosity of Boo¨tes I is about 100 times higher than
that of Segue 1, and nearly 105 L⊙ (Mun˜oz et al. 2006),
we also aim at addressing how the overall extent of chem-
ical evolution relates to the luminosity of the system. Ad-
ditional considerations besides the metallicity-luminosity
relationship (e.g., Kirby et al. 2008) that appears to ex-
ist for all dwarf galaxies may shed light on the formation
and evolution of these objects.
In § 2 we describe our new observations and in § 3 our
analysis techniques and corresponding chemical abun-
dance results. We apply a set of chemical abundances
criteria to our and literature abundances for assessing to
what extent Boo¨tes I is a surviving first galaxy in § 4. We
discuss our findings in § 5.
2. TARGET SELECTION AND OBSERVATIONS
Targets were selected from Norris et al. (2008) who
had taken medium resolution R ∼ 5, 000 AAOmega spec-
tra of Boo¨tes I stars with the Anglo-Australian Telescope
(AAT). Their wide survey covered ∼ 6 half-light radii
from the center of the galaxy. Radial velocity-confirmed
members were further analyzed (Norris et al. 2010b).
Star Boo-1137, with [Fe/H] = −3.7, was observed with
high spectral resolution for a detailed abundance analy-
sis (Norris et al. 2010c), as were seven additional stars
(Gilmore et al. 2013). Six of those stars were also ob-
served by Ishigaki et al. (2014). The only brighter star
that had not been observed was Boo-980. Being located
at 3.9 half-light radii from the center it was only cov-
ered by the Norris et al. (2010b) study. On the con-
trary, another star, Boo-127, was observed by all these
studies, first by Feltzing et al. (2009) and then by all
others. Discordant abundance ratios were found which
made this star interesting for re-observation. We ob-
served these latter two Boo¨tes I stars with the MIKE
spectrograph (Bernstein et al. 2003) on the Magellan-
Clay telescope in March 2010 and March 2011. Details of
the MIKE observations and photometry taken from Nor-
ris et al. (2008) are given in Table 1. MIKE spectra have
nearly full optical wavelength coverage over the range of
∼ 3500-9000A˚. Observing conditions during these runs
were mostly clear, with an average seeing of 0.8′′ to 1.0′′.
The 1.0′′ ×5′′ slit yields a spectral resolution of ∼ 22, 000
in the red and ∼ 28, 000 in the blue wavelength regime.
We used 2 × 2 on-chip binning. Exposure times were
∼ 4 and 7 hr, and the observations were typically bro-
ken up into 45 to 55min exposures to avoid significant
degradation of the spectra by cosmic rays.
Reductions of the individual MIKE spectra were car-
ried out with the MIKE Carnegie Python pipeline ini-
tially described by Kelson (2003)7. The orders of the
combined spectrum were normalized and merged to pro-
duce final one-dimensional blue and red spectra for fur-
ther analysis. The spectra have modest S/N which
ranges from 25 to 30 at ∼ 5300 A˚ and 37 to 43 at
∼ 6000 A˚. Figure 1 shows a representative portions of the
spectra around the CH G-band at 4313 A˚ and the Mg b
lines at 5170 A˚. For comparison we also add CD −38 245
with [Fe/H] ∼ −4.2 as well as Arcturus ([Fe/H] = −0.5).
They bracket the metallicities of the Boo¨tes I stars.
3. CHEMICAL ABUNDANCE ANALYSIS RESULTS
We measured the equivalent widths of metal absorp-
tion lines throughout the spectra using software from
Casey (2014). Blending features were treated with spec-
trum synthesis to obtain line abundances. Table 2 lists
the lines used and their measured equivalent widths and
abundances for all elements, together with 3σ upper lim-
its for selected elements. Full details of our analysis pro-
cedure are given in Frebel et al. (2014) who analyzed the
Segue 1 stars in exactly the same manner.
Using 1D plane-parallel model atmospheres with α-
enhancement from Castelli & Kurucz (2004) and the lat-
est version of the MOOG analysis code (Sneden 1973;
Sobeck et al. 2011), we computed local thermodynamic
equilibrium (LTE) abundances for our Boo¨tes I stars. Fi-
nal abundance ratios [X/Fe] are obtained using the solar
abundances of Asplund et al. (2009) and listed in Table 3.
Abundance uncertainties refer to the standard deviation
of the line abundance measurements for each element. In
case of elements with few lines and the standard devia-
tion resulting in very small values, we adopt a nominal
minimum uncertainty of 0.1 dex.
3.1. Stellar parameters
We derive stellar parameters spectroscopically from
Fe I and Fe II lines by minimizing abundance trends and
matching the Fe I to Fe II abundance. We follow the pro-
cedure described in Frebel et al. (2013) which adjusts
the spectroscopic temperatures to better match those
derived from photometry. In this way, no unphysically
cool temperatures and low surface gravities are obtained.
We note that the present analysis procedures are iden-
tical to those employed in the analysis of the stars in
Segue 1 (Frebel et al. 2014). This allows us to make
a homogeneous comparison between the Boo¨tes I and
Segue 1 stars. The final spectroscopic stellar parame-
ters are Teff = 4720K, log g = 1.4, vmicr = 2.2 km s
−1
and [Fe/H] = −3.06 for Boo I-980 and Teff = 4765K,
log g = 1.35, vmicr = 2.3 km s
−1 and [Fe/H] = −1.99
for Boo I-127. We estimate our temperature uncertain-
ties to be ∼ 100 - 150K, and those in log g and vmicr
to be 0.3 dex and 0.3 km s−1, respectively. The spec-
troscopic temperatures agree well with temperatures de-
rived from the (g−r)o color transformed to B−V (Jordi
et al. 2006), and using color-temperature relations from
7 Available at http://obs.carnegiescience.edu/Code/python
The chemical evolution of Boo¨tes I 3
TABLE 1
Observing Details
Star α δ UT dates slit texp g0 (g − r)0 S/N S/N
(J2000) (J2000) hr mag mag 5300 A˚ 6000 A˚
SDSS JBoo980 13 59 12.68 +13 42 55.8 2010-03-07/08/18/24 1.0′′ 7.0 18.51 0.61 25 37
SDSS JBoo127 14 00 14.57 +14 35 52.7 2011-03-10/12 1.0′′ 3.7 18.15 0.76 30 43
Note. — The S/N is measured per ∼ 33mA˚ pixel.
Fig. 1.— Magellan/MIKE spectra of our Boo¨tes I stars, shown near the G-band head at 4314 A˚ (left) and the Mgb lines around 5180 A˚
(right). Some absorption lines are indicated. The stars are bracketed in terms of their metallicity by Arcturus (bottom) and CD −38 245
(top).
Alonso et al. (1999). They are Teff = 4810K for Boo I-
980 and Teff = 4625K for Boo I-127. Our final spectro-
scopic stellar parameters agree very well with values of
a 12Gyr isochrone with [α/Fe] = 0.4 and metallicities of
[Fe/H] = −3.0 and −2.0 (Kim et al. 2002).
3.2. Chemical abundances
We here briefly summarize key aspects of our chemical
abundance analysis. Again, we refer the reader to Frebel
et al. (2014) for further details since our analysis follows
the same procedures. We use the 4313 A˚ and 4323 A˚ CH
features to determine the C abundance. Both Boo-980
and Boo-127 have slightly subsolar [C/Fe] ratios, when
taken at face value. However, one has to take into ac-
count that carbon is converted to nitrogen due to the
CN cycle operating on the red giant branch. Many of the
Boo¨tes I stars are indeed located on the giant branch and
suffer from this depletion. Calculating individual correc-
tions based on stellar evolutionary modeling (see Placco
et al. 2014 for more details) for each star to counter this
effect leads to final carbon abundances of [C/Fe] = 0.13
and 0.37, for the two stars, respectively. We also apply
corrections to the adopted literature carbon abundances
which range from 0dex for the warmer giants to 0.75dex
for giants with Teff = 4500K (see also Section 3.3 for
further discussion). The corrected carbon abundances of
all stars in Boo¨tes I together with corrected values of the
comparison stars are shown in Figure 2.
As can be seen in the figure, there is a clear trend
of increasing carbon abundance with decreasing [Fe/H],
very similar to what is found among halo stars (e.g.,
Barklem et al. 2005; Placco et al. 2014). Overall, a
range in carbon abundances of 2.7 dex is present among
Boo¨tes I stars, from −0.34 < [C/Fe] < 2.31 (after ap-
plying the carbon correction). In fact, there are eight
carbon-enhanced metal-poor stars [C/Fe] > 0.7 (here-
after “CEMP stars”; CEMP subclasses are discussed
further below) after the application of the carbon cor-
rections of Placco et al. (2014). (There are five CEMP
stars before the correction.) This means that the over-
all fraction of CEMP stars eight out of 39, and thus
∼ 21%. For stars with [Fe/H] < −2.5, the fraction is
seven out of 19, and thus 37%. This increases to 66%
(four out of six) for [Fe/H] < −3.0, and 100% (one star)
for [Fe/H] < −3.5. These fractions of CEMP stars are
considerably higher than the results from halo stars for
the respective metallicities ranges, 24%, 43%, and 60%
(Placco et al. 2014). Note, though, that the Boo¨tes I
results are of course based on significantly smaller sam-
ples. Nevertheless, we broadly conclude that Boo¨tes I
does not show a carbon abundance distribution that is
different from that of the Galactic halo.
Although also plagued by low number statistics, the
same (qualitative) behavior is found for the combined
population of Segue 1 (excluding their metal-rich CH star
which is not shown in Figure 2), Ursa Major II, Coma
4 Frebel et al.
TABLE 2
Equivalent width measurements of the Boo¨tes I stars (Table will be
available electronically)
El. λ χ log gf EW [mA˚] lg ǫ [dex] EW [mA˚] lg ǫ [dex]
[A˚] [eV] [dex] Boo-980 Boo-127
CH 4313 · · · · · · syn 4.97 syn 6.24
CH 4323 · · · · · · syn 4.97 syn 6.34
Na I 5889.950 0.00 0.108 153.6 3.56 218.6 4.19
Na I 5895.924 0.00 −0.194 129.9 3.46 199.4 4.27
Mg I 3829.355 2.71 −0.208 163.7 4.96 · · · · · ·
MgI 3838.292 2.72 0.490 199.2 4.61 · · · · · ·
MgI 3986.753 4.35 −1.030 · · · · · · 88.7 6.16
Mg I 4057.505 4.35 −0.890 · · · · · · 96.3 6.15
Mg I 4167.271 4.35 −0.710 · · · · · · 97.7 5.99
Mg I 4571.096 0.00 −5.688 46.2 4.92 122.6 6.06
Mg I 4702.990 4.33 −0.380 44.3 4.73 117.9 5.87
Mg I 5172.684 2.71 −0.450 166.3 4.90 · · · · · ·
MgI 5183.604 2.72 −0.239 169.5 4.73 · · · · · ·
MgI 5528.405 4.34 −0.498 49.8 4.94 114.5 5.91
Al I 3944.006 0.00 −0.644 88.4 2.42 · · · · · ·
Al I 3961.520 0.01 −0.340 119.2 2.81 171.9 3.57
Si I 3905.523 1.91 −1.092 167.6 4.87 · · · · · ·
Si I 4102.936 1.91 −3.140 63.3 4.90 138.7 6.28
Ca I 4226.730 0.00 0.244 183.4 3.47 · · · · · ·
Ca I 4283.010 1.89 −0.224 · · · · · · 107.4 4.75
Ca I 4318.650 1.89 −0.210 · · · · · · 100.4 4.59
Ca I 4425.440 1.88 −0.358 38.6 3.61 98.0 4.65
Ca I 4434.960 1.89 −0.010 53.9 3.55 · · · · · ·
Ca I 4435.690 1.89 −0.519 · · · · · · 96.9 4.79
Ca I 4454.780 1.90 0.260 60.9 3.42 124.4 4.56
Ca I 4455.890 1.90 −0.530 31.6 3.67 · · · · · ·
Ca I 5581.971 2.52 −0.555 · · · · · · 38.0 4.50
Ca I 5588.760 2.52 0.210 · · · · · · 84.1 4.49
Ca I 5590.120 2.52 −0.571 · · · · · · 38.3 4.51
Ca I 5594.468 2.52 0.097 · · · · · · 86.8 4.64
Ca I 5598.487 2.52 −0.087 · · · · · · 73.6 4.61
Ca I 5601.285 2.53 −0.523 · · · · · · 34.6 4.41
Ca I 5857.450 2.93 0.230 · · · · · · 74.9 4.79
Ca I 6102.720 1.88 −0.790 · · · · · · 86.2 4.70
Ca I 6122.220 1.89 −0.315 36.9 3.47 107.5 4.59
Ca I 6162.170 1.90 −0.089 63.0 3.70 133.8 4.79
Ca I 6169.055 2.52 −0.797 · · · · · · 34.7 4.66
Ca I 6169.559 2.53 −0.478 · · · · · · 48.7 4.59
Ca I 6439.070 2.52 0.470 53.6 3.71 93.3 4.33
Ca I 6449.810 2.52 −0.502 · · · · · · 41.6 4.47
Ca I 6499.649 2.52 −0.818 · · · · · · 36.1 4.69
Ca I 6717.685 2.71 −0.524 · · · · · · 29.3 4.48
Sc II 4246.820 HFS · · · 116.7 0.19 syn 1.01
Sc II 4314.083 HFS · · · 74.6 −0.04 syn 1.14
Sc II 4324.998 HFS · · · 59.0 −0.03 syn 1.15
Sc II 4400.389 HFS · · · 59.8 0.11 syn 1.15
Sc II 4415.544 HFS · · · 56.7 0.16 syn 1.05
Sc II 5031.010 HFS · · · · · · · · · syn 1.00
Sc II 5239.811 HFS · · · · · · · · · syn 1.00
Sc II 5526.785 HFS · · · 18.8 0.09 syn 1.00
Sc II 5657.907 HFS · · · · · · · · · syn 1.13
Ti I 3989.760 0.02 −0.062 74.5 2.34 · · · · · ·
Ti I 3998.640 0.05 0.010 72.8 2.27 · · · · · ·
Ti I 4512.730 0.84 −0.424 · · · · · · 45.4 3.06
Ti I 4518.020 0.83 −0.269 26.7 2.55 49.6 2.97
Ti I 4533.249 0.85 0.532 40.5 2.05 94.6 2.95
Ti I 4534.780 0.84 0.336 39.3 2.20 88.8 3.02
Ti I 4535.567 0.83 0.120 · · · · · · 81.5 3.10
Ti I 4544.687 0.82 −0.520 · · · · · · 51.6 3.23
Ti I 4656.470 0.00 −1.289 · · · · · · 62.6 3.17
Ti I 4681.910 0.05 −1.015 18.3 2.13 85.8 3.32
Ti I 4840.870 0.90 −0.453 · · · · · · 36.4 2.98
Ti I 4991.070 0.84 0.436 · · · · · · 101.4 3.08
Ti I 4999.500 0.83 0.306 · · · · · · 97.4 3.12
Ti I 5007.206 0.82 0.168 · · · · · · 89.6 3.10
Ti I 5020.024 0.84 −0.358 · · · · · · 62.5 3.22
Ti I 5024.843 0.82 −0.546 · · · · · · 37.6 2.99
Ti I 5035.902 1.46 0.260 · · · · · · 60.5 3.32
Ti I 5036.463 1.44 0.186 · · · · · · 59.6 3.35
Ti I 5038.396 1.43 0.069 · · · · · · 32.9 3.02
Ti I 5039.960 0.02 −1.130 · · · · · · 69.9 3.10
Ti I 5064.650 0.05 −0.935 · · · · · · 72.2 2.98
Ti I 5173.740 0.00 −1.062 32.3 2.40 83.3 3.19
Ti I 5192.970 0.02 −0.950 · · · · · · 82.9 3.10
Ti I 5210.390 0.05 −0.828 36.6 2.32 91.0 3.15
Ti II 4012.396 0.57 −1.750 86.4 2.25 120.7 3.01
The chemical evolution of Boo¨tes I 5
TABLE 3
Magellan/MIKE Chemical Abundances of our
two Boo¨tes I Stars
Species N log ǫ(X) σ [X/H] [X/Fe]
Boo-980
CH 2 4.97 0.20 −3.46 −0.40
Na I 2 3.51 0.10 −2.73 0.33
Mg I 7 4.83 0.13 −2.77 0.29
Al I 2 2.62 0.20 −3.83 −0.77
Si I 2 4.88 0.10 −2.62 0.44
Ca I 8 3.57 0.11 −2.77 0.30
Sc II 6 0.08 0.10 −3.07 −0.01
Ti I 8 2.28 0.15 −2.67 0.40
Ti II 31 2.32 0.19 −2.63 0.43
Cr I 6 2.13 0.17 −3.51 −0.45
Mn I 3 1.93 0.10 −3.50 −0.44
Fe I 125 4.44 0.20 −3.06 0.00
Fe II 11 4.46 0.18 −3.04 0.02
Co I 4 2.45 0.10 −2.54 0.53
Ni I 5 3.23 0.21 −2.99 0.07
Zn I 1 1.90 0.15 −2.66 0.40
Sr II 2 −1.14 0.20 −4.01 −0.95
Ba II 2 −2.11 0.20 −4.29 −1.23
Eu II 1 < −2.44 · · · < −2.96 <0.10
Boo-127
CH 2 6.29 0.20 −2.14 −0.15
Na I 2 4.23 0.10 −2.01 −0.02
Mg I 6 6.02 0.11 −1.58 0.41
Al I 1 3.57 0.20 −2.88 −0.89
Si I 1 6.28 0.20 −1.23 0.76
Ca I 21 4.60 0.13 −1.74 0.24
Sc II 9 1.07 0.10 −2.08 −0.09
Ti I 22 3.11 0.12 −1.84 0.15
Ti II 46 3.30 0.22 −1.65 0.33
Cr I 20 3.57 0.18 −2.07 −0.09
Cr II 2 3.64 0.11 −2.00 −0.01
Mn I 7 2.94 0.14 −2.49 −0.50
Fe I 204 5.51 0.21 −1.99 0.00
Fe II 26 5.53 0.17 −1.97 0.01
Co I 4 2.88 0.16 −2.11 −0.12
Ni I 16 4.17 0.44 −2.05 −0.07
Zn I 1 <2.52 · · · < −2.04 < −0.05
Sr II 1 −0.32 0.99 −3.19 −1.20a
Ba II 3 −0.53 0.10 −2.71 −0.72
Eu II 1 < −1.37 · · · < −1.89 <0.10
a
Derived only from the λ4215 line. The λ4077 line gave a spuriously
low abundance.
Berenices, and Leo IV stars. The overall fraction is 5
out 13 (38%). No stars with [Fe/H] > −3.0 are CEMP
stars but the five stars with [Fe/H] < −3.0 are all carbon
enhanced (100%).
Interestingly, two of the eight carbon enhanced stars in
Boo¨tes I for which barium abundances are available be-
long to the class of CEMP-no stars, that is, CEMP stars
without supersolar enhancement in neutron-capture el-
ements, i.e. [Ba/Fe] < 0 (Beers & Christlieb 2005).
Additional data would be needed to determine whether
the remaining six objects are also CEMP-no stars. It
should be noted, though, that no star with a barium
measurement in Boo¨tes I has enhanced neutron-capture
element abundances compared to the solar level, mak-
ing it likely that these six objects will turn out to be
CEMP-no stars also. This speculation is furthermore
supported by the fact that all CEMP stars in the group
of ultra-faint dwarfs of Segue 1 (SDSS J100714+160154,
SDSS J100652+160235, SDSS J100639+160008, Segue 1-
7), Ursa Major II (UMa II-S1), and Leo IV (Leo IV-
S1) stars are CEMP-no stars. For completeness,
we also note that one CEMP-s star was found in
Fig. 2.— [C/Fe] abundance ratios (corrected for the evolution-
ary status of the stars) as a function of [Fe/H] (top panel) and sur-
face gravity (bottom panel) of our Boo¨tes I stars (filled red circles)
and literature Boo¨tes I stars (open red circles; Ishigaki et al. 2014
values were used for the best possible internal consistency), in com-
parison with metal-poor halo stars (black circles; Barklem et al. 2005)
and other dwarf galaxy stars (Segue 1: blue circles Frebel et al. 2014;
UMa II, ComBer and Leo IV: green circles Frebel et al. 2010; Simon
et al. 2010). Representative error bars are also shown. The definition
of C-enhancement from Aoki et al. (2007) is shown with a dashed line.
Segue 1. Interestingly, several CEMP-r stars were re-
cently found in the ultra-faint dwarf galaxy Reticu-
lum II (DES J033523540407, DES J033607540235, DES
J033454540558) whose r-process elements likely stem
from a independent source than all other elements,
namely a neutron star merger or magnetar (Ji et al.
2016a).
Again, this behavior is very similar to that of halo stars
which adds evidence that the most metal-poor halo stars
could have originated in small dwarf galaxies such as the
systems discussed here (see e.g., discussion in Frebel &
Norris 2015). A further extensive discussion on the car-
bon abundances in Boo¨tes I in the context of chemical
evolution can be found in Gilmore et al. (2013) and will
not be repeated here.
We used the Mg I, Si I, Ca I and Ti II lines to determine
6 Frebel et al.
Fig. 3.— Abundance ratios ([X/Fe]) as a function of metallicity ([Fe/H]) for various elements detected in our Boo¨tes I stars
(filled red circles) in comparison with those of stars in Segue 1 (Frebel et al. 2014; Norris et al. 2010a) (blue open circles) and
of halo stars (black and small cyan circles) of Cayrel et al. (2004) and Yong et al. (2013), respectively.
α-element abundances from equivalent width measure-
ments and spectrum synthesis (in the case of Si). Our
two stars have enhanced, halo-typical α-enhancement.
Figure 3 shows the comparison of most of our elemental
abundances for the two stars, in comparison with those of
the extremely metal-poor star samples from Cayrel et al.
(2004), Yong et al. (2013), as well as the Segue 1 ultra-
faint dwarf galaxy stars (Frebel et al. 2014). Overall, the
agreement of the different groups of stars is remarkably
good.
Considering the full Boo¨tes I sample, a similar behav-
ior is found. About two-thirds of the Boo¨tes I stars
with high-resolution abundance measurements have halo-
typical α-element abundances, as can be seen in Figure 4
in the top three panels. A combined [α/Fe] vs [Fe/H]
(bottom two panels) then delivers an overall reasonable
agreement with halo star abundances. No obvious down-
turn to solar-level ratios is found for Mg, Ca, Ti and the
combined [α/Fe], despite some individual stars having
low abundances. Overall, though, the slope of [α/Fe]
does somewhat decrease with increasing [Fe/H].
Excluding Boo-119, a CEMP-no star with [Fe/H] =
−3.3 and [α/Fe] = 0.77, the [α/Fe] slope is −0.10 (rms
scatter of 0.12dex), based on the high-resolution abun-
dances (second lowest panel in the Figure). Also ex-
cluding Boo-41 (with [Fe/H] ∼ −1.9), as Gilmore et al.
(2013) did, increases the slope to −0.18. Gilmore et al.
find −0.19, for comparison. When adding the combined
[α/Fe] measurements of Lai et al. (2011), the slope de-
creases to −0.03 (bottom panel of Figure 4; excluding
The chemical evolution of Boo¨tes I 7
Fig. 4.— Abundance ratios ([X/Fe]) of α-elements of our two
Boo¨tes I stars (full red circles) together with 11 Boo¨tes I stars (filled
red circles) as analyzed by Gilmore et al. (2013), Ishigaki et al. (2014),
and Feltzing et al. (2009). Segue1 stellar abundances (Frebel et al.
2014) are shown with blue circles, and halo stars in black circles (Cayrel
et al. 2004) and cyan dots (Yong et al. 2013). The bottom two panels
show combined [α/Fe] ratios. The bottom panel additionally includes
the [α/Fe] of Lai et al. (2011). Fits to the Boo¨tes data points (with
Boo-119 excluded; [α/Fe] = 0.77) are also shown.
only Boo-119), with an rms scatter of 0.13 dex. Impli-
cations of this behavior will be further discussed in Sec-
tion 4.
Iron-peak element abundances were obtained from
lines as listed in Table 2. The abundances found for our
two Boo¨tes I stars are also in excellent agreement with
those of comparable halo stars (see Figure 3). This indi-
cates the robust production of these elements in Boo¨tes I,
in the same way as in other dwarf galaxies and in the
halo. This behavior thus appears to be independent of
environment.
Finally, Sr and Ba are the only neutron-capture el-
ements detectable in the two stars. Boo-980 has Sr
and Ba abundances similar to those of halo stars with
[Fe/H] ∼ −3, although they are on the lower end of that
range. They are actually rather similar to the abun-
dances of Leo IV-SI in Leo IV studied by Simon et al.
(2010) and two of the Ursa Major II stars (UMa II-S1,
UMa II-S2) of Frebel (2010). This can be seen in Fig-
ure 5, where we show both [Sr,Ba/Fe] and [Sr,Ba/H] to
best illustrate the behavior of neutron-capture elements.
Boo-127 shows a somewhat different behavior. Its Sr
abundance is about 1 dex below the general Sr trend set
by halo stars. However, based on about half a dozen
data points, Boo-127 seems to extend an emerging sec-
ond branch (marked in the figure) of Sr abundance about
one dex below the main halo trend. It currently consists
mainly of four Ursa Major II and Coma Berenices stars,
two more Boo¨tes I stars, together with a few halo stars.
Additional measurements of Sr in stars with [Fe/H] > −2
in ultra-faint dwarf galaxies would help to further inves-
tigate the reality of this putative branch. Especially for
Boo¨tes I, currently only five stars have Sr measurements.
Boo-1137 also has a Sr abundance indistinguishable
from halo stars. However, its [Sr/H] level is extremely
low at [Sr/H] ∼ −5. Boo-94 has an even lower value of
[Sr/H] ∼ −5.4. Compared to Boo-127, this indicates a
[Sr/H] spread of at least 2 dex. The low level of Sr abun-
dances in these two stars is comparable with that found
in Segue 1. Generally, the Ba abundances of all Boo¨tes I
stars are somewhat uniformly offset from the main halo
star trend. The Ba abundance spread is 1.5 dex, and as
in the case of Sr, the lowest Ba abundances are similar
to those found in Segue 1. This behavior will be dis-
cussed further in Section 4. Interestingly, six Boo¨tes I
stars including Boo-127 cluster around [Ba/H] ∼ −2.7
and [Fe/H] ∼ −2.2. These stars also have other similar
abundances and could perhaps represent a dissolved star
cluster candidate (Bland-Hawthorn et al. 2010; Karls-
son et al. 2012) although more accurate abundances are
needed to draw firm conclusions.
3.3. Comparison of Boo-127 abundances
Boo-127 was first observed by Feltzing et al. (2009)
who found the star to have unusually high Mg
([Mg/Fe] = 0.76) and low Ca ([Ca/Fe] = 0.02) abun-
dances. However, these results were not confirmed by
Gilmore et al. (2013) or Ishigaki et al. (2014) who also
included this star in their samples. Our independent
observation of Boo-127 also yields much more moderate
values of [Mg/Fe] = 0.41 and [Ca/Fe] = 0.24, in line with
the Gilmore et al. and Ishigaki et al. studies. We thus
conclude that Boo-127 has no unusual α-abundances, but
rather a halo-like signature in these elements.
Figure 6 shows a detailed comparison of various ele-
ments observed in Boo-127 that are common to these four
studies. The [C/Fe] abundance of Norris et al. (2010b)
8 Frebel et al.
Fig. 5.— Abundance ratios of neutron-capture-elements [Sr/Fe] and
[Ba/Fe] (top panels) and [Sr/H] and [Ba/H] (bottom panels) as a func-
tion of metallicity [Fe/H] of our Boo¨tes I stars (filled red circles) and
literature Boo¨tes I stars (open red circles; here Ishigaki et al. 2014 was
used for Sr and Gilmore et al. 2013 for Ba, for internal consistency) in
comparison with those of other ultra faint dwarf galaxy stars in Segue 1
(blue circles; Frebel et al. 2014; Norris et al. 2010a), UMa II, ComBer,
Leo IV (green circles; Frebel et al. 2010; Simon et al. 2010), Draco
and Hercules (cyan squares; Fulbright et al. 2004; Koch et al. 2008,
2013), stars in the classical dwarf spheroidal galaxies (pink triangles;
Venn et al. 2004), and the galactic halo (black open circles; Aoki et al.
2005, Franc¸ois et al. 2007, Barklem et al. 2005 and Yong et al. 2013).
The dashed line is meant to guide the eye regarding a putative second
dwarf galaxy star sequence. Note that both axes have the same scale,
showing the huge range of neutron-capture abundances in metal-poor
stars. Representative error bars are shown in the [Sr/Fe] and [Ba/Fe]
panels.
is also included for comparison. Except for the high Mg
abundance of Feltzing et al. (2009) and the low carbon
abundance of Ishigaki et al. (2014), the abundances agree
reasonably well and are within ∼ 0.3 dex of each other.
We note here that Ishigaki et al. (2014) have systemat-
ically lower carbon abundances compared to our study,
Norris et al. (2010b) and Lai et al. (2011) by 0.43dex
(based on five stars). We note here that the stellar pa-
rameters of all four studies are consistent, though. The
effective temperatures agree within 165K and the surface
gravities within 0.6 dex. The [Fe/H] abundances are re-
Fig. 6.— Abundance ratios ([X/Fe]) of the four different
studies that analyzed Boo-127 as a function of atomic num-
ber. In the case of iron, [Fe/H] − [Fe/H]Boo−127, this study is
shown on the y-axis. Symbols are as follows: red circles: this
work; green triangles: Gilmore et al. (2013); black crosses:
Ishigaki et al. (2014); blue squares: Feltzing et al. (2009);
and cyan asterisk: Norris et al. (2010b).
markably close and within 0.11dex. Excluding the Feltz-
ing et al. (2009) results decreases these ranges to 65K,
0.25dex and 0.07dex. Since the Mg b lines are gravity
sensitive (e.g., Hollek et al. 2011), the lower Feltzing et
al. gravity choice of log g = 1.0 might, in retrospect,
explain their high Mg abundance.
3.4. Literature data on Boo¨tes I
In Table 4 we list abundances of key elements common
to all studies of Boo¨tes I stars. We list each star’s name
as provided by the original authors. We note that this
results in an apparently inconsistent system where, e.g.,
two distinct stars having the names Boo-7 and Boo07.
In reverse, where applicable, two distinct names for the
same star are also indicated. Coordinates for these stars
can be found in the references given in the last column.
For many stars, duplicate studies exist. Given authors
who use different data and analysis methods, to obtain
a set of abundances as internally consistent as possible
between studies, and following the abundance compari-
son in Figure 6, we adopt a combination of (adjusted)
abundances for our analysis and in the figures shown.
Adjustments were made to reflect different iron abun-
dances, e.g., [C/Fe]adopted = [C/Fe]orig +∆Fe/H].
Generally, we adopted our own and the Gilmore et al.
(2013) abundances (we adopted an average of the “NY”
and “GM” abundances) together with the carbon values
from Norris et al. (2010b). We adjusted the [Sr/Fe] val-
ues from Ishigaki et al. (2014) to the Gilmore et al. scale.
We replaced the carbon abundance upper limits of Ishi-
gaki et al. (2014) with detections of Norris et al. (2010b).
We used the adjusted carbon abundance of Norris et al.
(2010b) together with the adopted abundances of Feltz-
ing et al. (2009) in the case of Boo-7. Explanations of
which elemental abundances have been used are given in
the table notes but for clarity we list our adopted abun-
dances for each star at the bottom of the table.
Regarding abundance uncertainties in [X/Fe], system-
atic uncertainties taking into account uncertainties in the
stellar parameters as well as random uncertainties are
around 0.2 to 0.3 dex for this kind of data quality (e.g.,
The chemical evolution of Boo¨tes I 9
Table 4 in Frebel et al. 2014). Keeping this in mind, the
various abundance averages are not significantly affected
by our attempt to construct a homogeneous abundance
set and vary by on order 0.1 dex. Nevertheless, even such
subtle systematic changes could affect the interpretation,
and more generally, are important for comparisons with
chemical evolution models.
4. ON THE ORIGIN AND EVOLUTION OF BOO¨TES I
Using all available abundances of stars in Boo¨tes I, as
described in the previous section, we now investigate the
global chemical abundance signatures to characterize the
origin and history of this ultra-faint dwarf galaxy. In
particular, we use four abundance criteria developed by
Frebel & Bromm (2012) to assess to what extent chemical
evolution has taken place in this system and how far
removed Boo¨tes I might be from being a surviving first
galaxy, like e.g., Segue 1. The abundance criteria concern
the a) metallicity distribution function (MDF), b) light
and iron-peak element abundances, c) α-elements, and
d) neutron-capture elements. Under e) we also add a
discussion on the role of carbon in these early systems.
All of these are discussed in detail below, and applied
to the body of abundance data presently available for
Boo¨tes I .
a) Iron abundance spread and metallicity distri-
bution function. This criterion stipulates that a large
spread of 2-3 dex in [Fe/H] as well as a not too steep
MDF shape (especially on the low-metallicity side) would
be signs of early inhomogeneous mixing. The upper
panel of Figure 7 shows the MDF of Boo¨tes I as obtained
from the sample of 39 stars that includes both medium-
resolution and high-resolution [Fe/H] abundances (Norris
et al. 2010b; Lai et al. 2011; Feltzing et al. 2009; Norris
et al. 2010c; Gilmore et al. 2013; Ishigaki et al. 2014, this
work). When available, a high-resolution value was used
for a given star.
A spread in [Fe/H] of 2 dex is apparent, ranging from
[Fe/H] ∼ −3.8 to [Fe/H] ∼ −1.8. The lack of stars
[Fe/H] > −1.8 (when considering high-resolution abun-
dances where available) is typical for ultra-faint dwarf
galaxies (e.g., Kirby et al. 2008; Frebel et al. 2010, 2014).
The mean value for Boo¨tes I of [Fe/H] = −2.6 based
on this combined sample is in line with previous results
([Fe/H] = −2.3, Belokurov et al. 2006; [Fe/H] = −2.5,
Mun˜oz et al. 2006, Siegel 2006, Martin et al. 2007 and
Norris et al. 2008; [Fe/H] = −2.6, Lai et al. 2011).
From a theoretical viewpoint, chemical inhomogeneity
likely resulted in a large spread in [Fe/H] ranging from
[Fe/H] ∼ −4.0 up to ∼ −1.0 (Greif et al. 2011), without
a very pronounced peak in the [Fe/H] distribution. This
behaviour is found in Segue 1 (Frebel et al. 2014) which
now presumably contains only those first Pop II stars.
In Boo¨tes I, a similar population of the earliest Pop II
stars could have formed, with their highest metallicity
stars having [Fe/H] ∼ −2.0. Any additional Boo¨tes I
stars located at and near the peak region of the MDF
might then be stars from the second and/or later stellar
generations formed from more homogeneous gas enriched
by any first/early Pop II supernovae. Still, star formation
must have been rapidly quenched to prevent stars with
[Fe/H] > −1.8 to form.
Regarding the MDF shape, there is a steady increase
from [Fe/H] ∼ −4.0 to the distinct peak of the distribu-
Fig. 7.— Metallicity distribution function for Boo¨tes I. Note the
lack of metal-rich stars above [Fe/H] = −1.8. Top panel: Complete
sample of 39 stars. Bottom panel: red histogram depicts CEMP
stars with [C/Fe] > 0.7, black histogram the remaining non-CEMP
stars. Additional discussions about the CEMP histogram in part
e) of this section.
tion at [Fe/H] ∼ −2.5. Then, there is a relatively sharp
drop-off towards higher metallicities suggesting that star
formation was extinguished rather abruptly, preventing
the formation of stars with [Fe/H] > −1.8. The same
shape of the metallicity distribution has also been found
in other, more luminous dwarf galaxies (e.g., Draco and
Carina, although they have higher peak metallicities;
Fig 17 of Norris et al. 2010c). Since this behavior is thus
not unique to Boo¨tes I we take it as a sign of chemical
evolution (as opposed to a limited chemical enrichment),
although likely in its early phases given the low peak
metallicity.
Lai et al. (2011) explored the nature of the Boo¨tes I
MDF with simple chemical evolution models, following
Kirby et al. (2011). Taking into account our slightly
more populated MDF which also contains more stars
with high-resolution [Fe/H] measurements, it appears
that their leaky-box model with pre-enriched initial gas
10 Frebel et al.
TABLE 4
Chemical Abundances of Key Elements of Boo¨tes I Stars from the Literature
Name Teff log g [Fe/H] [C/Fe] [Mg/Fe] [Ca/Fe] [Ti/Fe] [Sr/Fe] [Ba/Fe] Comment Ref.
Boo02 5114 2.00 −2.37 0.36 0.28a · · · · · · · · · · · · med-res LAI11
Boo03 5127 1.99 −3.09 0.79 0.43a · · · · · · · · · · · · med-res LAI11
Boo04 5210 2.68 −2.39 0.34 0.35a · · · · · · · · · · · · med-res LAI11
Boo05 5077 2.55 −2.89 <0.00 0.05a · · · · · · · · · · · · med-res LAI11
Boo06 5404 2.40 −2.20 0.64 0.32a · · · · · · · · · · · · med-res LAI11
Boo07 5200 2.54 −2.49 0.39 0.16a · · · · · · · · · · · · med-res LAI11
Boo-7 4800 1.60 −2.32 −0.50 · · · · · · · · · · · · · · · med-res NOR10
· · · · · · −2.33 · · · 0.30 0.23 · · · · · · −0.75 high-res FEL09b,c
Boo08 5178 2.30 −2.48 0.32 0.18a · · · · · · · · · · · · med-res LAI11
Boo-8 5090 2.30 −2.75 0.60 · · · · · · · · · · · · · · · med-res NOR10
Boo09 5563 2.44 −2.65 0.90 0.37a · · · · · · · · · · · · med-res LAI11
Boo-9 4630 1.10 −2.67 −0.55 · · · · · · · · · · · · · · · med-res NOR10
4750 1.40 −2.64 < −0.29 0.13 0.18 0.27 · · · −0.89 high-res ISH14b,c
Boo10 5086 2.42 −2.59 <0.00 · · · · · · · · · · · · · · · med-res LAI11
Boo11 5199 2.65 −2.43 −0.05 0.05a · · · · · · · · · · · · med-res LAI11
Boo12 5168 2.19 −2.48 0.24 0.04a · · · · · · · · · · · · med-res LAI11
Boo13 5631 2.35 −2.49 <0.00 0.23a · · · · · · · · · · · · med-res LAI11
Boo14 5971 2.56 −2.57 <0.00 0.34a · · · · · · · · · · · · med-res LAI11
Boo15 5117 2.31 −2.89 <0.00 0.30a · · · · · · · · · · · · med-res LAI11
Boo18 5287 2.27 −2.51 0.44 0.13a · · · · · · · · · · · · med-res LAI11
Boo19 5141 2.38 −3.29 0.40 0.28a · · · · · · · · · · · · med-res LAI11
Boo20 4931 2.44 −2.42 −0.35 0.13a · · · · · · · · · · · · med-res LAI11
Boo22 4866 1.93 −2.87 −0.16 0.12a · · · · · · · · · · · · med-res LAI11
Boo23 5475 2.83 −2.90 1.86 0.16a · · · · · · · · · · · · med-res LAI11
Boo25 5141 2.02 −2.76 1.34 0.25a · · · · · · · · · · · · med-res LAI11
Boo28 5449 3.29 −2.31 0.27 0.06a · · · · · · · · · · · · med-res LAI11
Boo30 5449 3.29 −1.86 0.38 0.51a · · · · · · · · · · · · med-res LAI11
Boo-33 4730 1.40 −2.29 0.30 · · · · · · · · · · · · · · · med-res NOR10
4600 1.00 −2.52 · · · 0.69 0.40 · · · · · · −0.40 high-res FEL09
4740 1.40 −2.32 · · · 0.26 0.14 −0.02 · · · −0.35 high-res GIL13b,c
Boo-34 4840 1.60 −3.10 0.55 · · · · · · · · · · · · · · · med-res NOR10
Boo-41 4750 1.60 −1.93 −0.65 · · · · · · · · · · · · · · · med-res NOR10
(=Boo24) 4798 1.63 −1.65 −0.80 0.46a · · · · · · · · · · · · med-res LAI11
4750 1.50 −1.88 · · · 0.50 0.28 0.78 · · · −0.39 high-res GIL13b,c
Boo-78 4950 1.90 −2.46 −0.15 · · · · · · · · · · · · · · · med-res NOR10
Boo-94 4570 0.80 −2.90 −0.45 · · · · · · · · · · · · · · · med-res NOR10
4600 0.50 −2.95 · · · 0.47 · · · 0.22 · · · · · · high-res FEL09
4560 0.80 −2.94 · · · 0.49 0.30 0.26 · · · −0.95 high-res GIL13b,c,d
4500 0.80 −3.18 <0.25 0.39 0.46 0.55 −2.16 −0.80 high-res ISH14
Boo-117 4700 1.40 −2.25 −0.30 · · · · · · · · · · · · · · · med-res NOR10
(=Boo01) 4716 1.65 −2.34 −0.50 0.12a · · · · · · · · · · · · med-res LAI11
4600 1.00 −2.29 · · · 0.22 0.29 · · · · · · −0.46 high-res FEL09
4725 1.40 −2.18 · · · 0.18 0.20 0.14 · · · −0.65 high-res GIL13b,c,d
4750 1.50 −2.15 −0.79 0.04 0.01 0.44 −1.12 −0.43 high-res ISH14
Boo-119 4770 1.40 −3.33 · · · 1.04 0.46 0.80 · · · −1.00 high-res GIL13c,e
(=Boo21) 4775 1.48 −3.79 2.20 0.27a · · · · · · · · · · · · med-res LAI11
Boo-121 4630 1.10 −2.37 −0.25 · · · · · · · · · · · · · · · med-res NOR10
· · · · · · −2.44 · · · 0.62 0.38 · · · · · · −0.43 high-res FEL09
4500 0.80 −2.49 < −0.24 0.20 0.24 0.07 · · · −0.63 high-res ISH14b,c
Boo-127 4670 1.40 −2.08f −0.25 · · · · · · · · · · · · · · · med-res NOR10
4600 1.00 −2.03 · · · 0.76 0.02 · · · · · · − 0.64 high-res FEL09
4685 1.40 −2.01 · · · 0.17 0.16 0.20 · · · −0.69 high-res GIL13
4750 1.60 −1.92 −0.77 0.11 0.19 0.20 −1.44 −0.87 high-res ISH14
4765 1.35 −1.99 −0.15 0.41 0.24 0.33 −1.20 −0.75 high-res this studyb
Boo-130 4750 1.40 −2.29 −0.40 · · · · · · · · · · · · · · · med-res NOR10
4775 1.40 −2.32 · · · 0.21 0.19 0.17 · · · −0.54 high-res GIL13b,c,d
Boo-911 4540 1.10 −1.98 −0.55 · · · · · · · · · · · · · · · med-res NOR10
· · · · · · −2.26 · · · 0.11 0.40 · · · · · · −0.56 high-res FEL09
4500 0.90 −2.16 −0.77 0.35 −0.01 0.02 · · · −0.64 high-res ISH14b
Boo-980 4890 1.70 −3.09 0.00 · · · · · · · · · · · · · · · med-res NOR10
4720 1.40 −3.06 −0.40 0.29 0.30 0.43 −0.95 −1.23 high-res this studyb
Boo-1069 5050 2.20 −2.51 −0.35 · · · · · · · · · · · · · · · med-res NOR10
Boo-1137 4710 1.20 −3.66 0.20 · · · · · · · · · · · · · · · med-res NOR10
4700 1.20 −3.66 0.25 0.47 0.52 0.52 −1.42 −0.59 high-res NOR10bb
Adopted abundances of high-resolution spectroscopic studies
Boo-7 · · · · · · −2.33 −0.49(0.07g) 0.30 0.23 · · · · · · −0.75 high-res FEL09,NOR10
Boo-9 4750 1.40 −2.64 −0.58(−0.02g) 0.13 0.18 0.27 · · · −0.89 high-res ISH14,NOR10
Boo-33 4740 1.40 −2.32 0.29(0.74g) 0.26 0.14 −0.02 · · · −0.35 high-res GIL13,NOR10
Boo-41 4750 1.50 −1.88 −0.78(−0.24g) 0.50 0.28 0.78 · · · −0.39 high-res GIL13,NOR10
Boo-94 4560 0.80 −2.94 −0.44(0.31g) 0.49 0.30 0.26 −2.43 −0.95 high-res GIL13,NOR10,ISH14
Boo-117 4725 1.40 −2.18 −0.50(0.04g) 0.18 0.20 0.14 −1.22 −0.65 high-res GIL13,NOR10,ISH14
Boo-119 4770 1.40 −3.33 1.85(2.16g) 1.04 0.46 0.80 · · · −1.00 high-res GIL13,LAI11
Boo-121 4500 0.80 −2.49 −0.13(0.62g) 0.20 0.24 0.07 · · · −0.63 high-res ISH14,NOR10
Boo-127 4765 1.35 −1.99 −0.15(0.37g) 0.41 0.24 0.33 −1.20 −0.75 high-res FRE14
Boo-130 4775 1.40 −2.32 −0.41(0.17g) 0.21 0.19 0.17 · · · −0.54 high-res GIL13,NOR10
Boo-911 4500 0.90 −2.16 −0.77(−0.01g) 0.35 −0.01 0.02 · · · −0.64 high-res ISH14
Boo-980 4720 1.40 −3.06 −0.40(0.13g) 0.29 0.30 0.43 −0.95 −1.23 high-res FRE14
The chemical evolution of Boo¨tes I 11
would best describe the low metallicity tail of Boo¨tes I.
That model begins with pre-enriched gas of low level of
[Fe/H]0 ∼ −4.0 which is close to the assumption of pris-
tine gas. However, it would simultaneously underpredict
the very tall peak region of the MDF. We speculate that
this could signify a subsequent population of stars that
formed after the initial generation emerged from gas en-
riched by the first/early Pop II supernovae.
b) Core-collapse supernova signatures This cri-
terion stipulates that light elements (including iron-peak
elements) observed in early dwarf galaxy stars should
match those of equivalent metal-poor halo stars as core-
collapse supernovae are believed to be the progenitors of
these elements. As can be seen in Figure 3, the observed
abundances of stars in Boo¨tes I are in good agreement
with metal-poor halo star abundances, in line with a fast
enrichment in the earliest dwarf galaxies.
c) α-element abundances: late-time star for-
mation? This criterion stipulates that the α-elements
(Mg, Si, Ca, Ti) should show enhanced abundances of ∼
0.4 dex due to core-collapse supernova as progenitors for
all stars, even at high(er) metallicities ([Fe/H] > −2.0)
since it implies that no stars formed after (any) late-time
enrichment of iron by supernovae Ia. This would sup-
port that only one/few generations of early Pop II stars
formed in a system.
As noted by previous author (e.g., Gilmore et al. 2013),
the α-element abundances of stars in Boo¨tes I do not
provide a clear picture. Consideration of the individual
α-elements, e.g., in Figure 4, shows that all stars have
[Mg/Fe] high-resolution abundance ratios in agreement
with halo stars. For [Ca/Fe] and [Ti/Fe], most Boo¨tes I
stars have typical halo values but there are also several
outliers with abundance near the solar ratio. However, in
all cases, the 2-3 most metal-rich stars do not have the
lowest abundances but rather halo-like values. In fact,
the highest metallicity star (Boo-41) has one of the high-
est α-element abundances ([α/Fe] = 0.52). This supports
Boo¨tes I containing those early Pop II stars, although not
exclusively.
The combined α-element abundances are then help-
ful to assess whether Boo¨tes I experienced late time,
extended star formation as evidenced by low(er) α-
abundances. This is illustrated in Figure 8. In the
bottom two panels, we show the Boo¨tes I abundances
together with abundances of other ultra-faint dwarf
galaxies as well as those of stars in the classical dwarf
spheroidal galaxies. Several Boo¨tes I stars have abun-
dances resembling those of the halo, Segue 1 and other
ultra-faint dwarf galaxy stars while many others have
values very similar to those of the classical dwarf galaxy
stars (21 have [α/Fe] < 0.3 and 16 [α/Fe] < 0.2).
It thus appears like Boo¨tes I is a system that was as-
sembled from or absorbed one (or a few) smaller building
block-type objects (e.g., minihalos) like Segue 1 before
forming more stars and experiencing a chemical evolu-
tion (as opposed to just chemical enrichment as, e.g.,
in Segue 1) that would eventually lead to stars with low
α-abundances. Such a two-population scenario is qual-
itatively in line what was found based on the CEMP
and non-CEMP star MDF components of Boo¨tes I. More
precise α-element abundances will be needed for all stars
(e.g., recall that Lai et al. 2011 only provided a combined
α-measurement) to conclusively determine whether any
Fig. 8.— Combined Mg-Ca-Ti α-element abundances as diagnos-
tic of early star formation; adapted from Frebel & Bromm (2012).
Top: Schematic representation of chemical enrichment in the [α/Fe]
vs [Fe/H] plane for different environments. The dotted line indicates
the solar ratio. Second: High-resolution α-abundances of metal-poor
stars from Cayrel et al. (2004) (halo) and Fulbright (2000) (thin/thick
disk). The diagonal dotted lines indicate the enrichment behavior of
the dSph galaxies (see top panel). A representative uncertainty is
shown. Third: The yellow shaded region around [α/Fe] = 0.35 de-
picts the predicted Frebel & Bromm (2012) one-shot enrichment be-
havior (with a 0.15 dex observational uncertainty) reflecting massive
core-collapse supernova enrichment. Several evolutionary paths are in-
dicated with dashed lines. High-resolution α-abundances of metal-poor
stars are shown; For ultra-faint dwarf galaxies: Boo¨tes I stars, this
study (filled red circles) Boo¨tes I, literature (open red circles; see text
for references), Segue1 (blue circles; Norris et al. 2010a; Frebel et al.
2014); Ursa Major II and Coma Berenices (green circles; Frebel 2010);
Leo IV (pink circle; Simon et al. 2011); Hercules (cyan squares; Koch
et al. 2008). For classical dSphs (small open black circles; see Frebel
& Bromm 2012 and references therein). Bottom: Same as above but
with only the Boo¨tes I stars that include the [α/Fe] measurements of
Lai et al. (2011).
late time enrichment by supernova typeI˙a did indeed oc-
cur in Boo¨tes I. This would indicate extended star forma-
tion possibly due to an gas accretion from other clouds
or cloud fragments in the system or an early merger with
a different system that would have brought in extra gas
(Smith et al. 2015).
d) Low and unusual neutron-capture element
abundances: evidence for one progenitor genera-
tion? This criterion stipulates that abundances of heavy
neutron-capture elements (e.g., Sr and Ba) should be
very low since it implies that no stars formed after late(r)
time enrichment associated with the s-process operating
in AGB stars. Small amounts of neutron-capture ele-
ments, primarily with [Sr/Ba] ratios characteristic of the
12 Frebel et al.
r-process (occurring in supernovae) could, however, be
present in the system without implying late time star
formation.
We have already discussed that the neutron-capture
elements Sr and Ba are of low abundance in Boo¨tes I
(Figure 5). Specifically, the Sr abundances are not like
those of halo stars but up to ∼ 1 dex lower. Ba does
roughly follow the halo star abundance trend but all
Boo¨tes I stars still have abundances that place them at
the lower edge of the region covered by halo data. In
addition, the spread in neutron-capture element abun-
dances in Boo¨tes I, as evidenced by only four stars, is
about 1 dex. This precludes overarching statements re-
garding the exact neutron-capture process responsible
for this material. The stars with [Sr/Ba] ∼ −0.5 ratio
technically fall in the regime of the r-process (we note
that most Segue 1 stars are also near this value). The
s-process is characterized by [Sr/Ba] < −1.0 and indeed
one Boo¨tes I star has such low [Sr/Ba] value. Overall,
this is consistent with Boo¨tes I being an old system that
formed its stars early and relatively quickly, before the
onset of any regular AGB star-based s-process element
enrichment. Consequently, fast neutron-capture enrich-
ment might have occurred by just one of the earliest gen-
erations of massive stars in the system.
In Figure 9, there exists no obvious correlation between
[Sr/Ba] and [Fe/H]. When plotting [Sr/Ba] vs. [Ba/Fe]
(see bottom panel). However, a clear trend emerges. In-
terestingly, the halo stars form a distinct branch while
the various dwarf galaxy stars form something that re-
sembles a second sequence below the halo branch. The
few Boo¨tes I stars with measured Sr abundances follow
this trend set by the UMa II, ComBer, Leo IV, and three
Segue 1 stars. Some halo stars do exhibit similar neutron-
capture element characteristics to those of the dwarf
galaxy stars. This might indicate that these halo stars
are stars that originated in small dwarf galaxies similar
to those in the surviving ultra-faint dwarfs. While more
data for ultra-faint dwarf galaxy stars is clearly needed to
more firmly establish this second sequence, perhaps also
more halo data could help shape up any of the differ-
ences between the halo star branch and the dwarf galaxy
sequence even more clearly.
Using neutron-capture element abundances is clearly a
promising way path to learn not only about the nature
of the host dwarf galaxy but also about the origin of halo
stars. Lighter element abundances have been repeatedly
shown to all agree with each other for halo and dwarf
galaxy stars making it more challenging to use them as
a discriminator.
Alternatively, the distinct patterns of halo stars and
dwarf galaxy stars could originate from highly inhomo-
geneous enrichment of their respective birth gas clouds.
However, while this would explain the large spreads in
[element/H] abundance ratios such as [Fe/H] or [Sr/H],
[element/element] ratios should in principle be hardly af-
fected (Frebel & Bromm 2012). Indeed, the lighter ele-
ments show tight [element/Fe] ratios and support this
possibility. In turn, the large spread in [Sr/Ba] might
indicate that the two elements are not produced with a
strong correlation between them as the fusion-produced
elements are, or that metal mixing did not play a primary
role.
e) Carbon enrichment: prevalent at early
Fig. 9.— [Sr/Ba] abundance ratios as a function of [Fe/H] (top
panel) and [Ba/Fe] (bottom panel) of our Boo¨tes I stars (filled red cir-
cles) and literature Boo¨tes I stars (open red circles; here the Ishigaki
et al. 2014 values were used for internal consistency) in comparison
with those of other ultra faint dwarf galaxy stars in Segue 1 (blue cir-
cles; Frebel et al. 2014; Norris et al. 2010a), UMa II, ComBer, Leo IV
(green circles; Frebel et al. 2010; Simon et al. 2010), and the galactic
halo (black open circles; Barklem et al. 2005 and Yong et al. 2013).
times? After applying carbon abundance corrections ac-
cording to stellar evolutionary status (see Table 4), there
are eight CEMP stars in Boo¨tes I with metallicities be-
low [Fe/H] . −2.6. Incidentally, the CEMP stars are all
CEMP-no stars and rather equally distributed among the
lower metallicity half of all Boo¨tes I stars. All stars have
[C/H] values higher than [C/H] = −3.4 placing them well
above the critical carbon and oxygen abundance criterion
of Dtrans = log(10
[C/H]+0.3×10[O/H]) = −3.5. that may
have facilitated their formation from carbon and oxygen
enriched gas in the early universe (Frebel et al. 2007).
The chemical evolution of Boo¨tes I 13
Carbon must thus have played an important role in the
early evolution of Boo¨tes I.
[C/H] ratios of the Boo¨tes I sample do not show a clear
trend with either [Fe/H] or log g, although the sample
does follow the overall carbon abundance range given by
halo stars (see Figure 2, also for references). The CEMP
stars in Boo¨tes I increasingly appear all at lower metallic-
ities, at [Fe/H] ≤ −2.6. Generally, the high-carbon out-
liers may indicate that the production of carbon and iron
and their subsequent mixing into the gas has been decou-
pled or that CEMP and non-CEMP signatures have been
produced at different sites. For a more extended discus-
sion on the origin of CEMP stars, we refer the reader to
Norris et al. (2013) or Frebel & Norris (2015).
Cooke & Madau (2014) suggested a minihalo envi-
ronment as a site for CEMP star formation with faint
fall-back first supernovae (e.g., Umeda & Nomoto 2003)
responsible for large quantities of lighter elements such
as carbon but for less or none of the heavier ones like
iron. The bottom panel of Figure 7 shows the sepa-
rate MDFs for CEMP stars (red) and non-CEMP stars
(black). The CEMP stars cluster at low metallicities
albeit with some spread in [Fe/H]. But chemical inho-
mogeneity could possibly account for this (e.g., Frebel
& Bromm 2012). These CEMP stars thus might have
formed in a minihalo environment that would later be-
come Boo¨tes I or part of Boo¨tes I.
The non-CEMP star MDF lacks the pronounced low-
metallicity tail and an average metallicity of [Fe/H] =
−2.5. This leaves the MDF to be more like the one pre-
dicted by the leaky box chemical evolution model with
extra gas (Lai et al. 2011; Kirby et al. 2011). This
incorporates an additional reservoir of gas available to
the galaxy from which to form stars. A merger of said
minihalo with another early halo (presumably prior to it
forming any stars) could have provided this gas. These
findings build in some way on the conclusions of Gilmore
et al. (2013) who suggest that the “CEMP-no” stars (our
“CEMP stars”) formed rapidly and prior to the “normal
branch” (our “non-CEMP” stars).
More generally, massive rotating Pop III “spinstars”
(Meynet et al. 2006; Chiappini et al. 2011) could have
dominated the enrichment of Boo¨tes I at the earliest
times by providing large amounts on carbon to the star
forming gas prior to their explosions. Fall-back super-
novae (Umeda & Nomoto 2003) could have been equally
responsible for high [C/Fe] ratios in environments more
massive than minihalos. Regardless, the earliest low-
mass stars would likely show a variety of [C/Fe] ratios,
and a clear trend would only emerge with the onset of
more regular Pop II supernovae and extended star for-
mation.
5. CONCLUSION
We have studied the stellar chemical abundances of
Boo¨tes I with high-resolution spectroscopy, including a
previously unobserved member star. Considering all
available abundance data, we examined Boo¨tes I in light
of several criteria for identifying the earliest galaxies.
The α- and carbon abundances suggest two potential
populations, although more precise abundances are re-
quired to draw firm conclusions. Considerations of these
chemical signatures alone suggests that Boo¨tes I is not a
surviving first galaxy but an already somewhat assem-
bled system that could have been built up from a mini-
halo (a first galaxy) in which the CEMP stars formed
(Cooke & Madau 2014), and then merged with another
metal-free or metal-poor halo (or large gas clouds or fil-
aments) that resulted in an injection of fresh gas trigger-
ing star formation and leading to a second population of
stars. Both populations show [Fe/H] spreads of > 1 dex
which is in line with predictions of chemical inhomogene-
ity (e.g., Frebel & Bromm 2012) and expected for such
early star formation.
Similar results have also been derived by Romano et al.
(2015) who modeled the formation and chemical evolu-
tion of Boo¨tes I. They conclude that Boo¨tes I formed from
accretion of a baryonic mass Mb ∼ 10
7M⊙ of gas over
a very short time scale of order 50Myr. This is also in
agreement with what was already qualitatively predicted
by Gilmore et al. (2013) who found that the CEMP stars
formed rapidly and prior to the non-CEMP stars.
All these findings are highly consistent with results
based on deep HST photometry. Brown et al. (2014)
found Boo¨tes I to currently consist of two ancient stel-
lar populations of 13.4Gyr (containing 3% of stars) and
13.3Gyr (97% of stars). We note, though, that these
ages are relative to the globular cluster M92, and assum-
ing its age to be 13.7Gyr. Nevertheless, it is apparent
that all the dwarf galaxies contain one ancient dominant
stellar population plus a small age spread. The two-
population scenario follows Koposov et al. (2011) who
also found that Boo¨tes I is best described with two com-
ponents, based on a kinematic analysis. They identified
a “colder” (∼ 3 km s−1) component (containing 70% of
stars) and a “hotter” ∼ 9 km s−1 component (30% of
stars). In addition, they suggested that the colder com-
ponent have “an extremely extended, very metal-poor,
low-velocity dispersion component”. The chemical abun-
dances and the MDF available for Boo¨tes I stars clearly
show the existence of such a low-metallicity tail (see Fig-
ure 7) that is mainly composed of CEMP stars.
In summary, evidence based on multiple independent
approaches suggest Boo¨tes I to be an ancient system with
two stellar populations of which at least the more metal-
poor one likely consisted of the first low-mass stars that
formed in its progenitor halo(s). This makes Boo¨tes I an
assembled system with a more complex evolution than
what would be expected of a surviving galactic building
block, such as Segue 1. Nevertheless, Boo¨tes I is likely
one of the earliest galaxies to have formed. Not just
for its age of ∼ 13.4Gyr (Brown et al. 2014) but also
because we seem to be witnessing what is the product
of two systems, one minihalo and one larger halo which
merged to form Boo¨tes I. To some extent this outcome
is already indicated by its larger luminosity of nearly
∼ 105 L⊙, compared to e.g., Segue 1 with only about
∼ 103 L⊙. This leaves the question as to whether an
average metallicity of an early galaxy is a useful quantify
if more than one population contributes stars in different
metallicity ranges. Ideally, one could cleanly disentangle
the population(s) to report separate results. This might
work for Boo¨tes I but will surely become impossible for
more complex, more luminous systems.
More ultra-faint galaxies need to be extensively stud-
ied with photometry and spectroscopy from a chemical
and kinematic point of view to assess the nature of as
many as possible of the faint dwarfs currently known.
14 Frebel et al.
Only then can be better understand the early evolution of
galaxies and the details of the associated star formation,
chemical enrichment and overall growth mechanisms and
timescales of these ancient systems.
A.F. is supported by NSF CAREER grant AST-
1255160. She also acknowledges support from the Silver-
man (1968) Family Career Development Professorship.
Studies at RSAA, ANU, of the Galaxy’s most metal-
poor stars and ultra-faint satellite systems are supported
by Australian Research Council grants DP0663562,
DP0984924, DP120100475, and DP150100862 which
J.E.N. gratefully acknowledges. This work made use
of the NASA’s Astrophysics Data System Bibliographic
Services.
Facilities: Magellan-Clay (MIKE)
REFERENCES
Ade´n, D., Eriksson, K., Feltzing, S., et al. 2011, A&A, 525, 153
Alonso, A., Arribas, S., & Mart´ınez-Roger, C. 1999, A&As, 140,
261
Aoki, W., Beers, T. C., Christlieb, N., et al. 2007, ApJ, 655, 492
Aoki, W., Honda, S., Beers, T. C., et al. 2005, ApJ, 632, 611
Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009,
ARA&A, 47, 481
Barklem, P. S., Christlieb, N., Beers, T. C., et al. 2005, A&A, 439,
129
Beers, T. C., & Christlieb, N. 2005, ARA&A, 43, 531
Belokurov, V., Zucker, D. B., Evans, N. W., et al. 2006, ApJL, 647,
L111
Bernstein, R., Shectman, S. A., Gunnels, S. M., Mochnacki, S., &
Athey, A. E. 2003, in Society of Photo-Optical Instrumentation
Engineers (SPIE) Conference Series, ed. M. Iye & A. F. M.
Moorwood, Vol. 4841, 1694
Bland-Hawthorn, J., Krumholz, M. R., & Freeman, K. 2010, ApJ,
713, 166
Brown, T. M., Tumlinson, J., Geha, M., et al. 2014,
Mem. Soc. Astron. Italiana, 85, 493
Casey, A. R. 2014, arXiv:1405.5968
Castelli, F., & Kurucz, R. L. 2004, arXiv:astro-ph/0405087
Cayrel, R., Depagne, E., Spite, M., et al. 2004, A&A, 416, 1117
Chiappini, C., Frischknecht, U., Meynet, G., et al. 2011, Nature,
472, 454
Cooke, R. J., & Madau, P. 2014, ApJ, 791, 116
Feltzing, S., Eriksson, K., Kleyna, J., & Wilkinson, M. I. 2009,
A&A, 508, L1
Franc¸ois, P., Depagne, E., Hill, V., et al. 2007, A&A, 476, 935
Frebel, A. 2010, Astronomische Nachrichten, 331, 474
Frebel, A., & Bromm, V. 2012, ApJ, 759, 115
Frebel, A., Casey, A. R., Jacobson, H. R., & Yu, Q. 2013, ApJ,
769, 57
Frebel, A., Johnson, J. L., & Bromm, V. 2007, MNRAS, 380, L40
Frebel, A., & Norris, J. E. 2015, ARA&A, 53, 631
Frebel, A., Simon, J. D., Geha, M., & Willman, B. 2010, ApJ, 708,
560
Frebel, A., Simon, J. D., & Kirby, E. N. 2014, ApJ, 786, 74
Fulbright, J. P. 2000, AJ, 120, 1841
Fulbright, J. P., Rich, R. M., & Castro, S. 2004, ApJ, 612, 447
Gilmore, G., Norris, J. E., Monaco, L., et al. 2013, ApJ, 763, 61
Greif, T. H., Springel, V., White, S. D. M., et al. 2011, ApJ, 737,
75
Hollek, J. K., Frebel, A., Roederer, I. U., et al. 2011, ApJ, 742, 54
Hughes, J., Wallerstein, G., & Bossi, A. 2008, AJ, 136, 2321
Hughes, J., Wallerstein, G., Dotter, A., & Geisler, D. 2014,
MNRAS, 439, 788
Ishigaki, M. N., Aoki, W., Arimoto, N., & Okamoto, S. 2014, A&A,
562, A146
Ji, A. P., Frebel, A., & Bromm, V. 2015, MNRAS, 454, 659
Ji, A. P., Frebel, A., Chiti, A., & Simon, J. D. 2016a,
arXiv:1512.01558
Ji, A. P., Frebel, A., Simon, J. D., & Geha, M. 2016b, ApJ, 817,
41
Jordi, K., Grebel, E. K., & Amman, K 2006, A&A, 460, 339
Karlsson, T., Bland-Hawthorn, J., Freeman, K. C., & Silk, J. 2012,
ApJ, 759, 111
Kelson, D. D. 2003, PASP, 115, 688
Kim, Y.-C., Demarque, P., Yi, S. K., & Alexander, D. R. 2002,
ApJS, 143, 499
Kirby, E. N., Lanfranchi, G. A., Simon, J. D., Cohen, J. G., &
Guhathakurta, P. 2011, ApJ, 727, 78
Kirby, E. N., Simon, J. D., Geha, M., Guhathakurta, P., & Frebel,
A. 2008, ApJL, 685, L43
Koch, A., Feltzing, S., Ade´n, D., & Matteucci, F. 2013, A&A, 554,
A5
Koch, A., McWilliam, A., Grebel, E. K., Zucker, D. B., &
Belokurov, V. 2008, ApJ, 688, L13
Koch, A., & Rich, R. M. 2014, ApJ, 794, 89
Koposov, S. E., Gilmore, G., Walker, M. G., et al. 2011, ApJ, 736,
146
Lai, D. K., Lee, Y. S., Bolte, M., et al. 2011, ApJ, 738, 51
Martin, N. F., Ibata, R. A., Chapman, S. C., Irwin, M., & Lewis,
G. F. 2007, MNRAS, 380, 281
Meynet, G., Ekstro¨m, S., & Maeder, A. 2006, A&A, 447, 623
Mun˜oz, R. R., Carlin, J. L., & Frinchaboy, P. M. a. 2006, ApJ,
650, L51
Norris, J. E., Gilmore, G., Wyse, R. F. G., et al. 2008, ApJ, 689,
L113
Norris, J. E., Gilmore, G., Wyse, R. F. G., Yong, D., & Frebel, A.
2010a, ApJ, 722, L104
Norris, J. E., Wyse, R. F. G., Gilmore, G., et al. 2010b, ApJ, 723,
1632
Norris, J. E., Yong, D., Bessell, M. S., et al. 2013, ApJ, 762, 28
Norris, J. E., Yong, D., Gilmore, G., & Wyse, R. F. G. 2010c, ApJ,
711, 350
Okamoto, S., Arimoto, N., Yamada, Y., & Onodera, M. 2012,
Ap.J., 744, 96
Placco, V. M., Frebel, A., Beers, T. C., & Stancliffe, R. J. 2014,
ApJ, 797, 21
Romano, D., Bellazzini, M., Starkenburg, E., & Leaman, R. 2015,
MNRAS, 446, 4220
Siegel, M. H. 2006, ApJ, 649, L83
Simon, J. D., Frebel, A., McWilliam, A., Kirby, E. N., &
Thompson, I. B. 2010, ApJ, 716, 446
Simon, J. D., Geha, M., Minor, Q. E., et al. 2011, ApJ, 733, 46
Smith, B. D., Wise, J. H., O’Shea, B. W., Norman, M. L., &
Khochfar, S. 2015, MNRAS, 452, 2822
Sneden, C. A. 1973, PhD thesis, The University of Texas at Austin
Sobeck, J. S., Kraft, R. P., Sneden, C., et al. 2011, AJ, 141, 175
Umeda, H., & Nomoto, K. 2003, Nature, 422, 871
Venn, K. A., Irwin, M., Shetrone, M. D., et al. 2004, AJ, 128, 1177
Webster, D., Frebel, A., & Bland-Hawthorn, J. 2016, ApJ, 818, 80
Yong, D., Norris, J. E., Bessell, M. S., et al. 2013, ApJ, 762, 26