EVIDENCE THAT HYDRA I IS A TIDALLY DISRUPTING MILKY WAY DWARF GALAXY

Observations of Hydra I were obtained on 2014 March 20–23 using the DECam imager installed on the CTIO Blanco 4 m telescope. The DECam imager consists of sixty-two 2048 × 4096 imaging CCDs (chip gaps of ∼50'') arranged in a circular layout in the focal plane. On the Blanco telescope, this layout provides imaging over a 21 diameter, resulting in a 3 square degree imaging area with 026 pixels.

We imaged Hydra I using the ugri filters, but to facilitate comparisons with other studies of Hydra I/EBS and the Galactic anti-center region, we only present the gr imaging in this work. We used a series of 6 × 180 s exposures in each of the g and r filters and dithered between exposures to fill in the chip gaps. Figure 1 shows the resulting footprint of the DECam imaging (∼9 deg²) on the Hydra I field. All imaging was obtained under photometric or near-photometric conditions. The mean FWHM of the image point-spread functions for the exposures ranged from 09 to 12.

The data were reduced using the NOAO Community Pipeline (CP; Valdes et al. 2014) and downloaded from the NOAO Science Archive. The CP performs both a standard instrumental calibration of the raw images (bias, overscan, cross-talk corrections; fringing corrections; dome flat and sky flat field corrections) and corrects for geometric distortions by reprojecting the images onto a common spatial scale.

We performed photometry on each DECam exposure by separately analyzing each of the 62 reduced, reprojected chips using the DAOPHOTII/ALLSTAR software suite (Stetson 1987, 1994). We derive a catalog of point sources using cuts on the DAOPHOT CHI and SHARP parameters which vary as a function of magnitude depth. To ensure a robust characterization of sources, we require that object must be point-like in each of the gri images.

The Hydra I/EBS region is located in the SDSS footprint, and so we derive calibrated magnitudes by matching directly to SDSS Data Release 8 (SDSS DR8; Aihara et al. 2011). We derived a photometric solution for each individual DECam exposure separately, after correcting the instrumental magnitudes for atmospheric extinction. We derived zero points and linear color terms using the maximum likelihood approach described by Boettcher et al. (2013), using only point sources which have high signal-to-noise ratio (S/N) in both the DECam and SDSS imaging and SDSS colors of $g-r\lt 1$. We impose the latter constraint in order to improve the photometric calibration over the color region of interest for this study. Lastly, all calibrated apparent magnitudes presented in this study were corrected for Galactic extinction using the reddening maps of Schlegel et al. (1998) and the reddening coefficients of Schlafly & Finkbeiner (2011). The mean color excess in the Hydra I region imaged by DECam is $E(B-V)$ ≃ 0.04 mag and varies by ±0.02 mag across the field of view.

The final Hydra I DECam point source catalog was constructed by calculating the weighted mean magnitude of each point source from the individual measurements on each frame. We used the uncertainties in the calibrated magnitudes (in flux space) as the weights and used an iterative sigma-clipping algorithm to reject any measurements which fell >4σ from the mean. The uncertainties in the weighted mean magnitudes were calculated from the inverse square of the flux errors. The final catalog contains ∼17,000 sources between $17\lt g\lt 22$ mag over a ∼9 deg² area. We present the entire photometric data set in Table 1, which is available in machine readable format from the publisher.

Table 1.  DECam Photometry

R.A.	Decl.	g0	${\sigma }_{{g}_{0}}$	r0	${\sigma }_{{r}_{0}}$	$E(B-V)$
(degree)	(degree)	mag	(mag)	(mag)	(mag)	(mag)
132.348949	3.153806	19.116	0.015	18.187	0.009	0.0318
132.350640	3.344006	18.993	0.013	18.308	0.009	0.0339
132.351339	3.363441	22.279	0.042	21.675	0.033	0.0340
132.351990	3.397592	22.226	0.068	21.581	0.038	0.0340
132.352598	3.438227	19.357	0.012	18.727	0.009	0.0338
132.352717	3.150096	22.082	0.037	22.058	0.050	0.0319
132.354884	3.307660	20.540	0.014	20.283	0.013	0.0341
132.355516	3.424209	20.756	0.016	20.193	0.011	0.0339
132.356115	3.223843	21.874	0.027	21.533	0.025	0.0334
132.358492	3.395449	23.487	0.106	22.591	0.070	0.0341

Note. DECam photometric point-source catalog of the EBS/Hydra I region (see footprint in Figure 1). The photometric data have been corrected for Galactic extinction (see Section 2.1) using the listed values of $E(B-V)$. We provide photometry only for stars with $g-r\lt 1$ due to the photometric calibration method employed in this study. Point-source selection was performed using the DAOPHOT CHI and SHARP parameters.

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

The number of point sources as a function of magnitude depth drops dramatically at g ≃ 23.5, r ≃ 22.8 mag. We therefore estimate that our imaging is complete at ∼1 mag brighter than these limits. A completeness limit at g ∼ 22–22.5 is sufficient for the purposes of this study. As discussed in Section 3, the main-sequence turn off (MSTO) in the color–magnitude diagram (CMD) of Hydra I is at g ≃ 18.75 (r ≃ 18.6) mag, significantly brighter than our estimated incompleteness depths.

We used MMT/Hectochelle to target stars throughout the Hydra I overdensity. Following the previous analysis by Grillmair (2011), we targeted stars from SDSS within 0.1 mag of a 13 Gyr, [Fe/H] = −1.8 isochrone (Dotter et al. 2008) at a distance of 9.7 kpc (however, see our conclusions in Section 4.3 regarding the age and metallicity of Hydra I). Figure 2 shows the selection filter on the SDSS color–magnitude-number density (Hess) diagram of stars within the ∼9 deg² footprint of the DECam imaging. We chose a CMD filter with a relatively blue edge in order to explore the possibility of younger (bluer) stellar populations in Hydra I. Because the expected red giant branch population of Hydra I overlaps with MW thick disk stars in color and magnitude (see Figure 2), objects fainter than g = 18 mag were given priority for our Hectochelle observations in order to minimize contamination.

Zoom In Zoom Out Reset image size

Spectroscopic observations were obtained over 7 nights using the Hectochelle spectrograph on the 6.5 m MMT. Hectochelle uses 240 fibers across a 1° diameter field of view to take single order spectra at a resolution of R ∼ 38,000. We used the RV31 order selecting filter to isolate the 150 Å around the Mg b triplet (5167–5184 Å). The field of view of Hectochelle is well suited to observing Hydra I. We obtained seven pointings around Hydra I, primarily in relatively bright or gray observing time, with total exposure times ranging from 4800 to 19,800 s per pointing. We targeted the areas on the primary overdensities and along the stream slightly southeast of the two main lobes. The footprint of the Hectochelle observations are shown in the right panel of Figure 1. We obtained spectra for a total of 796 stars with magnitudes in the range 17 < g < 21.3.

The multifiber spectra were reduced in a uniform manner. For each field, the separate exposures were debiased and flat fielded, and then compared before extraction to allow identification and elimination of cosmic rays through interpolation. Spectra were then extracted, combined and wavelength calibrated using contemporaneous exposures of a ThAr lamp (the spectra were not-linearized in dispersion). Each fiber has a distinct wavelength dependence in throughput, which can be estimated using exposures of a continuum source or the twilight sky. A background sky estimate was made by combining data from fibers placed randomly in the focal plane, but excluding those which accidentally fell on a source. This sky spectrum was finely interpolated onto the object spectrum's wavelength grid and then subtracted.

We model the Hectochelle spectra using the procedure introduced by Walker et al. (2015, hereafter W15) to model sky-subtracted Hectochelle spectra. The spectral model is given as

$$\begin{eqnarray}&&M(\lambda )={P}_{l}(\lambda )T\left(\lambda \left[1+\displaystyle \frac{{Q}_{m}(\lambda )+{v}_{{\rm{los}}}}{c}\right]\right),\end{eqnarray} \tag{ 1 }$$

where c is the speed of light and ${P}_{l}(\lambda )$ specifies the stellar continuum using an order-l polynomial of the form

$$\begin{eqnarray}{P}_{l}(\lambda ) & \equiv & {p}_{0}+{p}_{1}\left[\displaystyle \frac{\lambda -{\lambda }_{0}}{{\lambda }_{s}}\right]+{p}_{2}{\left[\displaystyle \frac{(\lambda -{\lambda }_{0})}{{\lambda }_{s}}\right]}^{2}\\ & & +\ldots +{p}_{l}{\left[\displaystyle \frac{(\lambda -{\lambda }_{0})}{{\lambda }_{s}}\right]}^{l}.\end{eqnarray} \tag{ 2 }$$

The factor of $T(\lambda [1+({Q}_{m}(\lambda )+{v}_{{\rm{los}}})/c])$ is a continuum-normalized template spectrum that is redshifted according to the line of sight velocity (v_los) and an order-m polynomial of the form

$$\begin{eqnarray}{Q}_{m}(\lambda ) & \equiv & {q}_{0}+{q}_{1}\left[\displaystyle \frac{\lambda -{\lambda }_{0}}{{\lambda }_{s}}\right]+{q}_{2}{\left[\displaystyle \frac{(\lambda -{\lambda }_{0})}{{\lambda }_{s}}\right]}^{2}\\ & & +...+{q}_{m}{\left[\displaystyle \frac{(\lambda -{\lambda }_{0})}{{\lambda }_{s}}\right]}^{m}\end{eqnarray} \tag{ 3 }$$

that compensates for systematic differences between wavelength functions of target and template spectra (see W15 for details). Following W15, we choose m = 2, providing sufficient flexibility to fit the observed continuum shape and to apply low-order corrections to the wavelength solution. We adopt scale parameters λ₀ = 5220 Å and ${\lambda }_{s}=60\;\mathring{\rm A}$, such that $-1\leqslant (\lambda -{\lambda }_{0})/{\lambda }_{s}\leqslant +1$ over the entire range (5160–5280 Å) considered in the fits.

Again following W15, we generate template spectra T using the synthetic library that was used for SEGUE Stellar Parameter Pipeline (SSPP) parameter estimation (Lee et al. 2008a, 2008b). The SSPP library contains rest-frame, continuum-normalized, stellar spectra computed over a grid of atmospheric parameters spanning $4000\leqslant {T}_{\mathrm{eff}}/{\rm{K}}\leqslant$ 10,000 in effective temperature (with spacing ${\rm{\Delta }}{T}_{\mathrm{eff}}/{\rm{K}}=250$), $0\;\leqslant {\mathrm{log}}_{10}[{\rm{g}}/(\mathrm{cm}/{{\rm{s}}}^{2})]\leqslant 5$ in surface gravity (${\rm{\Delta }}{\mathrm{log}}_{10}[{\rm{g}}/(\mathrm{cm}/{{\rm{s}}}^{2})]\quad =0.25$ dex) and $-5\leqslant [\mathrm{Fe}/{\rm{H}}]\leqslant +1$ in metallicity (${\rm{\Delta }}[\mathrm{Fe}/{\rm{H}}]=0.25$ dex). The library spectra are calculated for a range in [α/Fe]; however, this ratio depends on metallicity. The library has $[\alpha /{\rm{Fe}}]\;=\;+0.4$ for spectra with $[{\rm{Fe/H}}]\lt -1$, and then $[\alpha /{\rm{Fe}}]$ decreases linearly as metallicity increases from $-1\leqslant [{\rm{Fe/H}}]\lt 0$, with $[\alpha /{\rm{Fe}}]=0$ for $[{\rm{Fe/H}}]\geqslant 0$. The synthetic library spectra are calculated over the range 3000–10,000 Å at resolution 0.01 Å/pixel, which we degrade to 0.05 Å per pixel—twice as fine as our Hectochelle spectra.

Our spectral model, $M(\lambda )$, is specified fully by a vector of 13 free parameters. In order to fit the model and obtain estimates for each parameter, we follow W15's Bayesian approach and adopt the same priors specified in their Table 1. We use the software package MultiNest (Feroz & Hobson 2008; Feroz et al. 2009) to sample the posterior probability distribution functions (PDFs). As described in W15, the sampling of the PDFs allows one to evaluate the "Gaussianity" of the posterior PDFs by calculating the first four moments (mean, variance, skew, kurtosis) of the 1D marginalized PDFs of each parameter. W15 note that in the high S/N regime, the posterior PDFs become increasingly Gaussian and therefore allow one to define a measure of quality control. That is, the Gaussianity measure provides confidence that the calculated variance on the velocity is a good measure of the 1σ standard deviation ($68\%$ confidence interval). In practice, we keep only those spectra for which the model fits yield Gaussian shaped posterior PDFs as defined in W15 (see their Section 4.1 and Figure 4).

Table 2.  Hectochelle/MMT Spectroscopy

R.A.	Decl.	VR	${\sigma }_{{V}_{R}}$	g0	${\sigma }_{{g}_{0}}$	r0	${\sigma }_{{r}_{0}}$	$E(B-V)$
(degree)	(degree)	(km s−1)	(km s−1)	(mag)	(mag)	(mag)	(mag)	(mag)
134.208389	3.950312	96.84	0.44	17.241	0.005	16.744	0.003	0.0499
134.244110	3.806066	92.04	0.53	18.404	0.003	18.182	0.002	0.0479
134.257034	4.086879	39.57	1.10	19.455	0.005	19.146	0.004	0.0512
134.273636	4.103878	63.54	0.44	17.330	0.005	16.867	0.003	0.0510
134.260468	4.003685	97.81	0.44	17.555	0.005	17.140	0.003	0.0505
134.115387	3.871296	178.97	0.35	18.335	0.004	17.902	0.003	0.0514
134.117737	3.833826	31.85	1.47	20.291	0.006	19.858	0.004	0.0501
134.071915	3.766812	85.44	0.97	19.602	0.004	19.375	0.003	0.0486
134.084824	3.745721	65.47	0.34	17.731	0.003	17.330	0.003	0.0475
134.113266	3.739799	20.37	0.46	17.905	0.004	17.449	0.003	0.0464

Note. Hectochelle/MMT catalog of all 411 stars passing the spectroscopic quality control cuts and located within the SDSS CMD filter shown in Figure 2(a). Only a subset of these spectra were used in our analysis of Hydra I (see Section 3.1). Heliocentric radial velocities (V_R) and $1\sigma$ uncertainties were derived using the method described in Section 2.2. The DECam photometry has been corrected for Galactic extinction (see Section 2.1) using the listed values of $E(B-V)$.

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

While our fits to the Hectochelle spectra return simultaneous estimates of stellar-atmospheric parameters ${T}_{{\rm{eff}}}$, $\mathrm{log}g$, and [Fe/H], we find that these quantities exhibit systematic as well as random errors that are larger than those found, using the same technique on spectra from the same instrument, by W15. The reason for this difference is that our current spectra were acquired in relatively bright conditions and our sky-subtracted spectra suffer significantly more residual contamination by scattered sunlight. Therefore, in this work we shall use only the velocities obtained from our Hectochelle spectra; for stellar-atmospheric parameters we will rely on the SDSS/SEGUE catalog.

We examined the data quality and measurement reliability using both repeat observations and the pipeline data quality measures described above. To ensure a reliable dataset, we adopted cuts on the radial velocity error (v_err < 2.1 km s⁻¹, or 3 times the median velocity error of 0.7 km s⁻¹) and removed spectra with large residual contamination from scattered sunlight. After applying these cuts, 411 stars remained in the Hectochelle sample. These data are presented in Table 2 and include the DECam magnitudes (corrected for Galactic extinction). The spatial positions of stars fainter than g = 18.5 are shown with respect to the smoothed spatial map of Hydra I in Figure 3 (left panel). The radial velocity error of the Hectochelle data as a function of magnitude is shown in Figure 4. We obtain velocity errors less than ∼2 km s⁻¹ to a limiting magnitude of g ∼ 20.5, approximately two magnitudes below the main sequence turnoff (MSTO) of Hydra I.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 3.  SDSS/SEGUE Spectroscopy

R.A.	Decl.	VR	${\sigma }_{{V}_{R}}$	[Fe/H]	${\sigma }_{[\mathrm{Fe}/{\rm{H}}]}$	[α/Fe]	${\sigma }_{[\alpha /\mathrm{Fe}]}$	g0	${\sigma }_{{g}_{0}}$	r0	${\sigma }_{{r}_{0}}$	$E(B-V)$
(degree)	(degree)	(km s−1)	(km s−1)	(dex)	(dex)	(dex)	(dex)	(mag)	(mag)	(mag)	(mag)	(mag)
132.572900	3.247481	54.42	0.70	−0.35	0.051	0.18	0.014	17.217	0.006	16.636	0.006	0.0348
132.673340	3.591954	9.72	0.80	−0.20	0.056	0.13	0.016	16.880	0.006	16.377	0.004	0.0445
132.678310	2.828570	269.38	8.09	−0.87	0.066	0.46	0.018	19.960	0.013	19.574	0.009	0.0304
132.684040	3.827853	28.81	1.49	−0.47	0.023	0.31	0.012	18.131	0.011	17.584	0.006	0.0384
132.685030	3.294454	44.80	0.97	−0.36	0.040	0.21	0.014	17.402	0.006	16.843	0.003	0.0343
132.691000	3.215672	214.44	2.48	−1.57	0.065	0.24	0.018	18.027	0.008	17.526	0.007	0.0340
132.703350	2.779862	26.07	1.24	−0.58	0.031	0.43	0.013	16.762	0.010	16.250	0.007	0.0307
132.709140	3.524834	296.70	1.97	−1.01	0.074	0.34	0.014	18.124	0.005	17.600	0.003	0.0420
132.734190	3.252893	53.05	0.74	−0.40	0.011	0.22	0.013	16.587	0.005	16.118	0.004	0.0339

Note. SDSS/SEGUE catalog of all 292 stars located in the Hydra I region and within the SDSS CMD filter shown in Figure 2(a). All spectroscopic parameters were taken from the SSPP pipeline. Only a subset of these spectra were used in our analysis of Hydra I (see Section 3.1). The DECam photometry has been corrected for Galactic extinction (see Section 2.1) using the listed values of $E(B-V)$.

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

We complemented our Hectochelle data with archival spectroscopy from SDSS, adopting the spectroscopic parameters derived from the SSPP (Lee et al. 2008a, 2011). Unlike the MMT/Hectochelle target selection, the original SDSS/SEGUE survey spectroscopic selections were designed to meet a number of scientific goals. For example, because the SEGUE-1 survey was designed to broadly study all the major structural components of the MW (i.e., the disk, halo, and halo substructure), approximately 50% of the spectroscopic targeting excluded stars in the color range of the MSTO of Hydra I (Yanny et al. 2009). We analyze the same SEGUE pointing as Schlaufman et al. (2009, 2011; B-8/PCI-9/PCII-21), but because the SDSS data are more heterogeneously selected, we applied the same isochrone cut in color and magnitude used to select the Hectochelle targets—i.e., selecting only stars that might reasonably be Hydra I members. These data are presented in Table 3 and include the DECam magnitudes (corrected for Galactic extinction). Although SEGUE spectroscopic stars are a biased sampling of stars within this color–magnitude selection, they should still yield a unbiased measurement of its kinematic properties and a lower limit on its spread in [Fe/H] and [α/Fe]. The right panel of Figure 3 shows the spatial distribution of the SDSS/SEGUE stars fainter than g = 18.5.

Table 4.  Properties of Hydra I

Parameter	Value
R.A.	1339
Decl.	36
Galactic Longitude	2247
Galactic Latitude	291
Distance	12.7 ± 0.3 kpc
L	∼1000 ± 200 L⊙
MV	∼−2.5 ± 0.3 mag
Agea	∼5–6 Gyr
Vheliob	89.4 ± 1.4 km s−1
VGSRb	−55.5 ± 1.4 km s−1
σVb	8.4 ± 1.5 km s−1
Rotation Amplitudeb	≲6 km s−1
$\langle {\rm{[Fe/H]}}\rangle$	−0.91 ± 0.03 dex
${\sigma }_{{\rm{[Fe/H]}}}$ c	0.13 ± 0.02 dex
$\langle [\alpha /\mathrm{Fe}]\rangle$ c	0.14 ± 0.02 dex
${\sigma }_{[\alpha /\mathrm{Fe}]}$ c	0.09 ± 0.03 dex

Notes.

^aYoungest stellar populations only (see Section 4.3). ^bDerived from the faint Hectochelle sample. ^cDerived from the SDSS sample.

Download table as:  ASCIITypeset image

The Hectochelle targeting employed a relatively wide, SDSS-based CMD filter, and given that we have clearly detected the MSTO of Hydra I, we can remove likely contaminants using their location on the higher precision DECam CMD. Stars with Hectochelle/MMT observations are shown in Figure 6 and are colored by their measured radial velocities. The SDSS/SEGUE sample is shown on the same CMD in Figure 7, where stars are colored by their velocities 7(a), iron abundances 7(b), and alpha abundances 7(c), respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The radial velocities of stars along the MS are quite similar (∼90 km s⁻¹; see Figures 6(a) and 7(a)), whereas thick disk stars at brighter magnitudes show a larger velocity spread and have higher iron abundances (−0.5 < [Fe/H] < 0). At fainter magnitudes (g > 18.5), the SDSS sample shows that both the iron abundance and velocity of MSTO stars are well correlated. In contrast, stars at slightly redder colors (g −r ≳ 0.24) show a wider spread in velocity and may be on average more metal-poor than MSTO stars.

We define a "faint sample" of Hectochelle stars using a narrower CMD filter shown by the blue dashed lines in Figure 6(b). This filter has a red edge which is 0.06 mag bluer than our SDSS selection filter, minimizing contamination from redder stars ($g-r$ ≳ 0.24). The filter also removes thick disk stars at brighter magnitudes (g < 18.5), slightly brighter than the MSTO. For our faint sample, we only include stars within the central 2.3 sq. degrees centered on Hydra I, resulting in a total of 122 stars. Figure 6(d) shows the velocity histogram of the Hectochelle faint sample. For comparison we show stars at brighter magnitudes (g < 18.5) which would have passed the narrow CMD filter. The contamination is readily apparent in the velocity histogram of bright stars, which only shows a small excess of stars at the measured systemic velocity of Hydra I (V_rad ∼ 89 km s⁻¹). By contrast, the faint sample shows that Hydra I is a significant overdensity of stars in velocity space, consistent with the results of Schlaufman et al. (2009).

The SDSS faint sample is selected without the use of a narrower, bluer CMD filter due to the larger color measurement uncertainties. Although this may slightly increase contamination, we use the SSPP [Fe/H] measurements as an additional parameter in our membership probability calculations (see Section 3.2). We keep only stars fainter than g = 18.5 and within the 2.3 sq. degree region shown in Figure 3 for a total of 42 stars. Like the Hectochelle sample, the velocity histogram of the SDSS faint sample (Figure 7(e)) shows an strongly peaked overdensity of stars in velocity space, confirming the original discovery of this halo overdensity by Schlaufman et al. (2009).

We determine membership probabilities using the Expectation Maximization (EM) method described in Walker et al. (2009a). The EM method uses all available data to describe both the member and contamination distributions. We assume a Gaussian velocity distribution, allowing the algorithm to return membership probabilities for each star along with the mean and variance of the member distribution. The spatial probability distribution is described as a monotonically decreasing function of the stars' distance from an assumed center between the two primary lobes (R.A. = 1339, decl. = 36; see Figure 3; Grillmair 2011). The EM algorithm (and the derived membership probabilities) implicitly assumes that there are no gradients in velocity, metallicity, velocity dispersion, or metallicity dispersion. We use a Besancon model (Robin et al. 2003) to describe the Galactic contamination in our EM algorithm. For the faint Hectochelle sample, we use only the spatial and velocity information to determine membership probabilities. For the SDSS faint sample we also use the iron abundance as an additional parameter in determining memberships.

The result of the EM algorithm applied to the Hectochelle faint and SDSS samples are shown in Figures 8 and 9, respectively. Each point is colored by its membership probability (P_M). The EM algorithm infers a systemic heliocentric velocity of $\bar{V}=89.4\pm 1.9$ km s⁻¹ and dispersion $\sigma =8.4\pm 1.5$ km s⁻¹ for the Hectochelle faint sample, and $\bar{{V}_{\mathrm{SDSS}}}=89.6\pm 5.3$ km s⁻¹ and ${\sigma }_{\mathrm{SDSS}}=9.1\pm 3.1$ km s⁻¹ for the SDSS faint sample. Errors on the mean and dispersion were determined via bootstrapping. Schlaufman et al. (2009) found a similar mean velocity of 85 km s⁻¹ from their analysis of the SEGUE pointing coincident with Hydra I and a higher velocity dispersion of 14.9 km s⁻¹.

Zoom In Zoom Out Reset image size

The spatial positions of the Hectochelle and SDSS samples, colored by membership probabilities, are shown on the smoothed surface density map of Hydra I in Figure 10. The high probability members are well correlated with the photometric overdensities, although contamination from likely non-members is clearly present. As also noted by Grillmair (2011), the SDSS data show a clump of high P_M stars to the southeast of Hydra I. Examining the larger EBS region in Figure 1 shows that these stars fall along the southeast extent of the EBS stream.

The presence or lack of rotation in the Hydra I system is an important clue to its nature and possible ongoing destruction. To measure its rotation, we use a method similar to the one described by Lane et al. (2009, 2010a, 2010b) and Bellazzini et al. (2012) for stars in the Hectochelle faint sample and SDSS samples with membership probabilities ${P}_{{\rm{M}}}\quad \gt$ 0.5. In this method, the sample is divided in two by a line that passes through the center of Hydra I, and a difference in the mean velocity of each subsample is taken. This velocity difference is calculated for dividing lines with position angles (PA) ranging from 0 < PA < 360°. The analysis of both samples show no evidence for rotation at a statistically significant level. We find rotation amplitudes of ∼3–4 km s⁻¹ with bootstrap-derived uncertainties of ∼2 km s⁻¹. We can therefore set a 3σ upper limit on the rotation of ∼6 km s⁻¹.

We use SDSS DR8 proper motions (USNO B; Munn et al. 2004, 2008) to measure the ensemble proper motion of Hydra I stars in the Hectochelle sample. The typical,per star,random measurement uncertainty of this catalog is ∼4 mas yr⁻¹ (Munn et al. 2004). We find average proper motions of $\overline{{\mu }_{\alpha }\mathrm{cos}\delta }=-0.04\pm 0.57$ mas yr⁻¹ ($2.4\pm 34$ km s⁻¹) and $\bar{{\mu }_{\delta }}=-0.04\pm 0.45$ mas yr⁻¹ ($2.4\pm 27$ km s⁻¹) from Gaussian fits weighted by the EM membership probabilities. Using the uncertainties in proper motions,we set a $3\sigma$ upper limit to the proper motion of μ ∼ 2.2 mas yr⁻¹ (132 km s⁻¹). Although the measured proper motions are consistent with zero,they are very different than the proper motion of Hydra I predicted by Grillmair (2011): $\overline{{\mu }_{\alpha }\mathrm{cos}\delta }=-0.15\pm 0.57$ mas yr⁻¹, $\bar{{\mu }_{\delta }}=-2.67\pm 0.45$ mas yr⁻¹ for a prograde orbit. The differences between the measured and predicted proper motion are likely due to differences in methodology and sample selection: Grillmair (2011) use positions and radial velocities only to predict the proper motion of Hydra I, and we adopt only a subset of the SDSS stars (those within r = 086 centered on Hydra I) used by Grillmair (2011). A combination of increased numbers of stars and higher precision astrometry—perhaps from the Gaia mission—would be necessary for a refined proper motion measurement.

The mean iron- and alpha-abundances, and possible spreads, are important diagnostics for distinguishing between a globular cluster or dwarf galaxy progenitor for the EBS (Willman & Strader 2012; Kirby et al. 2013). As noted in Section 2.2, we use the SDSS sample to measure the iron and alpha-element abundances of Hydra I stars, adopting the [Fe/H] and [α/Fe] values from SSPP and membership probabilities determined using the EM algorithm. In the SSPP, the stellar iron abundances are determined from a calibration of Ca ii H+K and Ca ii triplet observations, while the value of [α/Fe] is determined primarily from the abundances of Mg i, Ti i, and Ti ii. Detailed descriptions of the pipeline iron- and alpha-abundance determinations are given in Lee et al. (2008a, 2011), respectively.

Figure 11 shows the [Fe/H] histograms for the SDSS sample. As expected from the [Fe/H]-velocity distribution (Figure 9(b)), stars with ${P}_{{\rm{M}}}\gt 50\%$ show a strong peak in iron abundance space, while lower probability stars show a wider spread in [Fe/H]. The alpha-element [α/Fe] abundance measurements from SSPP (Lee et al. 2011) provide an additional diagnostic on the star-by-star membership, and possible origin, of Hydra I. Figure 12 shows [α/Fe] versus [Fe/H] for the SDSS sample. The [α/Fe] distribution of high probability member stars peaks near ∼0.1 dex, whereas the low membership probability stars show a broad, non-peaked distribution of [α/Fe] values.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We determine the mean and dispersions of the iron and alpha-element abundances using the full sample of SDSS stars, adopting the EM membership probabilities as weights. We find mean abundances (with standard errors) of $\langle {\rm{[Fe/H]}}\rangle \ =-0.91\pm 0.03$ dex and $\langle [\alpha /\mathrm{Fe}]\rangle =0.14\pm 0.02$. We measure statistically significant dispersions in both abundances: ${\sigma }_{{\rm{[Fe/H]}}}=0.13\pm 0.02$ dex and ${\sigma }_{[\alpha /\mathrm{Fe}]}=0.09\pm 0.02$ dex, respectively. We discuss the implications of these measurements in Sections 5.1 and 5.3.

How sensitive are the measured spreads in the abundances to the individual measurement uncertainties? To assess this, we recalculate the abundance spreads using larger measurement uncertainties of 0.1 dex when the reported random errors on individual measurements are <0.1 dex. Systematic comparisons to high resolution spectra and repeat SEGUE measurements have shown that the systematic errors on individual measurements are closer to ∼0.1 dex, whereas the random errors on high S/N spectra are typically ∼0.05 (Allende Prieto et al. 2008). With these conservative measurement uncertainties, we find ${\sigma }_{{\rm{[Fe/H]}}}=0.09\pm 0.03$ dex and ${\sigma }_{[\alpha /\mathrm{Fe}]}=0.08\pm 0.02$ dex. The use of larger uncertainties still yield statistically significant dispersions, and the values themselves remain relatively unchanged at a spread of ∼0.1 dex in the iron and alpha-elements. We note that adopting errors larger than ∼0.12 dex effectively erases evidence of an abundance spread.

Lastly, we note that the likely contaminants in our sample—those stars with ${P}_{{\rm{M}}}\lt 50\%$—show distributions in [α/Fe] and [Fe/H] consistent with the expected stellar populations. Given the spatial proximity to or association with the EBS stream, we expect that the ${P}_{{\rm{M}}}\lt 50\%$ abundance distributions will be broad—containing stars from the EBS, the Galactic halo, and Hydra I itself. The small peak in the iron-abundance distribution at [Fe/H] ∼ −1.7 dex (Figure 11) is consistent with Galactic halo contamination. These stars also show slightly higher values for the alpha-element abundances ([α/Fe] ∼0.25 dex), consistent with the expectation for halo stars (e.g., Pritzl et al. 2005; Hawkins et al. 2014).

Given the measurement of a mean iron abundance of [Fe/H] = −0.9 dex from the SDSS sample, we explore the age range of Hydra I's stellar populations using the observed CMD of spectroscopic members and theoretical isochrones. Although we used a color–magnitude cut to select spectroscopic targets (see Figures 2 and 6), we are sensitive to MSTO stellar populations in Hydra I as young as 4 Gyr (for [Fe/H] = −0.9 dex).

Figure 13 shows the CMD of the Hectochelle and SDSS samples compared to isochrones with ages 5–6 Gyr, [Fe/H] =−0.9 dex, and [α/Fe] = +0.2 dex (Dotter et al. 2008). We infer a distance of $d=12.7\pm 0.3\;{\rm{kpc}}$ based on the prominent location of the MSTO in the CMD. We find that fitting the blue MSTO population of Hydra I stars ($g-r$ ≲ 0.2 mag) requires a stellar population with an age of 5–6 Gyr, assuming a mean iron abundance of [Fe/H] = −0.9 dex. Changes of ±0.1–0.2 dex in the iron and alpha abundances result in older/younger ages by ∼0.5 Gyr, which is below the age uncertainty of ∼1 Gyr implied by the photometric uncertainties in the DECam $g-r$ colors at the MSTO. The age of the ∼5–6 Gyr stellar populations correspond to a formation at redshift z ≃ 1–1.2 for the standard ΛCDM cosmological model (Planck Collaboration et al. 2014). We discuss the implications of intermediate age stellar populations in Hydra I in Sections 5.1 and 5.3.

Zoom In Zoom Out Reset image size

The bluest stars in Hydra I show good agreement with the ∼5–6 Gyr isochrones and are clearly inconsistent with a single old stellar population (age ≳7 Gyr). The Hectochelle sample shows a number of high probability stars at colors redder than the blue MSTO ($g-r$ ≳ 0.2), consistent with the expected colors of old, metal-poor MSTO stars. However, a number of factors limit our ability to draw conclusions about the presence of older stellar populations, including small numbers of candidate stars in both the Hectochelle and SDSS sample, as well as the limited/incomplete spectroscopic targeting of the reddest stars. We explored the possibility of a old metal-poor MSTO using a background subtracted Hess diagram of the region but found no statistically significant evidence for such a feature. Larger spectroscopic samples targeting the expect colors of older stars are necessary to further investigate the presence of complex stellar populations in Hydra I. We discuss the implications of the stellar populations analysis in Section 5.1.

We estimate a total of ∼400 stars brighter than g = 22 in Hydra I from a comparison of star counts in the overdensity to the adjacent field (Figure 5). As noted in Section 2.1, we expect no issues with completeness at this magnitude limit. We use the number count estimate, in combination with a luminosity function (LF) from Dotter et al. (2008), to place limits on the total luminosity of Hydra I. The LF of Hydra I was created with a Salpeter IMF, an age of 5.5 Gyr, [Fe/H] = −0.9 dex, and [α/Fe] = 0.2 dex. We calculated the total luminosity of Hydra I by normalizing the LF with our value of 400 stars brighter than g = 22. The normalized LF was then integrated to find the total luminosity of Hydra I. We determine that the object has a total luminosity of $\sim 1000\pm 200\quad {L}_{\odot }$, or ${M}_{g}\simeq -2.4\pm 0.3$ and ${M}_{r}\simeq -2.6\pm 0.3$. Our uncertainties are determined assuming Poisson statistics describe the number of star counts. Using the SDSS-Johnson filter transformations of Jordi et al. (2006), we find ${M}_{V}\simeq -2.5\pm 0.3$ mag. Because our background estimate is taken from an adjacent region not too far from Hydra I, it is possible that we have overestimated the background star counts by not sampling the "true" background. In this case, our measurement would be an underestimate of the true luminosity of Hydra I.

The measured velocity dispersion of Hydra I ($\sigma =8.4\pm 1.5$), in combination with a half-light or other characteristic radius, can yield an estimate of the dynamical mass of the system (Walker et al. 2009b; Wolf et al. 2010). The spatial morphology of Hydra I indicates that the object is likely undergoing significant tidal disturbance, so any measure of the dynamical mass is limited by the extent to which the present day stellar motions trace Hydra I's underlying gravitational potential. We discuss this assumption further in Section 5.1.

Within the inner $r=0\buildrel{\circ}\over{.} 86$ (190 pc) region of interest, we find a velocity dispersion $\sigma =8.4$ km s⁻¹. If we assume this radius correspond to the half light radius of Hydra I, we can use Equation (2) from Wolf et al. (2010) to estimate the dynamical mass within the half light radius, ${M}_{1/2}$. We derive a dynamical mass estimate of ${M}_{1/2}=9\times {10}^{6}$M_⊙. If this mass estimate is within two orders of magnitude of the true dynamical mass, then Hydra I must have a significant dark matter component given the luminosity of ∼1000 ${L}_{\odot }$ derived in 4.4.

Given our stellar and dynamical mass estimates: Is the observation that Hydra I appears to be experiencing stellar mass loss consistent with the expected tidal radius of the object at its distance from the Galactic center? The tidal radius r_tidal of an object can be approximated using Equation (2) from Bellazzini (2004):

$$\begin{eqnarray}&&{r}_{\mathrm{tidal}}=\displaystyle \frac{2}{3}{\left[\displaystyle \frac{{m}_{{\rm{Hydra}}}}{2{M}_{{\rm{MW}}}({d}_{{\rm{Hydra}}})}\right]}^{1/3}{d}_{{\rm{Hydra}}},\end{eqnarray} \tag{ 4 }$$

where m_Hydra is the mass of Hydra I, d_Hydra is the galactocentric distance of Hydra I (∼19 kpc), and ${M}_{\mathrm{MW}}({d}_{{\rm{Hydra}}})$ is the MW mass within the distance to Hydra I. We calculate the MW mass ${M}_{\mathrm{MW}}({d}_{{\rm{Hydra}}})$ assuming an isothermal sphere with circular velocity of V_circ = 220 ± 40 km s⁻¹ (Bellazzini 2004).

We calculate the tidal radius (r_tidal) using both dynamical and stellar mass estimates. For a dynamical mass of 9 × 10⁶M_⊙, we find tidal radius of 17 (∼370 pc). However, assuming a stellar mass to light ratio ${M}_{*}/L=1$ and $L=1000\quad {L}_{\odot }$, we find a tidal radius of ${r}_{\mathrm{tidal}}\sim 0\buildrel{\circ}\over{.} 1$ (∼22 pc).

The instantaneous tidal radii of both the "only stars" and the high dynamical mass scenarios appear to be consistent with our observations of Hydra I. The small tidal radius, expected if Hydra I only contains stellar mass, is easily in good agreement with the spatial extent (∼1° radius) and asymmetric morphology of Hydra I. The larger tidal radius of ∼1°–2°, expected from the estimated dynamical mass, is consistent with mass loss at Hydra I's edges (possibly resulting in the EBS stream). It could also be consistent with significant tidal disturbance within Hydra I's inner ∼1°–2°, given that the instantaneous tidal radius may exceed its orbit averaged tidal radius and that we may be overestimating the dynamical mass.

If, however, the system is unbound, we can estimate the dissolution time for the system based on the crossing time—the timescale for an unbound star to reach the observed extent of the object. For simplicity, we assume a characteristic size of 1° and that stars are moving at their measured velocity dispersion (∼8 km s⁻¹). We find that the system would have a dissolution time of only ∼30 × 10⁶ years. Given such a short dissolution timescale, it would be surprising to observe this stellar overdensity at the present day if Hydra I were truly unbound. Although this provides general support for Hydra I as a bound object, additional kinematic data and numerical modeling would be necessary to draw more detailed conclusions.

The hypothesis that Hydra I is a disrupting dwarf galaxy has been considered previously by Schlaufman et al. (2011) in the discussion of their cold halo objects (ECHOs) sample. They argued that the relatively high velocity dispersions of the ECHOs sample ($\langle {\sigma }_{{\rm{vel}}}\rangle \sim 20$ km s⁻¹) are difficult to explain with globular cluster progenitors, which (on average) show velocity dispersions of ∼7 km s⁻¹ (Kimmig et al. 2015). However, see the discussion below (Section 5.2) for a counterpoint to this broad dynamical argument.

If Hydra I is a dwarf galaxy, then the mass–metallicity and luminosity–metallicity relations for Local Group dwarf galaxies provide a means to estimate the stellar mass and luminosity of the stream progenitor (Kirby et al. 2013). For a mean iron abundance of [Fe/H] ≃ −1 dex, a dwarf galaxy stream progenitor would have been relatively massive—approximately ${M}_{*}\sim {10}^{9}{M}_{\odot }$—similar to a present-day Fornax dSph. Comparing the SSPP [Fe/H] to the Kirby et al. iron abundances could introduce small (∼0.1–0.2 dex) systematic differences due to the use of different lines and methodologies. The Kirby et al. abundances are directly measured from the iron lines, whereas the SSPP [Fe/H] values are determined via calibration of calcium lines. The mass–metallicity relation therefore only provides a first-order estimate of Hydra I's progenitor stellar mass.

Given our stellar estimate of ${M}_{*}\sim 1000\quad {M}_{\odot }$ for Hydra I (see Sections 4.4), this implies that a dwarf galaxy stream progenitor would have lost ≳99.99% of its stellar mass. If Hydra I has undergone such substantial stellar mass loss, it is unlikely that our analysis in Section 4.5 yields the true, gravitationally bound mass of the remnant. Smith et al. (2013) find that the velocity dispersion of a dwarf galaxy remnant is only a reasonable measure of the bound mass when the object has retained ≳10% of the initial mass. When significant mass loss (≳98%) has taken place and only a small bound object remains—the remnant mass can be overestimated by as many as three orders of magnitude. We therefore do not consider the large dynamical "mass" of Hydra I to provide support for a dwarf galaxy hypothesis.

The iron abundance spread ${\sigma }_{{\rm{[Fe/H]}}}$ is a potentially strong discriminator between dwarf galaxies and GCs (Willman & Strader 2012). Comparing a sample of MW dwarfs and GCs, Willman & Strader (2012) find ${\sigma }_{{\rm{[Fe/H]}}}\sim 0.3{\rm{\mbox{--}}}0.7$ dex for dwarfs, while GCs less luminous than ${M}_{V}=-10$ have almost no measurable spread above the measurement uncertainties. Our estimate of ${\sigma }_{{\rm{[Fe/H]}}}=0.13\pm 0.02$ dex is statistically significant and consistent with a dwarf galaxy progenitor hypothesis. Although the brightest (i.e., classical; ${M}_{V}\lt -8$) dwarfs generally have larger iron abundance spreads (∼0.2–0.3 dex) compared to Hydra I, a relatively small value ${\sigma }_{{\rm{[Fe/H]}}}$ is consistent with the apparent decrease in ${\sigma }_{{\rm{[Fe/H]}}}$ with increasing galaxy luminosity noted by Willman & Strader (2012).

All MW dwarf galaxies with comprehensive studies of their stellar populations have shown evidence for an old, metal-poor stellar population (e.g., Kirby et al. 2013; Brown et al. 2014; Weisz et al. 2014). The addition of this population to the observed metallicity distribution would significantly increase its dispersion in [Fe/H], making the value more consistent with other luminous dwarf galaxies. If Hydra I is indeed a dwarf galaxy, our non-detection of this population could be partially due to our selection function (see Section 4.3). Another possibility is that Hydra I might have once possessed such a population, but due to its larger spatial extent (e.g., Battaglia et al. 2006; Bate et al. 2015), it has preferentially been tidally stripped compared to the more metal-rich subpopulation. We emphasize that our analysis does not rule out the possibility that Hydra I contains an old, metal-poor population. Additional spectroscopic targeting of candidate stars in the expected region of color–magnitude space and/or additional wide-field imaging would be necessary to find evidence for an older population.

The presence of intermediate age stellar population in Hydra I (ages ∼5–6 Gyr; Section 4.3) is consistent with a dwarf galaxy progenitor hypothesis. Although MW dwarf spheroidals, on average, formed the majority of their stars prior to $z\sim 2$ (∼10 Gyr ago), there is significant variation from galaxy to galaxy. Moreover, more massive dwarfs form a larger fraction of their stellar mass at later times (Weisz et al. 2014). For example, ∼50% of the stellar mass in Fornax formed in the last ∼8 Gyr and ∼40% was formed in the last ∼6 Gyr (Weisz et al. 2014). If Hydra I is a disrupting dwarf galaxy and was relatively massive at infall (as implied by the measured [Fe/H]), then the presence of an intermediate age stellar population is consistent with our expectations from the SFHs of Local Group dwarfs.

The presence of tidal streams associated with MW GCs (e.g., Pal 5; Odenkirchen et al. 2001) raises the possibility that Hydra I could be a GC in the final throes of destruction. Grillmair (2011) discussed this scenario and concluded that the relatively high velocity dispersion of ∼15 km s⁻¹ (determined by Schlaufman et al. 2009) did not rule out the possibility of Hydra I being a GC. They note that, because EBS/Hydra I may be on a highly eccentric orbit that brings it into the inner Galactic potential, a large velocity dispersion could be evidence of dynamical heating from frequent encounters with massive structures (such as giant molecular clouds) or dark matter subhalos. In this study, however, we find a significantly smaller velocity dispersion of σ ≃ 8–9 km s⁻¹ (see Section 3.2), more consistent with GCs than dwarf galaxies.

To further explore the GC progenitor hypothesis, we consider the location of Hydra I in the age–metallicity relation (AMR) for MW globular and open clusters in Figure 14. The AMR of GCs has been well studied both in the MW (Forbes & Bridges 2010; Dotter et al. 2011) and M31 (Caldwell et al. 2011), as well as the Large Magellanic Cloud (LMC) and Small Magellanic Cloud (SMC) (Pagel & Tautvaisiene 1998). For the MW, GCs within 8 kpc of the Galactic center show a narrow range of old ages (∼13 Gyr) but span a wide metallicity range. Outside a galactocentric distance of 8 kpc, however, GCs exhibit an AMR more like that of the LMC and SMC, with a trend showing younger ages at higher metallicities.

Zoom In Zoom Out Reset image size

If Hydra I is a disrupting GC, then we expect it to contain (to first order) a single stellar population with the age of ∼5–6 Gyr as inferred from our stellar populations analysis (Section 4.3). The MW has few known, well studied GCs in the intermediate age (∼6–8 Gyr) range. These include Pal 1 (Rosenberg et al. 1998a, 1998b), Pal 12, Terzan 7 (Dotter et al. 2010), Whiting 1 (Valcheva et al. 2015), and 47 Tuc (Hansen et al. 2013). With the exception of Whiting 1, all of these GCs have galactocentric distances ≲18 kpc (Harris 1996), similar to that of Hydra I (d_MW = 19 kpc). The youngest GCs of this subset (Pal 1 and Whiting 1), however, have higher iron abundances than Hydra I.

Outside of the MW, some evidence exists for massive star clusters with intermediate ages and metallicities. However, there are significant measurement uncertainties associated with these observations. For example, M31 contains a large population of massive [Fe/H] ∼ −1 GCs, but age determinations from integrated light measurements (e.g., Lick indices) at this metallicity are highly sensitive to the presence of blue horizontal branch stars, resulting in artificially young and thus highly uncertain ages (see discussion in Caldwell et al. 2011). The LMC has a well known "age-gap" in its star cluster population (e.g., Harris & Zaritsky 2009): there are no massive clusters with ages from ∼5 to 12 Gyr, and only one relatively low mass, intermediate metallicity cluster ($M\sim {10}^{4}\quad {M}_{\odot };$ [Fe/H] ∼ −1) with an age ∼8–9 Gyr (Mackey et al. 2006). In contrast, the SMC may contain a few $M\sim {10}^{4}{\rm{\mbox{--}}}{10}^{5}\quad {M}_{\odot }$ clusters consistent with the age and metallicity of Hydra I (see Dias et al. 2010 for a recent review), although the wide range of age estimates for individual clusters highlights the large uncertainty in their observed ages.

Given Hydra I's outlying age and metallicity, relative to other known MW clusters, and the lack of strong evidence for analog clusters in M31 and the Magellanic Clouds, we conclude that the star cluster hypothesis is less likely than the dwarf galaxy hypothesis. If Hydra I is indeed a GC, the observed age and metallicity imply that it would have likely formed in dwarf galaxy (perhaps smaller than the LMC or SMC) which was subsequently accreted by the MW. However, the lack of known GCs with ages and metallicities consistent with Hydra I places limits on the plausibility of such a scenario.

Lastly, we consider the possibility that Hydra I and EBS may be part of a larger Galactic structure—in particular the Monoceros Ring stellar overdensity. The distance and metallicity of Hydra I are consistent with photometric and spectroscopic studies of the larger Monoceros Ring region (Conn et al. 2012; Meisner et al. 2012). Slater et al. (2014) recently presented a Pan-STARRS1 stellar density map for the Monoceros Ring, constructed using MSTO stars from a ∼9 Gyr population with [Fe/H] = −1 (colors in the range $0.2\lt {(g-r)}_{0}\lt 0.3$). They argue that the EBS and Anticenter streams are extended stellar arcs that are physically associated with the Monoceros Ring. Conversely, Grillmair (2011) argued that the EBS stream is kinematically distinct from the Monoceros Ring, given the high eccentricity orbital solution needed to explain EBS's observed spatial curvature. The Monoceros Ring is believed to be on a nearly circular orbit (Crane et al. 2003; Martin et al. 2006). Although our study cannot shed additional light on whether either Hydra I or EBS might be associated with the Monoceros Ring, a more refined orbital solution for the EBS stream—using GAIA proper motions and a large sample of radial velocity measurements—is a promising avenue for investigating this hypothesis.

Given the proximity of Hydra I to the Monoceros Ring, it is possible we have isolated a younger population of stars recently speculated to be part of this structure. Carballo-Bello et al. (2015) noted that along the line of sight to the Monoceros Ring (at $l\sim 180^\circ ,195^\circ ;b\sim 25^\circ$), there exists a population of stars bluer than the ∼9 Gyr MSTO of Monoceros Ring stars. They suggested that this could be a ∼4 Gyr population (with an iron abundance of [Fe/H] ∼ −0.9 dex) associated with the Monoceros Ring. Carballo-Bello et al. (2015) were unable to clearly detect a MSTO feature in the SDSS data but estimated a distance of ∼10 kpc for the population. The systemic velocities of the blue population—as well as Hydra I—are consistent with model predictions from Peñarrubia et al. (2005) for the origin of the Monoceros Ring as an accreted dwarf galaxy. Although suggestive, additional precision photometry over a much larger field will be necessary to further explore any spatial connections between the Carballo-Bello et al. (2015) detections and the EBS/Hydra I.

Another consideration is whether the iron- and alpha-element abundances of Hydra I are consistent with the MW's halo or thick disk populations. Although this comparison may provide a general picture of the origin of Hydra I, the conclusions should be taken with caution. First, the distribution in [Fe/H] of the Galactic halo population is quite extended and shows a large spread in alpha-element abundances at a given [Fe/H] (∼0.15 dex; Hawkins et al. 2014). Second, the metal-rich tail of halo stars and metal-poor thick disk stars overlap in both [Fe/H] and [α/Fe] (Venn et al. 2004; Pritzl et al. 2005), blurring the distinction between these two Galactic populations in this metallicity regime. Lastly, comparisons of the SSPP abundances to those from heterogeneous compilations may introduce small systematic differences of order ∼0.1 dex or less (Venn et al. 2004).

A comparison of iron- and alpha-element abundances was done by Schlaufman et al. (2011) in their analysis of the SDSS/SEGUE observations of a set of cold halo substructures (ECHOs), one of which was Hydra I. In Figure 14 of Schlaufman et al. (2011), they compare the mean iron and alpha abundances of ECHOs to samples of thin disk, think disk, and halo stars compiled by Venn et al. (2004). The mean [Fe/H] and [α/Fe] of the ECHO associated with Hydra I is more iron-rich and less alpha-enhanced than both (1) the smooth halo along that line of sight ([Fe/H]_halo ∼ −1.4; [α/Fe]_halo ∼ 0.25]) and (2) the mean values of the halo as compiled by Venn et al. (2004). Compared to the thick disk stars compiled by Venn et al. (2004), Hydra I is less alpha-enhanced (by ∼0.15 dex) than the metal-poor thick disk at [Fe/H] ∼ −1. As additional evidence pointing to Hydra I's extragalactic origin, Schlaufman et al. (2011) note that the [Fe/H] and [α/Fe] distributions of the ECHOs are in good agreement with the MW dwarf galaxy satellites Leo I and Sculptor (see their Figure 15). Taken together, the mean values of the iron- and alpha-elements suggest an extragalactic origin for Hydra I rather than a Galactic overdensity of halo or thick disk stars.