EXPLORING HALO SUBSTRUCTURE WITH GIANT STARS: SUBSTRUCTURE IN THE LOCAL HALO AS SEEN IN THE GRID GIANT STAR SURVEY INCLUDING EXTENDED TIDAL DEBRIS FROM ωCENTAURI

If the Galactic halo has been formed partly (Searle & Zinn 1978) or entirely (e.g., Majewski 1993) from accretion of smaller systems then one might expect groups of stars with halo-like motions having strong velocity coherence (e.g., Helmi & White 1999; Meza et al. 2005) near the Sun. Claims for nearby halo moving groups date back at least to Eggen & Sandage (1959) and include several retrograde candidates—including, in particular, the widely recognized Kapteyn Group—among the halo groups long discussed by Eggen (e.g., 1965, 1996a, 1996b). Majewski et al. (1994, 1996) analyzed nearby halo stars (selected by asymmetric drift) toward the north Galactic pole and claimed strong organization of their phase space distribution into a few clumps, including a prominent retrograde moving group initially identified via proper motions in Majewski (1992). A possible association of horizontal branch stars to this retrograde feature has been identified by Kinman et al. (2007). Eggen (1996b) suggests an association of the Majewski (1992) retrograde group to Kapteyn's group.

More recently, interest in the notion that the globular cluster ωCentauri ("ωCen") may be the remnant core of a tidally disrupted satellite galaxy (e.g., Lee et al. 1999; Majewski et al. 2000b; Bekki & Freeman 2003)—a notion inspired by (1) internal chemical and age distributions belying multiple ωCen stellar populations; (2) the example of the similarly massive cluster M54 located in/near/as the "core" of the disrupting Sagittarius galaxy (Majewski et al. 2000b) and apparent chemical similarities of M54+Sagittarius to ωCen (Carretta et al. 2010); and (3) the unusual, low-inclination, retrograde orbit of ωCen itself (Dinescu 2002)—has led to several ωCen tidal disruption simulations; these models generally produce retrograde-moving, ωCen debris relatively near the solar circle (Dinescu 2002; Tsuchiya et al. 2003, 2004; Mizutani et al. 2003; Bekki & Freeman 2003). This has prompted searches for retrograde halo stars possibly shed by the "cluster" and led to suggestions that an "ωCen" signal is present among local metal-weak stars (Dinescu 2002; Mizutani et al. 2003; Meza et al. 2005). In fact, Eggen (1978) speculated a connection between his Kapteyn's star group and ωCen (see also Kotoneva et al. 2005), a dubious connection (Proust & Foy 1988) before the breadth of ωCen's metallicity distribution function was fully recognized, but more recently bolstered by detailed chemical analysis of Kapteyn group stars (Wylie-de Boer et al. 2010).

Despite these recent claims for stripped ωCen stars in the solar neighborhood, searches for extratidal stars near ωCen itself have been less promising. While the photometric search by Leon et al. (2000) seemed to suggest a "significant" pair of tidal tails extending from ωCen, these results were cast in doubt when the substantial foreground differential reddening was assessed (Law et al. 2003).⁵ An expansive spectroscopic search for ωCen stars beyond its tidal radius by Da Costa & Coleman (2008) reveals only six candidates among more than 4000 stars selected from the ωCen giant branch in the color–magnitude diagram; these authors suggest that this meager haul is consistent with models where most stripping took place long ago, and with the lost stars now widely distributed about the Galaxy. Stronger support for the tidally disrupted dwarf galaxy model, and for confidently linking solar neighborhood candidate members with ωCen itself, would come from actually being able to trace debris along the satellite's orbit, and, eventually, from ωCen continuously to the solar neighborhood.

Here, we report detection of a kinematically coherent "tidal debris" signature spanning ≳ 60° of Galactic longitude in a large radial velocity (RV) survey of giant stars mostly (∼95%) within ∼5 kpc of the Sun. Stars within this dynamically coherent group show a specific chemical marker thought to be unique to ωCen, as well as distances and velocities consistent with models of ωCen tidal debris. Though still mostly only a few kiloparsecs away, these extended tidal debris stars provide a start at tracing the ωCen stream and provide crucial, though still crude, dynamical constraints on the disruption of the closest known dwarf galaxy to the Sun. At minimum, these discovered debris stars suggest that ωCen may be a principal source of local retrograde stars.

Our analysis uses the Grid Giant Star Survey (GGSS), a partially filled, all-sky search for giant stars using the Washington M, T₂+DDO51 photometric selection technique described in Majewski et al. (2000a). The GGSS had as one goal the identification of bright but distant, metal-poor giants suitable for the Astrometric Grid of the (now defunct) Space Interferometry Mission. To this end, a specific subset of potential SIM Grid stars was drawn from the GGSS: the four most distant giant stars with M < 13.5 in each of the 1302 evenly spaced, ∼0.4–0.6 deg² GGSS fields, where photometric distances were estimated by assigning absolute magnitudes to stars from their position in the (M − T₂, M − DDO51) diagram (which separates dwarf and giant stars and, within these luminosity classes, sorts by metallicity; Majewski et al. 2000a). This particular selection biases this "SIM Grid sample" to more metal poor giants, because these are farther at a given apparent magnitude. An echelle resolution study of 774 Grid candidates selected in this way verified 100% of them to have giant branch surface gravities, and indicated a median [Fe/H] ∼−0.7 and distance of ∼2 kpc (Bizyaev et al. 2006), but with long tails to more distant and metal-poor stars. Thus, the sample contains a mix of old thin disk, Intermediate Population II (IPII) thick disk, and a smaller fraction of halo stars in mean proportions varying by Galactic latitude. Our initial analysis here relies on R ∼ 2600 spectroscopy of a larger sample of 3318 GGSS "SIM Grid" giants from our southern observing campaign (sky distribution shown in Figure 1(a)); further discussion of the GGSS is given in Patterson et al. (2001) and S. Majewski (in preparation).

Zoom In Zoom Out Reset image size

In 1999, a spectroscopic follow-up campaign for these candidate Astrometric Grid stars began at Las Campanas Observatory using the Modular Spectrograph on the Swope 1 m telescope (and the DuPont 2.5 m for 68 stars). The typical setup used the 600 line mm⁻¹ grating to sample a ∼2000 Å spectral region including both Hα and Hβ at 1.0 Å pixel⁻¹ (in a small fraction of cases the Ca infrared triplet region was observed; details of both setups are in Majewski et al. 2004). The signal-to-noise ratio of the spectra exceeds 10 in almost all cases and reaches >50 in some cases. RV cross-correlation of the spectra used the methodology discussed in Majewski et al. (2004). Typical RV errors are 5–10 km s⁻¹, estimated a variety of ways including repeat measures of many stars and comparison to echelle resolution RVs.

Figure 1(b) shows the distribution of Galactic Standard of Rest (GSR) RVs (v_GSR) for the "SIM Grid" stars, assuming a solar motion in right-handed Galactic coordinates of (+ 10.0, +225.3, +7.2) km s⁻¹. For stars moving predominantly in Galactic planar orbits (e.g., disk stars and putative ωCen debris), the approximate "planar RV" (i.e., the observed RV the star would have were it on the Galactic equator) is given by v_b = v_GSR/cos (b) (this latitude normalization breaks down at high latitude, where a star's Z motion dominates the observed RV, so Figure 1(c) is limited to stars with |b| < 60°). In this planar projection stellar populations more clearly sort by their relative asymmetric drifts, and Figure 1(c) for the most part shows the expected longitudinal distribution of v_GSR for a predominantly thin disk/IPII mix of stars.

However, stars not following disk kinematics are evident, including a number having retrograde velocities. Among stars with halo-like velocities (and particularly among likely retrograde stars), outlier groupings (e.g., at [l, v_b] ∼ [280°, 300 km s⁻¹]) or thin, coherent strands of stars (e.g., from [20°, −125 km s⁻¹] to [300°, −275 km s⁻¹]) can be seen in Figures 1(b) and/or (c) (where we have highlighted some interesting features with shading). Such cold and coherent RV trends with sky position are characteristic of long tidal streams, such as the Sagittarius system (e.g., Majewski et al. 2004; Law et al. 2005); in this case, however, the substructure is found among relatively nearby giants and therefore corresponds to stars with a much broader sky distribution than typical for more distant streams (Figure 1(a)). Such substructure among the GGSS giants is not surprising given that they probe distances similar to the mostly main-sequence stars in the Majewski et al. (1994, 1996) study, which also showed significant halo substructure (but in only a single pencil beam). Moreover, a new "all-sky" study of bright M giants by A. Sheffield (in preparation) and probing comparable distances show analogous, though even more striking, [l, v_b] coherences among halo-like stars (most likely because M giants probe typically younger tidal streams). Together, the GGSS giants, Sheffield M giants, Majewski et al. (1994, 1996) subdwarfs, and Kinman et al. horizontal branch stars point to the high degree of substructure in the halo even at the solar circle. Indeed, the degree of velocity coherence and substructure of the local halo does not differ much from that seen in the distant halo (Majewski 2004).

Are there stars in the GGSS sample that can be associated with ωCen? The studies of solar neighborhood ωCen debris mentioned in Section 1 were able to make use of the expected (E, L_z) distributions to trawl for the best local representatives. Unfortunately, our giant stars are far enough away that most available proper motions (and therefore complete space velocity determinations) are unreliable. Therefore, we winnow our search to those [l, v_b] ranges expected to be populated by any realistic model of ωCen tidal disruption. To do so we create a suite of N-body simulations of satellites undergoing tidal disruption along ωCen-like orbits in the static Milky Way (MW) potential given by Johnston et al. (1995), with 30,000 particles representing the parent satellite in an initially Plummer configuration. To account for observational uncertainties that prohibit us from accurately knowing the true position, space motion, and therefore orbits of both ωCen and the Sun, we create a grid of models spanning ranges of uncertainty around typical mean values for each critical interaction parameter (from Dinescu et al. 1999): ωCen distances of [4.9, 5.1, 5.3] kpc, Galactic Cartesian velocities (right-handed system in the Galactic rest frame) of V_x = [42, 53, 64, 75, 86] km s⁻¹, V_y = [ − 52, −43, −34, −25, −16] km s⁻¹, and V_z = [ − 6, 4, 14] km s⁻¹, a solar Galactocentric distance of [7.0, 7.75, 8.5] kpc, and, to set the MW mass scale, a local circular velocity of [220, 254] km s⁻¹, where the latter value is that suggested by Reid et al. (2009). Within this grid we adopt satellites of three different initial masses, [3e7, 3e8, 3e9] M_☉, and evolved for ∼0.4 Gyr, ensuring that each model orbit places the satellite at the current (l,b) position and RV of ωCen.

These 4050 simulations produce a variety of tidal streams from which we can establish the range of possible [l, v_b] distributions; in fact, the sum of these models (Figure 1(d)) gives us something like a "probability distribution function" (PDF) of where the last 0.4 Gyr of ωCen tidal debris might most likely lie (Figure 1(d)). Comparing this to the GGSS distribution, and ignoring regions dominated by disk stars, we call attention to a particularly striking match of the PDF to a relatively tight sequence of likely retrograde stars over l ∼ 280°–360°, highlighted with green shading in Figure 1(c). That these stars form a coherent velocity structure orbiting in a near-Galactic planar orbit is demonstrated by the fact that, despite their broad sky distribution (green and red points, Figure 1(a)), they show a coherent, string-like configuration in planar, v_b projection.

A second-order polynomial fit to the Figure 1(c) sequence with iterative rejection settles on 35 stars in the feature with an observed RV dispersion of only ∼40 km s⁻¹. Most interestingly, the sequence passes through the (l, v_b) position of ωCen, also shown.

To test whether ωCen stars might fall among those GGSS stars lying within this retrograde feature, high-resolution (R ∼ 55, 000), echelle spectra for eight of them, as well as a control sample of four other halo GGSS stars (including two very extreme v_b stars at similar longitudes), were obtained using the Sandiford echelle spectrometer (McCarthy et al. 1993) on the McDonald 2.1 m Struve telescope. Given the modest wavelength coverage and S/N ∼ 25–50 of the spectra, iron abundances were derived from a set of unblended Fe i lines using measured equivalent widths. The stellar parameters T_eff and log g were taken from the analysis of GGSS stars by Bizyaev et al. (2006), with determinations of the microturbulence velocities (ξ) set by the Fe i lines measured for this study.

The well-defined Ba ii line at 5854 Å was used as an s-process abundance indicator because (1) it is the most easily detectable in these generally mediocre spectra, and (2) the distribution of [Ba/Fe]–[Fe/H] for ωCen is quite distinctive (Figure 2(a))—indeed it is unique among all star systems studied to date in the extreme overabundances of s-process elements characterizing its more metal-rich stellar population (see Figure 11 of the Geisler et al. 2007 review). Figure 2(b) shows the derived [Ba/Fe]–[Fe/H] pattern for the 12 GGSS stars; abundances and other data derived for these stars are given in Table 1. The quoted velocities are from the original medium-resolution spectra and have typical random errors of ∼10 km s⁻¹; photometric uncertainties are about 0.01 mag. Abundance uncertainties (shown in Figure 2(b)) were set by the sensitivities of Fe i and Ba ii abundances to changes in stellar parameters of ±100 K in T_eff, ±0.3 dex in log g, and ±0.5 km s⁻¹ in ξ. As is vividly demonstrated, the Table 1 stars most likely to be kinematically associated with ωCen (Figure 1(d)) clearly follow the characteristic ωCen trend in [Ba/Fe] versus [Fe/H], while two stars least likely to be kinematically associated with ωCen lie along the MW trend. The combination of kinematical consistency and possession of the hallmark barium abundance trends for the former group of stars are strong evidence that an extended part of the ωCen tidal debris stream has been found.⁶

Zoom In Zoom Out Reset image size

It is worth noting that within this rather small sample of ωCen stream candidates is a carbon-rich star, G1358-16.167. This red giant exhibits strong Swan C₂ bands (e.g., at λ5165 Å; Figure 3) and is strongly barium-enhanced. Such C-rich halo giants at this modestly low metallicity are relatively rare and constitute only about 1%–2% of halo giants. However, ωCen has at least five known C-rich giants (see Table 3 in Bartkevicius 1996, which includes both subgiants and giants) whereas the only other globular clusters known to harbor C-rich giants are M22 (with two—as well as a spread in heavy-element abundances and s-process enrichment, similar to, but not to the degree of, ωCen), M2 (with one), and M55 (with one—Smith & Norris 1982). Given the likely association of most Table 1 stars to ωCen it is not too surprising that one is found to be C-rich; this observation only strengthens the tie to their chemically peculiar parent stellar system.

Zoom In Zoom Out Reset image size

Based on the [Fe/H]–[α/Fe] and age–metallicity distributions for ωCen given by Stanford et al. (2006)—adopting a 4 Gyr ωCen age span—and the derived [Fe/H] and Two Micron All Sky Survey (2MASS) photometric data for the ten good GGSS ωCen candidates we estimate their distances using matching Dotter et al. (2008) isochrones. With these distances we can place the stars in their Galactic planar positions relative to ωCen and the Sun (Figure 4). Superposed on this distribution we plot the model debris and satellite orbit from one of several ωCen models in our grid (the example model parameters are given in the figure legend) that provide a reasonable match to the positions and RVs of the GGSS stars. This model, based on a satellite orbit with peri-/apo-Galactica limits of (1 kpc)/(7 kpc), respectively, not only demonstrates how the stars of interest very plausibly trace ωCen debris, but also how it might be possible for ωCen tidal debris to reach the solar neighborhood, as suggested by the various claims for this discussed in Section 1.

Zoom In Zoom Out Reset image size

In fact, though, as stated earlier, available proper motions for the GGSS ωCen stars are typically of low quality, the UCAC astrometry (Zacharias et al. 2010) does hint at further tantalizing connections to previous claims for nearby ωCen debris: the GGSS ωCen stars with the smallest derived (E, L_z) uncertainties happen to fall in, or quite near, the "ωCen debris expectation" box defined in (E, L_z) by Dinescu (2002, using the same gravitational potential), whereas the weighted mean L_z of all ten GGSS ωCen stars, −179 ± 135 km s⁻¹ kpc, matches extremely well the "ωCen peak" identified within the stellar sample explored by Meza et al. (2005, their Figure 9).

In the GGSS sample of bright giant stars within ∼5 kpc across the southern two-thirds of the celestial sphere and biased toward metal-poor stars we have shown evidence that those having halo-like velocities, and particularly those with retrograde velocities, show a highly substructured, rather than random, distribution. This result is a further demonstration that even in the inner halo, at the position of the solar circle, the Galactic halo is not well mixed, but shows the signature of multiple minor accretion events. We identify one group of stars kinematically and chemically consistent with being ωCen debris, which we use to get a rough constraint on the ωCen tidal stream in the inner Galaxy, a model demonstrating how ωCen stars can reach the solar neighborhood. We have probed with high-resolution spectroscopy less than half of the several dozen stars we associate with ωCen [l, v_b] features, yet all of these best candidates are found to be chemically consistent with being from ωCen; we can project that most likely a large fraction of the fuller sample of "ωCen" candidates are authentic. Considering that only a few dozen clearly retrograde stars are found in the southern GGSS at all (Figure 1(b) and (c)), our results suggest that ωCen tidal debris is a primary contributor of retrograde stars near the Sun, and likely the overwhelmingly dominant contributor in the inner two Galactic quadrants.

We gratefully acknowledge support by NASA/JPL contracts 1201670 and 1228235, and the David and Lucile Packard Foundation to build the GGSS, as well as NSF grants AST-0307851 and AST-0807945. This work would not have been possible without the generous access to the Swope telescope for development of the SIM Astrometric Grid granted by Carnegie Observatories Directors Augustus Oemler and Wendy Freedman.