EVIDENCE FOR A MILKY WAY TIDAL STREAM REACHING BEYOND 100 kpc

As noted in DR13, the MLS data have the same photometric sensitivity as the CSS data since all observations are taken unfiltered with the same type of CCD camera. All images are processed with the same software. From DR13, the color transformation from Catalina photometric system V_CSS magnitudes to Johnson V is given by

$$\begin{equation} V = V_{{\rm CSS}} + 0.31 \times (B-V)^2 + 0.04. \end{equation} \tag{ 1 }$$

The average B − V color of RRL is about 0.3 mag with stars varying between about 0.1 and 0.5 as they pulsate (e.g., Clube et al. 1969; Stepien 1972; Cacciari et al. 1987; Layden 1997; Nemec 2004). As the MLS data have no color information, we adopt the average values in our analysis. This gives rise to a maximum uncertainty of ∼0.07 mag in the V magnitudes. Combining this with the photometric uncertainty, we expect a color-based dispersion of σ = 0.09 mag in our RRL photometry.

To find RRL candidates we follow the analysis undertaken in DR13. That is, we first selected variable sources using a Welch–Stetson variability index I_WS (Welch & Stetson 1993). A total of 3.1 million MLS sources were selected as variable candidates using I_WS > 0.6. All sources were processed for periodicity using the Lomb–Scargle (LS; Lomb 1976; Scargle 1982) periodograms. In total 170 thousand sources were found to exhibit significant periodicity with a false alarm probability p₀ < 1 × 10⁻⁵. However, the bulk of these detections are due to the systematic sampling of the data and occur at integer day frequencies.

Here, as with DR13, we are primarily interested in RRab's since they have well-known magnitudes and characteristic light curves that are not easily mistaken for other periodic variables. After excluding the period aliases near 0.5 and 1 day, 3087 of the objects were found to have periods between 0.34 and 1.4 days. This period region was deliberately chosen to be broader than the RRab range in order to include sources found by the LS technique at aliases of their true periods.

As with the CSS sources of DR13, each of the RRab candidates was run through the AFD software (G. Torrealba et al. 2013, in preparation) to select the best period from among the best 15 given by the LS and AoV software (Schwarzenberg-Czerny 1989). Briefly, this process involves Fourier fitting and iteratively rejecting bad data to determine the best period based on reduced χ² values. Non-RRab sources were rejected using the M-test statistic (Kinemuchi et al. 2006, Equation (8)), as well as the Fourier fit order. This initial selection resulted in the detection of 1125 RRab candidates.

Upon reviewing the light curves of the RRab candidates we discovered a number of faint RRab's in the MLS data with V ∼ 20.5. Such RRL were found to be mainly concentrated in the constellation Gemini. In Figure 1, we plot the phased light curves of eight of these faint stars. These distant sources are very important for defining distant streams within the Galactic halo. From simulations based on adding artificial RRab light curves to our data, we discovered that our detection sensitivity was <50% for RRab's with V > 20.

Zoom In Zoom Out Reset image size

To improve our completeness we investigated the light curves of MLS sources with SDSS spectroscopic types A0 and F5 that had 19 < V < 21. In addition, we used SDSS photometry to select stars with the same color range as the RRab stars that we had discovered. These stars were matched with MLS sources at lower variability and periodicity thresholds (I_WS > 0.3 and p₀ < 0.01) to select RRab candidates. To better constrain the variability, we combined the MLS photometry with shallower CSS data. Likewise, we searched the CSS data for additional faint RRab candidates by selecting objects with RRab colors in SDSS photometry. Furthermore, since our analysis in DR13 showed that ∼17% of RRab's were missed because the selection was based on LS periods, we reprocessed the CSS and MLS photometry using the AoV period-finding software.

We inspected the light curve of each RRL candidate to assess variability and then carried out period finding for all the new variable candidates. In total, approximately 6000 additional sources were searched. Each of the new sources exhibiting some sign of periodicity were searched for improved periods using our Fourier-fitting AFD process. In a number of cases the RRL were found to exhibit Blazhko phase and amplitude variations (Blazhko 1907).

In total the additional searches yielded an extra 219 MLS RRL and 2051 CSS RRab's. Of these, only 425 were previously known. As distant RRL are important for defining Halo streams we retain the 17 clear c-type RRL (RRc) that were discovered in MLS data beyond 70 kpc. A number of these have SDSS spectra. After removing non-RRL sources the final set of MLS RRL consists of 1207 stars and is presented in Table 1. The new RRab's found in CSS data are given in Table 2.

In Figure 2, we present the magnitude distribution of the MLS and the CSS RRL. The MLS histogram shows clear bumps near V = 19.5 and V = 20.5. The first bump is easy to understand as it is clearly due to the Sagittarius leading and trailing arms (as seen in DR13). The second peak is due to the more distant RRL that are concentrated near the Galactic anti-center. We hereafter call this the Gemini stream. In Section 5, we will investigate the origin of this feature.

Zoom In Zoom Out Reset image size

The SEKBO survey discovered 2016 candidate RR Lyrae in a survey covering 1675 deg² along the ecliptic (Keller et al. 2008). Like the MLS survey, the SEKBO survey covered a band within 10° of the ecliptic. However, SEKBO RRL candidate selection was based on the two colors observed simultaneously by the MACHO camera and variability was determined from between 2 and 8 epochs of images. In contrast, the MLS RRab candidates were observed unfiltered and have an average of 130 observations (sampling ranges from 42 to 342 measurements).

We matched the MLS RRab with the SEKBO RRL candidates and found only 103 matches. However, this is not unexpected since the sampling of MLS is highly concentrated within a few degrees of the ecliptic with few observations having been taken near the ∼10° limit. For example, 70% of the MLS RRab's are within 2° of the ecliptic compared to 27% of the SEKBO RRL candidates.

To extend the comparison, we also matched the SEKBO candidates with RRab's from the DR13 sample and the SSS RRab's (G. Torrealba et al. 2013, in preparation). This yielded an additional 538 RRab matches. Because of overlap between the CSS, MLS, and SSS RRab catalogs there were 540 unique RRab stars matching SEKBO RRL candidates. Clearly this number falls far short of the number of SEKBO candidates.

We compared average V magnitudes for the Catalina sources with those of the SEKBO sources. In Figure 5, we present the V magnitude differences between the measurements. The average difference between magnitudes is −0.02 mag with σ_V = 0.16 mag, corresponding to a 7.6% uncertainty in source distances. The internal dispersion for CSS RRab magnitudes from overlapping fields is σ = 0.04 mag and thus does not significantly contribute to the observed dispersion.

Zoom In Zoom Out Reset image size

In order to investigate the SEKBO candidates not found in the Catalina RRab catalogs, we matched the remaining 1481 unmatched SEKBO objects to the SDSS DR8 spectroscopic catalog. We found 40 matches; of these, 35 had measured $\log ({\tt g})$ values and 23 had $\log ({\tt g}) < 3.75$ consistent with RRL (DR13). We also matched the SEKBO sources with SDSS DR8 photometry. Of the 276 sources with SDSS photometry, approximately one third had SDSS colors beyond the limits observed for DR13 RRab's. Next, we matched the 1481 SEKBO sources to photometry from MLS, CSS, and SSS surveys. Among the matches, we sub-selected the objects exhibiting significant variability based on the I_WS > 0.6 and also objects with LS periodic significance p₀ < 1 × 10⁻⁵. In total, we find matches for 90% of the SEKBO objects, including 88% of the sources that were not selected as RRab's. Combining the known Catalina RRab's and other sources with apparent variability, we find that 74.5% of the 1833 SEKBO sources we cover exhibit significant variability. In Table 3, we present the numbers of matches with each of the Catalina surveys, as well as the number of objects selected by our variability and periodicity significance thresholds.

For each of the 1314 SEKBO RRL candidates with Catalina photometry, that were not in our RRab catalogs, we determined an LS period (regardless of their I_WS variability). We inspected both the observed and phase folded light curves of each source and determined a classification based on this photometry. As there are 2759 Catalina light curves matching the 1314 SEKBO candidates, the majority of these sources have multiple light curves. We discovered 551 additional RRL candidates, 140 eclipsing binary candidates (mainly WUMa), and 140 other types of variable sources (including QSOs, CVs, δ Scuti variables, and objects where the classification was unclear). This discovery of many non-periodic sources is not unexpected since a number of these are already known, including QSOs and cataclysmic variables. For example, SEKBO 106646.2532 (CSS080623:140454-102702) is a CV that was discovered in outburst by CRTS in 2008 (Kato et al. 2009). Additionally, based on SDSS DR8 spectroscopy, SEKBO-105832.627 is a CV, SEKBO-117146.1412 is a QSO at z = 1.02, and SEKBO-096514.1360 is an eclipsing WD–MD binary system (Rebassa-Mansergas et al. 2010).

In order to obtain the best periods for the clear RRL, and to find RRL that may have been missed during our inspection, we ran every light curve through the AFD software (G. Torrealba et al. 2013, in preparation). For 307 of the 551 RRL candidates we found good Fourier fits and periods consistent with RRab's and RRc's. In Table 4, we present all the new SEKBO RRL found using Catalina data (including RRL candidates where the period is uncertain). This table contains 263 RRab, 282 RRc, and 11 RRd candidates.

Table 4. SEKBO RR Lyrae

ID	R.A.	Decl. (J2000)	$\bar{V}$	P	A	Var Type
	(°)	(°)		(days)	(mag)
128412.1045	3.65453	3.85789	18.30	...	...	RRd
103591.297	6.04219	2.31516	17.40	0.3044727	0.46	RRc
127806.438	6.34742	−3.25643	17.13	...	...	RRc
116383.660	27.46520	10.92001	17.79	...	...	RRc
117146.1412	29.69399	12.44214	18.60	0.3974842	0.34	RRc
115646.851	44.35595	15.18618	18.35	...	...	RRc
115511.269	50.62031	13.51653	16.77	0.5326870	0.69	RRab
117289.371	64.21176	18.87230	15.57	0.2668644	0.42	RRc
104856.1587	80.69952	13.69941	17.27	...	...	RRc
118386.758	110.85990	21.62161	13.61	0.4071660	0.15	RRc?
118692.330	120.07333	15.15337	14.84	...	...	RRc?
117700.1008	120.99545	15.81325	15.52	...	...	RRc
117775.379	121.92978	15.42998	15.78	0.3299652	0.36	RRc
118801.2653	122.37960	15.61271	19.15	0.4493209	1.23	RRab
117776.476	122.70074	16.11215	15.36	...	...	RRc

Notes. Column 1 gives the SEKBO survey ID. Columns 2 and 3 give the right ascension and declination. Column 4 gives the average magnitude from the AFD fit, if a period is found. Otherwise the weight mean value. Column 5 gives the period. Column 6 gives the amplitude of the AFD fit. Column 7 gives the type of RRL. If a classification is uncertain, a "?" is included.

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

Combining the new RRab's with the 540 from Catalina catalogs we find that ∼36% were missing from our catalogs in good agreement with the simulations presented in DR13. The overall sample suggests that 24% of the RRL are RRc stars whereas Keller et al. (2008) estimated that ∼10% of the SEKBO candidates would be RRc's. However, we note that Pietrukowicz et al. (2012) found 30% RRc among their ∼15, 000 bulge RRL. Nevertheless, it is possible that some fraction of the RRc candidates we identify may be W UMa variables (due to the similarity of their light curves). The uncertain separation of RRc's and W UMa types is the main reason why RRc's are usually excluded as distance indicators.

Our results suggest that 60% of the SEKBO RRL candidates are likely RRL based on the 1833 objects covered by Catalina data. This is 2.2σ smaller than the 24 ± 7% non-RRL contamination estimated by Prior et al. (2009b). However, we note that the Prior et al. (2009b) sample was based on an average of just nine photometric measurements for 106 (∼5%) of the SEKBO stars. Prior et al. (2009a) followed 21 additional SEKBO RRLs overlapping the Sgr tidal stream. More recently, Akhter et al. (2012) followed 137 SEKBO candidates and found 57 to be RRLs, although they found a high fraction with colors matching the colors of RRL candidates from SDSS data.

From Catalina photometry we have an average of 121 observations per light curve (250 per object) covering 90% of the SEKBO sources. Our analysis suggests that ∼25% of the SEKBO sources are either not variable or have significantly less variability than expected for RRL. Nevertheless, with >1000 likely RRL from the SEKBO survey, the results shows that large numbers of RRL can be found using very small numbers of observations and color selection.

To determine the significance of the Gemini stream it is necessary to compare the source density with that expected from halo models. We adopt the halo density model from Sesar et al. (2010) for our comparison. The local density of RRL is determined using the Nth nearest neighbor method. We find the eighth nearest star to the point where we want to calculate the density and calculate the area that bounds these stars. We use eight stars motivated by Ivezic et al. (2005) who found that this number gives the best results for the underlying density on their improved Bayesian method. We also saw that this number was a good combination of precision and computational efficiency.

We compute the observed number density within our grid (r) and density from the Sesar et al. (2010) halo model (r_m). To visualize the overdensities on the Sgr plane, we produce a 100 × 100 grid on the Majewski et al. (2003) Sgr coordinates at Z = 0 (for the X–Y plot) and B = 0 (for the Λ–D_Sgr plot). In Figure 10 we plot the resulting density ratios in the Λ–D_Sgr system and in Figure 11 we plot the densities in the X–Y plane of the Sgr system. The densities here can be readily contrasted sources in Figure 9.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The main density features are the leading and trailing Sgr arms (marked by A and B, respectively). The Gemini stream (marked as C), largely has a higher density ratio than the trailing stream system (marked by B). Another possible feature is seen near Λ = 220, D_Sgr = 50 kpc (marked D). This has much lower significance than the other features, yet is located where models predict the Sgr trailing arm should cross the Galactic plane and meet with the trailing arm B. Since the Gemini feature, like much of the Sgr stream, is offset from the Sgr plane, we have collapsed the source Z positions to visualize the stream.

Unlike the Gemini stream, the leading and trailing Sgr arms A and B are already well known (Law & Majewski 2010, and references therein). To quantify the significance of the Gemini stream, we estimated the number of stars expected by the model at the position of the stream. For simplicity, we selected the volume that bounds the stream in the Sgr coordinate system; that is, 190 < Λ < 240, 1 < B < 17, and 70 < R_Sgr < 130. Based on the Sesar et al. (2010) model density, within this volume we would expect to find 373 RRL assuming 100% detection efficiency. However, only 106 RRL were found. Assuming pure Poisson uncertainties, this suggests the Gemini area has ∼14σ underdensity. As our detection efficiency is close to 70%–80% based on DR13 this short fall cannot be explained purely by missing RRL. The detection of no distant RRL across most of the MLS survey area strongly suggests that the halo is much less dense than expected by the model. This result is in agreement with the results of Watkins et al. (2009) and Sesar et al. (2010) based on halo RRL discovered in SDSS stripe-82. This confirms that the halo density declines more rapidly than suggested by the Juric et al. (2008) halo model.

As a second model comparison, we compare the RRL densities with the Watkins et al. (2009) halo model. Their results suggest a break in density occurs d_GC = 23 kpc. Matching the model to the data we find that the Watkins et al. (2009) model requires normalization by a factor of 11 to match our data. This is in good agreement with the factor of 10 found by Sesar et al. (2010). In Figure 12, we plot the observed radial density compared with the Sesar et al. (2010) model and the normalized Watkins et al. (2009) model. The observations are found to be in good agreement with the Watkins et al. (2009) model, although the break in density appears to occur nearer to 50 kpc when averaged over a large area. However, this is because the densities are enhanced by RRL in the Sgr leading and trailing arms. Based on the Watkins et al. (2009) model we expect to find 50 RRL in the region selected above. The number of Gemini stream RRL we find is thus 7.9σ larger than expected from this model.

Zoom In Zoom Out Reset image size

To investigate differences between the Gemini RRab sources and the overall population, one can infer the Oosterhoff type based on the RRab period–amplitude relationship (Smith et al. 2011 and references therein). However, since average colors were used to transform the MLS light curves, we need to account for the effect of color variation of the RRab light curves. In DR13 we found that RRab amplitudes were systematically reduced by 0.15 due to pulsational color variations within the broad bandpass of CSS images compared to V-band. We also found that the CSS Oosterhoff type-I (OoI) RRab's exhibit a well-defined amplitude limit that is 0.1 mag higher than the Zorotovic et al. (2010) period–amplitude relationship.

To separate OoI RRab candidates from Oosterhoff type-II's (OoII's), we correct the MLS amplitudes to values by adding 0.15 mag. Next we define a set of RRab's that are between 0.1 and 0.25 mag above the average OoI period–amplitude relationship. These objects are a mixture of Oosterhoff types (although some may be Oosterhoff intermediate sources) that we remove from consideration. We then investigated the variation in MLS RRab's Oosterhoff types with distance. In Figure 13, we plot the period–amplitude distribution of the MLS RRab's.

Zoom In Zoom Out Reset image size

By dividing these sources into a nearer sample consisting of RRab with 70 < d_G < 95 kpc and the more distant RRab d_G > 95 kpc we see a division in types. In Figure 13, we also plot the period–amplitude distribution for the Gemini RRab's. We also overplot the Zorotovic et al. (2010) OoI line and an OoII line that is offset by +0.07 in log (P) to match the CSS observations. The nearer RRab's exhibit a mixture of Oosterhoff types when compared to the Zorotovic et al. (2010) period–amplitude relationships. This is significant evidence that the RRab's in the range 70 < d_G < 95 kpc do not come from a single population. In contrast, the distant MLS RRab's all lie near the OoI line, suggesting that they come from a single population that is more metal-rich.

In Figure 14, we show the locations of Gemini stream RRab's after removing the MLS RRab's with ambiguous Oosterhoff type. The CSS RRab's with 70 < d_G < 95 kpc have also been included, but are not separated by Oosterhoff type since their amplitudes and periods are less certain than the MLS RRab's. This figure shows that the Gemini RRab's are spread across the width of the Sgr stream system. It also shows that the OoII RRab's selected among the nearer RRL set (70 < d_G < 95 kpc) are distributed across the region. The nearer OoI group ends around α ∼ 125° where the stream divides between the nearer and farther groups. Analysis of this figure suggests that there may be two or more overlapping populations: one that follows a steep distant gradient and is predominantly OoI, and another that has a shallower gradient and is an OoII population. With respect to this figure, we once again note that the MLS RRL are naturally concentrated toward the ecliptic because of sampling, so it is not possible to make inferences about changes in density across the Sgr stream.

Zoom In Zoom Out Reset image size

While the SDSS photometry has little of the repeated photometry required to unambiguously identify RRab's, it is deeper than both CSS and MLS data (reaching HB stars to g ∼ 22). Additionally, SDSS data do cover most of the Gemini stream observed in the MLS data as well as the main region where MLS data do not overlap the Sgr stream. Therefore, it is possible to use these data to bridge the gap in between the depth of CSS data and coverage of MLS photometry.

Unlike previous authors who searched for BHB stars covering the Sgr tidal streams, here we sought to select both BHB and potential RR Lyrae stars. Based on the colors of the distant MLS RRab's and prior BHB work, we investigated the observed colors of RRL in SDSS photometry and selected SDSS DR8 stars within 0.95 < (u − g)₀ < 1.5, −0.2 < (r − i)₀ < 0.2, −0.35 < (g − r)₀ < 0.22, SDSS object type = 6 (star), and 17 < g₀ < 22. Based on DR13 we know that the bulk of RRab's are located near (g − r)₀ = 0.25, (r − i)₀ = 0.1 and (i − z)₀ = 0.05. However, RRab's with these colors are outnumbered by MSTO stars by a factor of >100 (Koposov et al. 2012). Using maps of SDSS source density, we sub-selected stars in the range 2.7 × (r − i)₀ + 0.25 > (g − i)₀ > 2.7 × (r − i)₀ − 0.1 for (g − i)₀ < 0.05. This selection retains most of the BHB candidates of Ruhland et al. (2011) as well as 22% of the distant MLS RRab sample and many BS stars.

This color selection reduces the initial number of SDSS point sources in our selection from 1.65 million to 81,552. By selecting stars in the Sgr stream region, with −11° < Λ < 11°, the number reduces to 23,507 HB candidates. To match the MLS photometry we transform the SDSS photometry to V magnitudes using Ivezic et al. (2007). As with Newberg et al. (2003, 2007) and Ruhland et al. (2011) we found the spatial distribution of SDSS HB candidates suffers from a significant crowding and background that is best viewed in source density. In Figure 15, we provide the HB candidates in the form of the Hess (point-density) diagram. This figure shows evidence that the density of HB candidates closely follows the fit to the distant Gemini stream of RRab's, in agreement with the distribution of BHB candidates from Ruhland et al. (2011) and others. Here the scaling has been set to match that in Figure 8 where the main streams are visible. In addition to the RRL, the paths expected for BS stars that pass the color cuts and form a shadow that is ∼2 mag fainter than the HB stars is also shown.

Zoom In Zoom Out Reset image size

To obtain another view of MLS RRL in the Gemini stream in relation to the SDSS HB candidates, we divided the SDSS sources into three groups. These were bright sources, with 17 < V < 18.5, intermediate-brightness sources, with 18.5 < V < 19.8, and faint sources, with 19.8 < V < 20.7. In Figure 16, we present the locations of these sources. The bright sources were selected to show the nearby HB stars as well as the BS stars in the Monoceros stream (near the Galactic anti-center at α = 110°–120° limit of the SDSS coverage). The intermediate brightness sources were selected to be indicative of the HB stars in the leading arm of the Sgr stream. The faintest sources were selected to include the HB candidates in the Gemini stream. However, this also includes BS stars that mirror the distribution of HB stars along the Sgr leading arm. The figure clearly shows the overdensity associated with the Sgr streams system. The separation into two streams is not clear here since there are far fewer BHB stars than the MSTO stars used by Belokurov et al. (2006).

Zoom In Zoom Out Reset image size

We matched the entire MLS RRL catalog with the SDSS DR8 spectroscopic catalog and found 89 matches. A much larger sample of RRL with spectra is given in DR13. However, here our main purpose was analysis of the distant halo RRL. We found that 16 of these matches are sources beyond 70 kpc. As noted earlier, many of the faint RRL candidates were found based on their SDSS spectra. The 16 RRL with SDSS spectra include 12 RRab's and four RRc's.

To separate the radial velocities from the velocities due to pulsation, we use the SDSS observation times and the Fourier fits to derive the phase at which the RRL spectrum was observed. As noted in DR13, for SDSS spectra, radial velocities are determined by averaging both Balmer and metallic lines (mainly Ca lines). We follow DR13 by applying Sesar (2012) velocity corrections for pulsation based corrections derived from both hydrogen and metallic lines. The average velocity correction for the 12 RRab is 13 km s⁻¹, in good agreement with the uncertainty derived from 905 RRab spectra in DR13 (σ = 14.3 km s⁻¹). We adopt this level of uncertainty for all the SDSS RRab spectra.

For the four RRc stars that pulsate in the first overtone mode the velocity of the pulsation is much smaller than for the RRab's. Based on the RRc's observed by Liu & Janes (1989) and Jones et al. (1988), the amplitude is expected to be approximately 20 km s⁻¹. To account for this factor, for these RRL we increase the observed SDSS radial velocity uncertainties by an additional 10 km s⁻¹. Following Law & Majewski (2010) we transform the radial velocities to the Galactic standard of rest assuming a solar peculiar motion of (U, V, W) = (9, 12 + 220, 7) km s⁻¹ in the Galactic Cartesian coordinate system.

In Figure 17, we plot the radial velocities for the outer halo MLS RRab's with spectra, the CSS RRab's with d_G > 40 kpc, and the 10 PTF RRab's found by Sesar et al. (2012). We also plot the velocities predicted by Law & Majewski (2010) N-body simulations. The figure shows that two of the RRab's observed near α = 190° appear to be associated with the leading arm of the Sgr stream. However, for the other Gemini stream RRab's, the velocities, like the distances, are not explained by the Law & Majewski (2010) N-body simulations. In this regard, Sesar et al. (2012) noted that the RRab velocity measurements they found suggested the RRab's belonged to two distinct groups. Additionally, sources associated with our group D of Figure 9 (occurring near α = 150°) have radial velocities that appear to vary rapidly with right ascension. These velocities are also inconsistent with the Law & Majewski (2010) model, suggesting that they do not belong to the Sgr leading arm. The association between these feature-D RRab's and the Gemini stream RRab's is unclear, although values are similar.

Zoom In Zoom Out Reset image size

As noted by Newberg et al. (2003), the outer halo BHB candidates they discovered reside near the unusual globular cluster (GC) NGC 2419. This system is located at α = 11453, δ = 3888 and distance d_h = 82.6 kpc (Harris 1996, 2010 edition). The corresponding coordinates in the Majewski et al. (2003) Sgr coordinate system are Λ = 2017, B = −85, making it well within the limits of the Sgr stream system (Koposov et al. 2012). The Galactocentric radial velocity is given by Newberg et al. (2003) as −14 km s⁻¹, and Baumgardt et al. (2009) find an internal velocity dispersion of 4 km s⁻¹.

NGC 2419 is noted as being one of the most metal-poor GCs ([Fe/H] ∼ -2.1; Mucciarelli et al. 2012). The cluster is notably old, with age 12.3 Gyr according to Forbes & Bridges (2010). The system has the highest luminosity (M_V ∼ −9.6 mag) of any GC with a Galactocentric distance R > 15 kpc, apart from the likely Sgr-dwarf-associated GC, M54 (Cohen et al. 2010). The half-light radius of this cluster is 19 pc making it significantly more extended in the log (R_h) versus M_V plane than other GCs with R > 15 kpc (Mackey & van den Bergh 2005). Indeed, the exceptional nature of NGC 2419 relative to outer halo GCs led van den Bergh & Mackey (2004) to suggest that the object is the stripped core of a former dwarf spheroidal galaxy (dSph). Based on abundance studies, Cohen et al. (2010) and Cohen & Kirby (2012) also found that the NGC 2419 appears like no other globular cluster, but rather the core of an accreted dwarf galaxy.

From Figures 7 and 17 we see that the location and radial velocity of NGC 2419 are a relatively good match for the Gemini tidal stream. The average velocities and metallicities of the Sesar et al. (2012) Cancer group B (〈v〉_gsr = 16.3 ± 7.1 km s⁻¹) and [Fe/H] = -2.1 ± 0.4 dex) and the metallicity of the Cancer group B are in reasonable agreement with the values expected for a stream from NGC 2419. However, the velocities and metallicities do not provide a strong enough association to link these sources.

Another important point to consider in the possible association between the Gemini RRL and NGC 2419 is the proximity of the sources. Some of the RRL have δ < 20° near the right ascension of the NGC 2419 (α = 11453). Thus, if one was to assume that the Gemini tidal stream proceeds in the direction of the Sgr stream (as suggested by SDSS HB candidates), the stream stars would have to be dispersed across the entire ∼20° of the Sgr system between the MLS RRab's and NGC 2419. Sources within this gap are not covered by MLS observations because of the coverage limits of the survey. It thus remains unclear whether there is any link between the Gemini stream and NGC 2419. Also, as noted earlier, our data show that the most distant Gemini RRab's are OoI type stars while NGC 2419 is well known to be OoII type. Indeed, Mucciarelli et al. (2012) found that NGC 2419 exhibits very little spread in [Fe/H] (σ = 0.11 dex) suggesting it could not be linked with metal-rich RRL in the Gemini structure.