Stellar Streams Discovered in the Dark Energy Survey

We searched for stellar streams using an unweighted matched-filter algorithm in color–magnitude space (Rockosi et al. 2002; Grillmair 2006; Bonaca et al. 2012; Jethwa et al. 2017). Our matched filter is based on the synthetic isochrone of an old ($\tau =13\,\mathrm{Gyr}$), metal-poor (Z = 0.0002, $[\mathrm{Fe}/{\rm{H}}]=-1.9$), stellar population as constructed by Dotter et al. (2008) and implemented in ugali (Bechtol et al. 2015; Drlica-Wagner et al. 2015). We selected stars within a range of colors around the isochrone according to the criteria

$$\begin{eqnarray}{(g-r)}_{\mathrm{iso}} & + & E\times \mathrm{err}({g}_{\mathrm{iso}}+\mu +{\rm{\Delta }}\mu /2)-{C}_{1}\\ & & \lt (g-r)\lt \\ {(g-r)}_{\mathrm{iso}} & + & E\times \mathrm{err}({g}_{\mathrm{iso}}+\mu -{\rm{\Delta }}\mu /2)+{C}_{2},\end{eqnarray} \tag{ 4 }$$

where ${(g-r)}_{\mathrm{iso}}$ represents the predicted color from the synthetic isochrone at a distance modulus of $\mu =m-M$. We parameterize the magnitude-dependent spread in color due to measurement uncertainties as

$$\begin{eqnarray}&&\mathrm{err}(g)=0.001+{e}^{(g-27.09)/1.09},\end{eqnarray} \tag{ 5 }$$

where the normalization coefficients were derived from fitting the median photometric error as a function of magnitude in the g band (${\mathtt{PSF}}\_{\mathtt{MAG}}\_{\mathtt{G}}$ vs. ${\mathtt{PSF}}\_{\mathtt{MAG}}\_{\mathtt{ERR}}\_{\mathtt{G}}$). We parameterize the selection region around the isochrone with a symmetric magnitude broadening, Δμ, an asymmetric color broadening, ${C}_{\mathrm{1,2}}$, and a multiplicative factor for broadening based on photometric uncertainty, E. We set the values Δμ = 0.5, ${C}_{\mathrm{1,2}}=(0.05,0.10)$, and E = 2 by comparing to the color–magnitude diagrams (CMDs) of old, metal-poor globular clusters and dwarf galaxies (Figure 2).

Zoom In Zoom Out Reset image size

In Figure 2 we show our matched-filter selection overplotted on binned CMDs from the globular clusters NGC 1260 and NGC 7089 (M2), the ultrafaint dwarf galaxy Reticulum II, and the classical dwarf galaxy Sculptor. Our selection retains ≳90% of stellar sources fainter than the MSTO in an annulus of $0\buildrel{\circ}\over{.} 07\lt r\lt 0\buildrel{\circ}\over{.} 12$ around NGC 1260 and NGC 7089. This is a conservative estimate of the true efficiency of our selection, since it does not account for contamination from Milky Way foreground stars within this annulus.

In our initial search for stellar streams in Y3A2, we applied our matched-filter isochrone selection over a grid of distance moduli from $14\lt m-M\lt 19$ spaced at intervals of ${\rm{\Delta }}(m-M)=0.3$. This spacing between distance modulus steps was chosen so that sequential isochrone selections overlap by ≳75% to ensure that streams at intermediate distance moduli were detectable (Grillmair 2017). For each distance modulus, we binned stars passing our matched-filter selection into equal-area HEALPix pixels with area of $\sim 0.013\,{\deg }^{2}$ (${\mathtt{nside}}=512$). We divided the number of stars selected in each HEALPix pixel by a map of the coverage fraction at equivalent resolution to produce a coverage-corrected map of stellar density. We smoothed the density maps with a 2D Gaussian symmetric beam with $\sigma =0\buildrel{\circ}\over{.} 3$.⁵⁴ We later repeated this procedure at spacings of ${\rm{\Delta }}(m-M)=0.1$ and confirmed that all stream candidates were still detected. We show the smoothed density map after the isochrone selection in Figure 3.

Matched filters are often weighted by the ratio of the density of stars in color–magnitude space between a target stellar population and the background population (e.g., Rockosi et al. 2002). However, this weighting disrupts the Poisson-distributed nature of the counts and does not account for changes in the foreground and background stellar populations across the footprint (e.g., close to the Galactic disk, LMC, or Sagittarius stream). For these reasons, we perform an unweighted matched-filter search, where stars are simply accepted or rejected based on their location in color–magnitude space (e.g., Erkal et al. 2017). As a validation, we repeated our search following the procedure of Rockosi et al. (2002) to build a matched-filter isochrone from the CMDs of the globular cluster NGC 7089 and the average Milky Way foreground population. We apply an additive factor of ${\rm{\Delta }}(m-M)$ to shift the CMD of NGC 7089 in distance modulus and thus select for stellar populations over a range of distances. We find that this procedure yields very similar results to our primary unweighted synthetic isochrone selection technique, and all stream candidates reported here were detected by both analyses.

The Sagittarius ($l,b\,\sim \,160^\circ ,-60^\circ$), ATLAS ($l,b\,\sim \,-130^\circ ,-80^\circ$), Phoenix ($l,b\sim -70^\circ ,-65^\circ$), and Tucana III ($l,b\,\sim -45^\circ ,-55^\circ$) streams are clearly visible in smoothed density maps (Figure 3). To further increase our sensitivity to faint stellar streams, we created a smooth model for the stellar foreground and misclassified background galaxies. We mask the dense stellar regions around the Sagittarius stream, the LMC, and the Galactic plane (Figure 4) and fit a 2D, fifth-order polynomial to the distribution of smoothed stellar counts. We subtracted this model from the stellar density to create smooth maps of the residual stellar density (Figure 4). Visual inspection of the foreground-subtracted residual density maps served as the primary technique for identifying new streams.

To facilitate the detection of faint streams, we repeated the procedure described above to generate residual stellar density maps for a range of isochrone selections with distance moduli between $14\leqslant m-M\leqslant 19$ in steps of ${\rm{\Delta }}(m-M)=0.3$. We assembled these sequences of residual density maps into animations of the isochrone-selected stellar density as a function of heliocentric distance. We required that stellar stream candidates appear in the residual density maps for at least two sequential distance moduli. The animations associated with Figures 3 and 4 contain a wealth of information about stellar structure in the Milky Way halo. In this paper we specifically focus on the the most prominent stellar streams, leaving other studies of the outer halo to future work.

We perform our visual search by assembling residual density maps of the full DES footprint and in smaller subregions, which we call "quadrants." Candidates identified in the residual density maps were further examined in color–magnitude space for evidence of a stellar population distinct from the Milky Way foreground (Section 3.3). Only candidates that showed a distinct stellar locus consistent with an old, metal-poor isochrone were included in our list of stellar stream candidates. This search resulted in the detection of the four previously known narrow stellar streams (ATLAS, Phoenix, Tucana III, and Molonglo) and 11 new stream candidates. We report the measured and derived parameters of our stream candidates in Tables 1 and 2, and we discuss each candidate in more detail in Section 4.

Table 1.  Measured Parameters of Stellar Streams

Name	Endpoints	Great-circle Pole	Length	Width	$m-M$	Age	Z	N*	Significance
	(${\alpha }_{2000}$, ${\delta }_{2000}$) (deg)	(${\alpha }_{2000}$, ${\delta }_{2000}$) (deg)	(deg)	(deg)		(Gyr)
Tucana III	(−6.3, −59.7), (3.2, −59.4)	(354.2, 30.3)	4.8	0.18	17.0	13.5	0.0001	700	17.0
ATLAS	(9.3, −20.9), (30.7, −33.2)	(74.3, 47.9)	22.6	0.24	16.8	11.0	0.0007	1600	13.9
Molonglo	(6.4, −24.4), (13.6, −28.1)	(62.3, 51.0)	7.4	0.32	16.8	13.5	0.0010	700	5.2
Phoenix	(20.1, −55.3), (27.9, −42.7)	(311.2, 14.0)	13.6	0.16	16.4	13.0	0.0004	700	11.1
Indus	(−36.3, −50.7), (−8.0, −64.8)	(24.8, 21.6)	20.3	0.83	16.1	13.0	0.0007	9700	21.4
Jhelum	(−38.8, −45.1), (4.7, −51.7)	(359.1, 38.2)	29.2	1.16	15.6	12.0	0.0009	4600	18.6
Ravi	(−25.2, −44.1), (−16.0, −59.7)	(53.2, 11.7)	16.6	0.72	16.8	13.5	0.0003	2300	10.3
Chenab	(−40.7, −59.9), (−28.3, −43.0)	(255.5, 14.4)	18.5	0.71	18.0	13.0	0.0004	1700	15.1
Elqui	(10.7, −36.9), (20.6, −42.4)	(64.0, 38.5)	9.4	0.54	18.5	12.0	0.0004	700	18.4
Aliqa Uma	(31.7, −31.5), (40.6, −38.3)	(94.5, 36.7)	10.0	0.26	17.3	13.0	0.0004	400	9.1
Turbio	(28.0, −61.0), (27.9, −46.0)	(297.8, −0.1)	15.0	0.25	16.1	13.0	0.0004	1000	7.9
Willka Yaku	(36.1, −64.6), (38.4, −58.3)	(316.0, 4.7)	6.4	0.21	17.7	11.0	0.0006	600	7.1
Turranburra	(59.3, −18.0), (75.2, −26.4)	(123.5, 53.3)	16.9	0.60	17.2	13.5	0.0003	1300	14.4
Wambelong	(90.5, −45.6), (79.3, −34.3)	(328.7, −27.3)	14.2	0.40	15.9	11.0	0.0001	500	5.9
Palca	(30.3, −53.7), (16.2, 2.4)	(286.6, −9.9)	57.3	...	17.8	13.0	0.0004	...	...

Note. Measured characteristics of stellar streams detected in DES Y3A2 data. The first section reports DES measurements of previously known streams, while the second section reports narrow streams discovered by DES. The broad stream/stellar overdensity, Palca, is given its own section. Endpoints and great-circle poles are reported in equatorial coordinates and are derived from the residual stellar density analysis described in Section 3.2. Stream lengths are calculated as the angular separation between endpoints, while stream widths, w, come from the standard deviation of a Gaussian fit to the transverse stream profile. Distance moduli, ages, and metallicities were calculated by fitting a Dotter et al. (2008) isochrone to the Hess diagrams described in Section 3.3. The number of stars is calculated by summing MS stars in the background-subtracted Hess diagram. The significance is calculated as the signal-to-noise ratio between the on-stream and off-stream regions.

Download table as:  ASCIITypeset image

After candidates are identified in the residual density images, we perform an iterative process to fit the characteristics of each stream:

• 1.  
  Define stream endpoints from the residual density maps.
• 2.  
  Fit the transverse stream width to define an "on-stream" region.
• 3.  
  Fit isochrone parameters to the CMD of on-stream minus off-stream stars.
• 4.  
  Refit the transverse stream width using the best-fit isochrone selection.

We describe each of these steps in more detail below.

We defined the endpoints of each stream from the residual stellar density map. The residual stellar density map has a pixel scale of $\sim 0\buildrel{\circ}\over{.} 1$ (${\mathtt{nside}}=512$) and is smoothed by a Gaussian kernel with standard deviation $\sigma =0\buildrel{\circ}\over{.} 3$. For narrow, prominent streams the endpoints can be measured with an accuracy of better than $0\buildrel{\circ}\over{.} 1;$ however, for fainter and/or more diffuse structures, measuring endpoints becomes more uncertain. The stream length reported in Table 1 was calculated as the angular separation between the endpoints assuming that the streams follow a great circle on the sky.

For each stream, we calculate the pole of a great circle passing through the endpoints and rotate into a coordinate system where the fundamental plane is aligned with the long axis of the stream. Following the convention of Majewski et al. (2003), we define (heliocentric) longitudinal and latitudinal coordinates (${\rm{\Lambda }},B$) for the rotated coordinate system associated with each stream.

As an initial estimate for the width of each stream, we rotate the HEALPix pixels of our raw and residual density maps into the frame of each stream. We then sum the content of HEALPix pixels along the transverse stream dimension to provide the transverse stream profile. We fit the transverse stream profile with a linear foreground component and a Gaussian stream model with free normalization and standard deviation. The standard deviation of the best-fit Gaussian was taken as an estimate of the stream width, and was used to define signal and background regions for the color–magnitude analysis in the following section. We repeated this procedure after selecting stars consistent with the best-fit isochrone to derive the final stream width.

Stream candidates were examined in color–magnitude space to confirm the presence of a distinct stellar population matched to an old, metal-poor isochrone. An "on-stream" region was selected along the great circle connecting the stream endpoints. For most streams, this region had a width of $\pm 2w$, where w is the stream width derived from the standard deviation of the best-fit Gaussian. Two "off-stream" regions were selected with the same shape as the on-stream region, but offset perpendicular to the stream axis by $\pm 4w$. In some cases this on-stream and off-stream geometry was impossible owing to the boundary of the survey or the presence of large resolved stellar populations (other streams, large dwarf galaxies, etc.). The precise on- and off-stream regions selected for each stream are described in Table 5 in the Appendix. When building on- and off-stream regions, we excise regions around known globular clusters and dwarf galaxies to avoid contaminating the CMD analysis. We calculate the effective solid angle of each region accounting for the excised regions and incomplete survey coverage using the maps described in Section 3.1. We binned the stars in the on-stream region in color–magnitude space with bin size ${\rm{\Delta }}g=0.167$ mag and ${\rm{\Delta }}(g-r)=0.04$ mag. We calculated the effective foreground contribution in each bin of the CMD using the off-stream regions and correcting the difference in effective solid angle. Hess diagrams were smoothed by a 0.75 pixel Gaussian kernel, and the resulting smooth residual CMD was examined for the presence of a distinct stellar population.

We performed a binned maximum-likelihood fit of the smoothed 2D background-subtracted Hess diagrams using a synthetic isochrone from Dotter et al. (2008) weighted by a Chabrier (2001) initial mass function (IMF). We built a binned Poisson likelihood function for the observed number of stars in each CMD bin given the number of stars predicted by our isochrone model convolved by the empirically determined photometric measurement uncertainties (Section 2). We simultaneously fit the richness, distance modulus, age, and metallicity of the isochrone model to the observed excess counts in the Hess diagram. The richness is a normalization parameter representing the total number of stream member stars with mass $\gt 0.1{M}_{\odot }$ (Bechtol et al. 2015). For roughly half the stream candidates, the data were unable to reliably constrain all four parameters simultaneously, and we fixed the age ($\tau =13\,\mathrm{Gyr}$) and metallicity (Z = 0.0004) while fitting richness and distance modulus. Furthermore, we note that there is a significant degeneracy between the age, metallicity, and distance modulus. We estimate a systematic uncertainty on the distance modulus of $\sigma (m-M)\sim 0.4\,\mathrm{mag}$, while spectroscopic observations are essential to break the degeneracy between age and metallicity.

Several streams, specifically those closer to the Galactic plane, suffer from over- or undersubtraction due to gradients in the surrounding stellar density. Mis-subtraction will bias estimates of the richness and total luminosity. To mitigate these issues, we estimate the stellar content of each stream based on the number of MS stars within a region around the best-fit isochrone. We apply a narrow isochrone selection based on Equation (4) using the best-fit age, metallicity, and distance modulus (Table 1) and the selection parameters ${\rm{\Delta }}\mu =0.5$, ${C}_{\mathrm{1,2}}=(0.05,0.05)$, and E = 1.⁵⁵ We sum the content of the background-subtracted Hess diagram within this selection region. We then correct the number of stars for the fraction of the stream width contained in the spatial selection region, to estimate the total number of MS stream stars within the spatial and magnitude range of DES. We record this value as N_* in Table 1. We then use the isochrone model along with a Chabrier (2001) IMF to estimate the total stellar mass, luminosity, and absolute magnitude in Table 2.

Table 2.  Derived Parameters of Stellar Streams

Name	Distance	Length	Width	Stellar Mass	MV	μV	Progenitor Mass
	(kpc)	(kpc)	(pc)	(${10}^{3}{M}_{\odot }$)	(mag)	(mag $\,\mathrm{arcsec}$−2)	(${10}^{4}{M}_{\odot }$)
Tucana III	25.1	2.1	79	3.8	−3.8	32.0	8
ATLAS	22.9	9.0	96	7.4	−4.5	33.0	12
Molonglo	22.9	3.0	128	3.5	−3.7	33.0	30
Phoenix	19.1	4.5	53	2.8	−3.6	32.6	3
Indus	16.6	5.9	240	34.0	−6.2	31.9	650
Jhelum	13.2	6.7	267	13.0	−5.1	33.3	1300
Ravi	22.9	6.6	288	10.4	−5.0	33.4	520
Chenab	39.8	12.9	493	18.3	−5.7	34.1	780
Elqui	50.1	8.2	472	10.4	−4.9	34.3	320
Aliqa Uma	28.8	5.0	131	2.3	−3.4	33.8	18
Turbio	16.6	4.3	72	3.5	−3.9	32.6	10
Willka Yaku	34.7	3.9	127	4.6	−4.1	32.9	14
Turranburra	27.5	8.1	288	7.6	−4.7	34.0	180
Wambelong	15.1	3.7	106	1.6	−3.0	33.7	26
Palca	36.3	36.3	...	...	...	...	...

Note. Derived physical parameters of stellar streams detected in DES Y3A2 data. Heliocentric distances are calculated by fitting a Dotter et al. (2008) isochrone to the Hess diagrams described in Section 3.3 and have an error of ∼20%. Other physical parameters are derived from the measured parameters in Table 1 assuming this distance. Stream lengths are calculated from the angular distance between the stream endpoints, while the widths represent the standard deviation of the best-fit Gaussian. Stellar masses, absolute magnitudes, and surface brightnesses are derived from the richness of the best-fit isochrone model assuming a Chabrier (2001) IMF. The absolute magnitude is derived from the DES g and r bands following the prescription of Bechtol et al. (2015). The surface brightness is derived assuming that 68% of the luminosity is contained within the stream width of ±w. The progenitor masses are estimated using the results of Erkal et al. (2016b).

Download table as:  ASCIITypeset image

We refit the stream width after applying an isochrone selection consistent with the best-fit isochrone. We also calculate the statistical significance of each stream from the on-stream and area-corrected off-stream regions, $S\,\equiv (\mathrm{on}\mbox{--}\mathrm{off})/\sqrt{\mathrm{off}}$. Note that Poisson statistics are valid in this case owing to our use of an unweighted matched filter. For consistency, all on-stream regions had a width of $\pm w$ for this calculation. We report the measured parameters of each stream in Table 1 and derived physical parameters in Table 2. In addition, we use this absolute magnitude to calculate an average surface brightness, which we estimate assuming that 68% of the luminosity is contained within $\pm w$ of the stream axis.

In Table 2 we also provide estimates of the progenitor masses. We use the relation between the stream width and the progenitor mass derived in Erkal et al. (2016b). More precisely, we use Equation (27) from Erkal et al. (2016a), where the progenitor mass is given in terms of the stream width as viewed from the Galactic center and the enclosed mass of the Milky Way at the stream's location. For the mass of the Milky Way, we use the best-fit model in McMillan (2017), who used a range of data to constrain the Milky Way potential. We then use galpot (Dehnen & Binney 1998) to evaluate the circular velocity (and hence the enclosed mass) as a function of galactocentric radius. We note that this method assumes that the streams are on a circular orbit and only works on average for streams on eccentric orbits (Erkal et al. 2016b). Furthermore, this method also assumes that the streams have not fanned out significantly as a result of being in a nonspherical potential (Pearson et al. 2015; Erkal et al. 2016b). As such, this method should be seen as giving a rough estimate of the progenitor mass.

In Table 3 we present stream parameters in galactocentric coordinates, assuming that the Sun is located $8.3\,\mathrm{kpc}$ from the Galactic center (Gillessen et al. 2009; de Grijs & Bono 2016). Galactocentric Cartesian coordinates are provided for endpoints of each stream assuming the heliocentric distance derived in Table 2. We provide longitude and co-latitude (ϕ, ψ) for the pole of a galactocentric orbit passing through the endpoints of each stream.⁵⁶ Assuming a single heliocentric distance for each stream naturally introduces a gradient in the galactocentric radius. We use the average galactocentric radius when calculating galactocentric great-circle orbits. These galactocentric parameters are primarily used to identify potential associations in Section 5.1.

Table 3.  Galactocentric Parameters of Stellar Streams

Name	${x}_{1},{y}_{1},{z}_{1}$	${x}_{2},{y}_{2},{z}_{2}$	$\overline{{R}_{\mathrm{GC}}}$	(ϕ, ψ)
	(kpc)	(kpc)	(kpc)	(deg)
Tucana III	(2.7, −9.4, −20.5)	(0.8, −10.2, −21.1)	23	(285.9, 64.6)
ATLAS	(−8.5, 2.8, −22.7)	(−11.7, −5.6, −22.0)	25	(157.3, 68.5)
Molonglo	(−6.9, 2.1, −22.8)	(−8.3, −0.5, −22.9)	24	(152.3, 72.7)
Phoenix	(−4.5, −8.3, −16.7)	(−8.4, −6.4, −17.9)	20	(235.4, 60.6)
Indus	(2.9, −2.6, −12.0)	(−0.6, −7.3, −12.7)	14	(321.3, 72.0)
Jhelum	(0.9, −0.8, −9.4)	(−4.3, −4.0, −11.9)	11	(298.6, 83.1)
Ravi	(4.7, −1.4, −18.8)	(3.4, −7.9, −18.0)	20	(350.6, 75.4)
Chenab	(18.9, −12.5, −26.3)	(15.5, −1.5, −31.9)	35	(28.7, 68.0)
Elqui	(−2.4, −6.3, −49.4)	(−5.2, −13.9, −48.0)	50	(159.6, 89.9)
Aliqa Uma	(−13.4, −6.7, −27.6)	(−13.5, −11.4, −26.0)	31	(171.2, 66.0)
Turbio	(−5.0, −9.0, −13.5)	(−7.8, −6.3, −15.4)	18	(208.7, 57.3)
Willka Yaku	(−1.5, −21.4, −26.4)	(−4.7, −20.0, −28.1)	34	(229.7, 56.9)
Turranburra	(−24.6, −9.8, −19.9)	(−23.6, −16.6, −15.8)	33	(155.3, 47.5)
Wambelong	(−12.3, −12.8, −6.9)	(−15.0, −10.7, −8.3)	20	(183.1, 28.1)
Palca	(−4.6, −17.5, −31.6)	(−19.8, 13.8, −31.5)	38	(205.8, 69.5)

Note. Galactocentric parameters of stellar streams detected in DES Y3A2 data. Transformation into galactocentric coordinates is performed assuming that the Earth resides at (8.3 kpc, 0, 0). The galactocentric azimuthal and polar angles, (ϕ, ψ), are defined as in Figure 1 of Erkal et al. (2016b). Both endpoints are assumed to be at the heliocentric distance quoted in Table 1.

Download table as:  ASCIITypeset image

The first quadrant (Q1) covers the western portion of the DES footprint from $-45^\circ \lesssim {\alpha }_{2000}\lesssim 10^\circ$ and $-65^\circ \lesssim {\delta }_{2000}\,\lesssim -40^\circ$ (Figure 5). While the DES data extend to ${\alpha }_{2000}\gtrsim -55^\circ$, the low-order polynomial background fit has difficulty modeling the rapidly varying stellar density at these lower Galactic latitudes. Q1 includes the Tucana III satellite and stream, the two most prominent new stellar streams, Indus and Jhelum, and two lower significance streams, Chenab and Ravi. In addition, diffuse stellar overdensities are found at the northern (${\alpha }_{2000},{\delta }_{2000}\sim -28^\circ ,-42^\circ$) and southern (${\alpha }_{2000},{\delta }_{2000}\,\sim \,-10^\circ ,-63^\circ$) edges of Q1; however, the footprint boundary makes it difficult to perform a quantitative evaluation of these structures.

4.1.1. Tucana III Stream

The Tucana III stellar stream is located at a distance of $\sim 25\,\mathrm{kpc}$, extending at least $\pm 2^\circ$ from the ultrafaint satellite Tucana III (Drlica-Wagner et al. 2015). We find that Tucana III appears prominently in the Q1 residual density maps with a projected length of 5° extending from $(-6\buildrel{\circ}\over{.} 3,-59\buildrel{\circ}\over{.} 7)$ to $(3\buildrel{\circ}\over{.} 2,-59\buildrel{\circ}\over{.} 4)$ (Figure 5).

In Figure 6, we show a Hess diagram calculated by subtracting a local background estimate derived from off-stream regions on either side of the Tucana III stream. Despite a well-defined MS and visible RGB, our likelihood analysis has trouble simultaneously fitting the richness, distance modulus, age, and metallicity of the Tucana III stream. Assuming a distance modulus of $m-M=17.0$ and a metallicity of Z = 0.0001 ($[\mathrm{Fe}/{\rm{H}}]=-2.24$) from Drlica-Wagner et al. (2015), we find that the MSTO of Tucana III is well described by an age of $\tau =13.5\,\mathrm{Gyr}$. This is older than the $\tau =10.9\,\mathrm{Gyr}$ reported by Drlica-Wagner et al. (2015), which is due in part to a correction to the synthetic isochrones using an updated version of the DECam filter throughput (T. S. Li et al. 2018, in preparation). In addition, the change in photometric calibration between the DES Y2Q1 and Y3A2 data sets is found to introduce a small color shift.

Zoom In Zoom Out Reset image size

Spectroscopic observations have been unable to conclusively classify Tucana III as an ultrafaint galaxy or star cluster (Simon et al. 2017). The unresolved velocity dispersion (${\sigma }_{v}\lt 1.5\,\mathrm{km}\,{{\rm{s}}}^{-1}$ at 95.5% confidence) and metallicity spread (${\sigma }_{[\mathrm{Fe}/{\rm{H}}]}\lt 0.19$ at 95% confidence) are both low for an ultrafaint dwarf galaxy. However, the mean metallicity ($[\mathrm{Fe}/{\rm{H}}]\,=-{2.42}_{-0.08}^{+0.07}$) and large physical size (${r}_{1/2}=44\pm 6\,\mathrm{pc}$) are both unusual for a globular cluster. In addition, Simon et al. (2017) argue that the mass-to-light ratio of the core of Tucana III is larger than that of a globular cluster, ${\text{}}M/L\gt 20{M}_{\odot }/{L}_{\odot }$, based on its proximity to the Galactic center and the nondetection of a velocity gradient out to 90 $\,\mathrm{pc}$. The core of Tucana III lies slightly offset from the luminosity–metallicity relationship for ultrafaint galaxies (Kirby et al. 2013). Simon et al. (2017) note that if Tucana III has been stripped of $\sim 70 \%$ of its stellar mass, then it would lie directly on the metallicity–luminosity relation of ultrafaint dwarfs. We find that the total stellar mass of the Tucana III stream, including the core and the tidal tails, is $3.8\times {10}^{3}{M}_{\odot }$, which is 4.75 times the stellar mass of the Tucana III core (Drlica-Wagner et al. 2015). This corresponds to a mass loss of 79%, which moves the Tucana III progenitor system onto the luminosity–metallicity relationship for dwarf galaxies.

We search for indications of a distance gradient following a similar procedure to that applied to the Sagittarius stream by Koposov et al. (2012). We transform to a coordinate system oriented along the stream axis and divide the stellar counts into eight longitudinal bins. Within each longitudinal bin, we examine the mean magnitude of MSTO stars satisfying the criteria $0.20\lt (g-r)\lt 0.24$. We find that the mean magnitude of the MSTO changes from $g\sim 16.75$ at the western end of the stream to $g\sim 17.19$ on the eastern end. Fitting a linear gradient model to these data yields a distance gradient of $0.16\pm 0.06\,\mathrm{mag}\,{\deg }^{-1}$. This measurement implies that the Tucana III stream spans $\sim 4\,\mathrm{kpc}$ in distance with a total physical extent of $\sim 4.5\,\mathrm{kpc}$ and that it is on a radial orbit.

4.1.2. Indus Stream

Indus is the first of four new stellar stream candidates detected in Q1. Indus has an angular extent of $20\buildrel{\circ}\over{.} 3$ with a projected width of $0\buildrel{\circ}\over{.} 83$. It may extend beyond the southern edge of the DES footprint in the direction of the SMC; however, at a best-fit distance of 16.6 kpc ($m-M=16.1$), it is unlikely that there is a physical association between the Indus stream and the Magellanic Clouds (located $\sim 3$ times farther away). The physical width of the Indus stream, $\sigma =240\,\mathrm{pc}$ ($\mathrm{FWHM}=565\,\mathrm{pc}$), is comparable to that of the Orphan stream ($\mathrm{FWHM}=688\,\mathrm{pc};$ Belokurov et al. 2007b) and considerably larger than known globular cluster streams. There is no obvious progenitor for Indus (Section 5.1); however, its width may indicate that the Indus stream is the disrupted remains of a faint dwarf galaxy.

The MS of the Indus stream is seen prominently in color–magnitude space (Figure 7), with an estimated absolute magnitude of ${M}_{V}=-6.2$ (Table 2). The measured metallicity of the Indus stream, Z = 0.0007 ($[\mathrm{Fe}/{\rm{H}}]=-1.4$), is considerably higher than would be expected for a dwarf galaxy with similar luminosity (Kirby et al. 2013). The proximity of the Galactic bulge makes it difficult to model the stellar foreground in the vicinity of Indus, and it is possible that foreground contamination in the RGB of Indus may be artificially inflating the measured metallicity (an even more pronounced example can be seen in Jhelum).

Zoom In Zoom Out Reset image size

The southern portion of the Indus stream becomes confused with a more distant diffuse stellar structure ($m-M\geqslant 16.5$) that extends toward the Tucana III stream. Due to the incomplete southern coverage of DES, we cannot determine whether this is the signature of another stream or a diffuse stellar cloud. Other DECam imaging in the regions of the Magellanic Clouds—e.g., the Survey of the Magellanic Stellar History (Nidever et al. 2017) and the Magellanic Satellites Survey (Drlica-Wagner et al. 2016; Pieres et al. 2017, MagLiteS;)—may be able to clarify this question in the near future. However, kinematic information will be necessary to test for any physical connection between Indus/Tucana III and this putative diffuse structure.

4.1.3. Jhelum Stream

4.1.4. Ravi Stream

4.1.5. Chenab Stream

The second quadrant (Q2) spans $0^\circ \lt {\alpha }_{2000}\lesssim 60^\circ$ and $-42^\circ \lt {\delta }_{2000}\lt 7^\circ$ (Figure 8). This quadrant contains the Sagittarius stream, which we have masked from our analysis in order to increase our sensitivity to fainter new streams. We detect four narrow streams in this region, including the previously known ATLAS (Koposov et al. 2014) stream, a possible extension of the Molonglo (Grillmair 2017) stream, and two newly detected streams that we name Elqui and Aliqa Uma. In addition, we find a long, diffuse structure that extends along the height of Q2 intersecting the Eridanus–Phoenix stellar overdensity lower in the DES footprint (Li et al. 2016). We name this structure the Palca stream and discuss it in more detail in Section 4.5.

4.2.1. ATLAS Stream

The ATLAS stream is a narrow stellar stream discovered in the first data release of the VST ATLAS survey, which covered a decl. slice around the stream from $-37^\circ \lesssim {\delta }_{2000}\lesssim -25^\circ$ to a limiting magnitude of $r\sim 22$ (Koposov et al. 2014). The ATLAS stream was later studied by Bernard et al. (2016) using the larger sky coverage provided by Pan-STARRS. The DES analysis is deeper than both VST ATLAS and Pan-STARRS, g = 23.5, and extends the sky coverage around ATLAS to lower decl. The ATLAS stream does not appear to extend significantly beyond the length described by Koposov et al. (2014), ending at $({\alpha }_{2000},{\delta }_{2000})\sim (30\buildrel{\circ}\over{.} 7,-33\buildrel{\circ}\over{.} 2)$. At higher decl. it becomes difficult to disentangle the ATLAS stream from the much more luminous Sagittarius stream before hitting the boundary of the DES footprint at $({\alpha }_{2000},{\delta }_{2000})\sim (9\buildrel{\circ}\over{.} 3,-20\buildrel{\circ}\over{.} 9)$. Using Pan-STARRS data, Bernard et al. (2016) have extended ATLAS to ${\delta }_{2000}\sim -15^\circ$, leading to a total length of $\sim 28^\circ$, of which $22\buildrel{\circ}\over{.} 6$ is contained within DES.

We follow the procedure described in Section 3.2 to characterize the physical properties of the ATLAS stream. The deeper DES data prefer a slightly larger distance of 22.9 kpc ($m-M\sim 16.8$), which is marginally consistent with the previously measured distance, $20\pm 2\,\mathrm{kpc}$, derived using the VST ATLAS data (Koposov et al. 2014). At a distance of $22.9\,\mathrm{kpc}$, the visible portion of the ATLAS stream extends to 9.0 kpc. We independently fit the distance modulus to each half of the ATLAS stream and find evidence that the southwestern portion of ATLAS has a distance modulus that is $\sim 0.3\,\mathrm{mag}$ larger than the northeastern portion. The southwestern portion of ATLAS is detectable in the residual density maps at a distance modulus of $m-M\gt 18.0$. The stellar density is not uniform along the length of ATLAS, and we note a roughly spherical overdensity in the southwestern portion at $({\alpha }_{2000},{\delta }_{2000})=(25\buildrel{\circ}\over{.} 37,-30\buildrel{\circ}\over{.} 13)$. This overdensity is visible at lower significance in Figure 1 of Koposov et al. (2014). The DES data suggest a fainter absolute magnitude for the ATLAS stream, ${M}_{V}=-4.5$, compared to that estimated in the VST ATLAS data, ${M}_{V}\sim -6$ (Koposov et al. 2014).

Figure 10 shows the spatial distribution of stars in a coordinate system aligned with the endpoints of the ATLAS stream. Following the procedure described in Section 3.3, we fit the width of the ATLAS stream with a Gaussian model on top of a linear background. Our measured width of the ATLAS stream, $w=0\buildrel{\circ}\over{.} 24$, is consistent with $w=0\buildrel{\circ}\over{.} 25$ reported by Koposov et al. (2014), where w corresponds to the Gaussian standard deviation. However, it is also clear in Figure 10 that the ATLAS stream deviates appreciably from a great circle on the sky, which would lie along the equator. We find that the ridgeline of the ATLAS stream is well described over the range $9\buildrel{\circ}\over{.} 3\lt {\alpha }_{2000}\lt 30\buildrel{\circ}\over{.} 7$ by a second-order polynomial of the form

$$\begin{eqnarray}&&{\delta }_{2000}=-15.637-0.545{({\alpha }_{2000})-0.001({\alpha }_{2000})}^{2}.\end{eqnarray} \tag{ 6 }$$

Interestingly, Figure 10 also appears to show an underdensity in the stream at ${\rm{\Lambda }}\sim 4^\circ$ that is approximately $2\buildrel{\circ}\over{.} 5$ in size. We caution that this underdensity occurs in a region where the polynomial background fit is complicated by the proximity of the Sagittarius stream. Oversubtraction of the background could manifest as an underdensity in the residual map. To check for the reality of this gap, we also analyzed the raw isochrone-selected counts without any background subtraction and found that in this region there is in fact a deficit with respect to the mean stream density. While the existence of this underdensity remains uncertain, if it is real, it could be due to perturbations by subhalos around the Milky Way (e.g., Ibata et al. 2002; Johnston et al. 2002). Erkal et al. (2016a) estimated the typical size and number of gaps in the ATLAS stream due to subhalos and found a characteristic gap size of $\sim 4^\circ$ with 0.1 gaps expected. However, this prediction depends on the length and orbital trajectory of the ATLAS stream; therefore, given the increased length of the ATLAS stream detected in this work and in Pan-STARRS (Bernard et al. 2016), as well as the uncertainty in its trajectory, the predicted number of gaps is likely an underestimate. If the underdensity is confirmed, then the gap can be used to infer the properties of the subhalo that created the gap (Erkal & Belokurov 2015b), and the statistical properties of the stream density can be used to place constraints on the number of subhalos in the Milky Way (Bovy et al. 2017).

4.2.2. Molonglo Stream

The Molonglo stream was identified as a faint, narrow feature in data from Pan-STARRS (Grillmair 2017). We extrapolate Equation (2) from Grillmair (2017) to detect a narrow, $\sim 8^\circ$ extension of the Molonglo stream in the DES data offset by $\lesssim 2^\circ$. Molonglo is the only stream that is detected based on prior information from another survey, and there is some risk of confirmation bias. In fact, Molonglo is the least apparent feature in the residual density maps, due in part to its proximity to the Sagittarius stream. However, the Hess diagram for Molonglo shows a clear MS and MSTO (Figure 9), and the stream has a detection significance of $5.2\sigma$. After fixing the age ($\tau =13.5\,\mathrm{Gyr}$) and metallicity (Z = 0.0010) to match the Hess diagram, we fit the distance modulus and richness of Molonglo. We determine a distance of 22.9 kpc ($m-M=16.8$), which agrees with the estimated distance of 20 kpc from Grillmair (2017). However, the foreground subtraction in this region is difficult, and the remaining contamination may have artificially inflated the metallicity. Our best-fit width of $0\buildrel{\circ}\over{.} 32$ is narrower than the $\sim 0\buildrel{\circ}\over{.} 5$ reported by Grillmair (2017); however, it is unclear whether his width was measured after convolving with a $0\buildrel{\circ}\over{.} 4$ Gaussian kernel. We do not find evidence of the other three streams identified by Grillmair (2017).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.2.3. Elqui Stream

4.2.4. Aliqa Uma Stream

The third DES quadrant (Q3) covers the region from $5^\circ \lesssim {\alpha }_{2000}\lesssim 60^\circ$ and $-65^\circ \lesssim {\delta }_{2000}\lesssim -42^\circ$ (Figure 11). This quadrant resides above the LMC and SMC and overlaps heavily with the EriPhe stellar overdensity (Li et al. 2016). Several linear structures are detected in this region; however, it is difficult to conclusively differentiate them from a diffuse component of EriPhe. In addition to the Phoenix stream (Balbinot et al. 2016), we identify two new stream candidates, Turbio and Willka Yaku.

4.3.1. Phoenix Stream

4.3.2. Turbio Stream

We find a linear feature near the center of the EriPhe stellar overdensity, which constitutes a candidate stream named Turbio. While Turbio is detected above the background of EriPhe with a significance of $7.9\sigma$, it would be very surprising if the two were not physically associated. It is likely that Turbio stands out more prominently in our analysis compared to that of Li et al. (2016) owing to improved photometric calibration (∼0.7% vs. ∼2%), increased depth ($g\lt 23.5$ vs. $g\lt 22.5$), and improved spatial resolution (smoothing kernel of $0\buildrel{\circ}\over{.} 3$ vs. $0\buildrel{\circ}\over{.} 5$).

The MS of Turbio is detected in a Hess diagram that subtracts a neighboring region of EriPhe to the east of the structure (Figure 12). We fit an isochrone with fixed age and metallicity ($\tau =13.0\,\mathrm{Gyr}$, Z = 0.0004) and find a best-fit distance modulus of $m-M=16.1$, which is indistinguishable from that of EriPhe and the nearby globular cluster NGC 1261. Li et al. (2016) suggest that EriPhe may have a common origin with the Virgo overdensity (Jurić et al. 2008) and the Hercules–Aquila cloud (Belokurov et al. 2007a). However, the orientation of Turbio is nearly perpendicular with the orbit necessary to connect these three diffuse structures. In addition, Turbio may constitute a dense portion of the larger Palca structure that is seen to extend northward from the EriPhe cloud.

Zoom In Zoom Out Reset image size

4.3.3. Willka Yaku Stream

The fourth quadrant of DES (Q4) spans $60^\circ \lesssim {\alpha }_{2000}\lesssim 95^\circ$ and $-65^\circ \lt {\delta }_{2000}\lt -15^\circ$ (Figure 13). It contains the globular clusters NGC 1851 and NGC 1904, and structure from the Monoceros Ring can be seen in the direction of the Galactic anticenter. The proximity of the Monoceros and the Galactic plane makes background subtraction difficult near the eastern edge of this region. Inspection of the residual density maps yields two stellar stream candidates, Turranburra and Wambelong, which are reasonably wide ($w\sim 0\buildrel{\circ}\over{.} 5$) and detected at moderate significance.

4.4.1. Turranburra Stream

The Turranburra stream stretches across the northern portion of Q4, extending from $({\alpha }_{2000},{\delta }_{2000})=(59\buildrel{\circ}\over{.} 3,-18\buildrel{\circ}\over{.} 0)$ to $({\alpha }_{2000},{\delta }_{2000})=(75\buildrel{\circ}\over{.} 2,-26\buildrel{\circ}\over{.} 4)$. The Hess diagram of this stream shows a prominent MS and a hint of an RGB (Figure 14). We estimate that this structure resides at a distance of $27.5\,\mathrm{kpc}$ and has a width of $288\,\mathrm{pc}$. The physical width of Turranburra is more consistent with a dwarf galaxy progenitor, and the photometric estimate of metallicity of Z = 0.0003 is comparable to photometric metallicities determined for ultrafaint galaxies (e.g., Bechtol et al. 2015). The extent of Turranburra is somewhat uncertain owing to its proximity to the edge of the DES footprint and its diffuse nature, especially at its northwestern end. The orbit of Turranburra appears to be distinct from other known streams and globular clusters, without any obvious association or potential progenitor.

Zoom In Zoom Out Reset image size

4.4.2. Wambelong Stream

While the search described here is optimized for ≲1°-wide stellar features, we note that we are sensitive to more diffuse stellar systems. Without the directionality of a narrow stream, interpreting the origin and physical parameters of these diffuse structures becomes much more difficult. However, we do note that the DES Y3A2 data unambiguously confirm the existence of the EriPhe overdensity (Li et al. 2016) and show strong indications of additional substructure around or within this system. Apart from the aforementioned Phoenix, Turbio, and Willka Yaku streams, there are several lower-significance linear structures that overlap EriPhe at a heliocentric distance of $\sim 16.6\,\mathrm{kpc}$ ($m-M=16.1$). One pronounced feature runs nearly east–west with $\delta \approx -56^\circ$ (Figure 11). Unfortunately, image-level masking around the super-saturated star Achernar (${\alpha }_{2000},{\delta }_{2000}=24\buildrel{\circ}\over{.} 43,-57\buildrel{\circ}\over{.} 24$) complicates the interpretation of the southern portion of EriPhe.

The Palca⁶⁰ stream is a broad curvilinear overdensity in the Y3A2 data that extends northward from EriPhe at ${\alpha }_{2000}\sim 30^\circ$. Palca extends from the Turbio stream in the south to the northern boundary of the DES footprint, crossing the Sagittarius stream at ${\alpha }_{2000}\sim 20^\circ$. Palca is more diffuse than the other streams discovered in DES Y3A2 (FWHM $\sim 2^\circ$). To increase our sensitivity to this broad feature, we convolve the residual density maps by a 1° kernel and plot in equatorial coordinates (Figure 15). We find that the northern part of Palca is at a significantly larger distance, $\sim 36\,\mathrm{kpc}$ ($m-M\sim 17.8$), than EriPhe. It is unclear whether this is a signature of a distance gradient from EriPhe to Palca, or whether these are two distinct systems. Assuming that Palca spans the DES footprint (i.e., overlapping with EriPhe), the ridgeline of this structure can be approximated with a second-order polynomial of the form

$$\begin{eqnarray}&&{\alpha }_{2000}=17.277-0.495{\delta }_{2000}-0.0046{\delta }_{2000}^{2}.\end{eqnarray} \tag{ 7 }$$

The above equation is found to be valid for $-55^\circ \,\lt {\delta }_{2000}\lt 2\buildrel{\circ}\over{.} 5$. Palca shows appreciable curvature on the sky, suggesting that it may be in a modestly elliptical orbit. The orbital trajectory of Palca is strongly misaligned with the proposed polar orbit connecting EriPhe with the Virgo overdensity and the Hercules–Aquila cloud (Li et al. 2016). A physical association between Palca and EriPhe would disfavor this hypothesis. However, the orbit of Palca may be broadly consistent with another scenario proposed by Li et al. (2016), suggesting that EriPhe may be the remnants of a disrupted dwarf galaxy originally associated with NGC 1261 and the Phoenix stream.

Zoom In Zoom Out Reset image size

The DES Y3A2 footprint contains five classical globular clusters (NGC 288, NGC 1261, NGC 1851, NGC 1904, and NGC 7089) and four more distant clusters (Whiting 1, AM-1, Eridanus, and Reticulum). While a full investigation of globular clusters is outside the scope of the current paper, we note that our analysis is sensitive to stellar features around these clusters. Four of the classical globular classical clusters (NGC 288, NGC 1261, NGC 1851, and NGC 1904) show hints of extended stellar structure.⁶¹ These features are detectable with the generic isochrone selection described in Section 3 and can be seen in the animations associated with Figure 4. However, to optimize our sensitivity to faint features, we built individual matched-filter selections for each cluster using the CMD of stars within an annulus of $4\buildrel{\,\prime}\over{.} 2\lt r\lt 7\buildrel{\,\prime}\over{.} 2$ around each cluster. We create an optimal weighting by taking the ratio between the density (in color–magnitude space) of cluster member stars compared to the Milky Way foreground population averaged over the DES footprint,

$$\begin{eqnarray}&&{w}_{i,j}={f}_{\mathrm{gc}}(i,j)/{f}_{\mathrm{mw}}(i,j),\end{eqnarray} \tag{ 8 }$$

where $i,j$ index the color and magnitude bins, ${f}_{\mathrm{gc}}(i,j)$ is the normalized density of cluster stars per bin, ${f}_{\mathrm{mw}}(i,j)$ is the normalized density of Milky Way stars, and ${w}_{i,j}$ is the weighting (Rockosi et al. 2002). We mask circular regions comparable to the Jacobi radii of each cluster (Table 4) and convolve the selected stellar density with a Gaussian kernel with $\sigma =0\buildrel{\circ}\over{.} 25$. We follow the same procedure to derive a global polynomial fit to the smoothed density of selected stars and create residual density maps from the difference between the data and the polynomial fit (Figure 16).

Zoom In Zoom Out Reset image size

Table 4.  Globular Cluster Parameters

Name	${\alpha }_{2000}$	${\delta }_{2000}$	$\,{{\rm{D}}}_{\odot }$	rJ	${\mu }_{\alpha }\cos (\delta )$	μδ
	(deg)	(deg)	(kpc)	(pc)	(mas yr−1)	(mas yr−1)
NGC 288	13.189	−26.583	76.4	8.9	4.48	−6.04
NGC 1261	48.068	−55.216	146.4	16.3	1.33	−3.06
NGC 1851	78.528	−40.047	166.5	12.1	1.29	2.38
NGC 1904	81.046	−24.525	153.8	8.9	2.12	−0.02

Note. Centroids, heliocentric distances, Jacobi radii, and proper motions for four classical globular clusters in the DES footprint. Values are taken from Harris (1996, 2010 edition), Balbinot & Gieles (2018), Dinescu et al. (1997), Dambis (2006), and Dinescu et al. (1999).

Download table as:  ASCIITypeset image

We compare the observed stellar features to predictions about orbital motion and tidal tail formation in each globular cluster. We simulated the orbits of the globular clusters using the spray-particle implementation by Küpper et al. (2012), where the escape velocity was modified to match that observed in N-body simulations with realistic tidal fields (Claydon et al. 2017). We assume a Milky Way potential similar to the best-fit Palomar 5 model (Küpper et al. 2015), but with a Jaffe bulge. The cluster initial mass is obtained using the method outlined in Balbinot & Gieles (2018). The tidal tail formation was simulated for the last 6 Gyr of the cluster history, and particles were released every 1 Myr.

The heliocentric distance, sky position, line-of-sight velocity, and integrated magnitude for each cluster were taken from Harris (1996, updated 2010). Proper motions for NGC 288 and NGC 1851 are taken from Dinescu et al. (1997), for NGC 1904 we used values from Dinescu et al. (1999), and for NGC 1261 we used values from Dambis (2006). These parameters are summarized in Table 4. Our simulations assumed a galactocentric solar position of $(8.3\,\mathrm{kpc},0,0)$, a local reflex motion of $U,V,W=(11.1,12.24,7.25)\,\mathrm{km}\,{{\rm{s}}}^{-1}$ (Schönrich et al. 2010), and a circular solar velocity of $U,V,W=(0,233,0)\,\mathrm{km}\,{{\rm{s}}}^{-1}$ (Küpper et al. 2015).

4.6.1. NGC 288

4.6.2. NGC 1261

4.6.3. NGC 1851

4.6.4. NGC 1904

We use the recent catalog of stellar streams compiled by Mateu et al. (2018), augmented with the recently discovered Jet stream (Jethwa et al. 2017), to assess whether any of our stream candidates may be associated with previously detected streams located in other regions of the sky. We begin by transforming the endpoints of each stream into galactocentric Cartesian coordinates. The ellipticity of these streams is poorly constrained, so we assume only that the streams orbit in a plane around the Galactic center. We find the pole of each stream, $(\phi ,\psi )$, defined as the positive normal vector of a plane containing both endpoints. Uncertainties in the pole location for our newly found streams were estimated by assuming a 20% uncertainty on the heliocentric distance before converting to galactocentric coordinates. The poles for the new and previously discovered streams are shown in Figure 17.

Zoom In Zoom Out Reset image size

The distribution of DES stream poles shown in the left panel of Figure 17 is clearly nonuniform. We do not find any strong association of stream poles coinciding with the proposed vast polar structure (VPOS; Pawlowski et al. 2012, 2015). Specifically, in the left panel of Figure 17 we plot the VPOS+new pole (Table 1 of Pawlowski et al. 2015) transformed into galactocentric coordinates assuming a distance of $100\,\mathrm{kpc}$. However, the limited sky coverage of the DES footprint will bias the observable distribution of stream poles. To estimate this bias, we generate a uniform random sample of galactocenric great-circle orbits with a radius of 25 kpc. We calculate the fraction of each great-circle orbit contained within the DES footprint as a function of galactocentric orbital pole and show this in the right panel of Figure 17. We find that the observed distribution of stream poles is consistent with the predictions from our simple simulation. The DES footprint is clearly biased against detecting streams having poles with ${\psi }_{\mathrm{pole}}\lt 30^\circ$ and $-60^\circ \lt {\phi }_{\mathrm{pole}}\lt 120^\circ$. We find qualitatively similar results for random samples of orbits with galactocentric radii of 15 and 50 kpc.

To investigate potential associations for the new DES streams, we plot galactocentric great-circle orbits for the DES streams and other known streams with similar orbital poles (Figure 18). Full phase-space information is necessary to definitively match between streams systems; however, we do note several tentative associations based on the photometrically measured properties of the DES streams. Figure 18 shows a strong correspondence between Ravi and the tentative candidate RR Lyrae stream 24.5-1 (Mateu et al. 2018). The poles of these two stream candidates match within $3^\circ$, while their distance moduli differ by ${\rm{\Delta }}(m-M)\sim 0.2\,\mathrm{mag}$ (well within the systematic error associated with our isochrone fitting). Such an association supports the robustness of tentative candidates identified below the conservative $\gt 4\sigma$ significance threshold of Mateu et al. (2018).

Zoom In Zoom Out Reset image size

The orbital pole of the Hermus stream (Grillmair 2014) is only $3\buildrel{\circ}\over{.} 4$ from that of Willka Yaku, but the galactocentric distances of the two streams differ by $\sim 16\,\mathrm{kpc}$. It has previously been suggested that Hermus may be a northern extension of the Phoenix stream (Grillmair & Carlberg 2016), which resides along a slightly different orbital plane but is well matched in distance. Kinematic information would help to resolve this ambiguity.

Globular clusters and dwarf galaxies may provide possible progenitors for the newly discovered streams. The globular cluster candidate with the smallest separation is IC 4499. It is within $1\buildrel{\circ}\over{.} 2$ of the great-circle orbit of Turbio and has a galactocentric radius that differs by less than $2\,\mathrm{kpc}$. IC 4499 is a moderate-mass, low-density cluster with signatures of an extratidal stellar halo (Walker et al. 2011). Cetus II, an ultrafaint dwarf galaxy (Drlica-Wagner et al. 2015), is another candidate for association. It lies $1\buildrel{\circ}\over{.} 7$ off of the great-circle orbit of Aliqa Uma, has a galactocentric radius that is $1.1\,\mathrm{kpc}$ larger, and is located $43^\circ$ from the nearest endpoint of the stream. We note also that the tidal tails of Palomar 5 (${R}_{\mathrm{GC}}=18.2\,\mathrm{kpc}$) have an orbital pole within $6^\circ$ of the ATLAS stream ($25\,\mathrm{kpc}$).

Agnello (2017) identified four stellar stream candidates overlapping the DES footprint using a WISE-Gaia multiple search. It was proposed that two of these streams, WG3 and WG4, may have associated counterparts in DES (Agnello 2017). While WG3 and WG4 appear qualitatively similar to Indus and Jhelum, their absolute coordinates are offset by ${\rm{\Delta }}{\alpha }_{2000}\sim -20^\circ$ (an angular separation of $\sim 15^\circ$). We do not see any significant stellar overdensities associated with the positions of WG3 or WG4 reported by Agnello (2017), though their proximity to the Galactic plane and unknown distance make it difficult to quantify the lack of a DES counterpart. If we allow ${\rm{\Delta }}{\alpha }_{2000}$ offsets of $\sim 10^\circ$, then we find a possible correspondence between WG1 and Wambelong. WG1 is offset from Wambelong by ${\rm{\Delta }}{\alpha }_{2000}\sim -8^\circ$ (angular separation of $\sim 6^\circ$) and extends both northeast and southwest along a similar path. This observation provides circumstantial evidence in support of the longer extent of Wambelong proposed in Section 4.4.2. WG2 does not correspond to any of the high-significance stream candidates reported here. However, WG2 appears qualitatively similar to a lower-significance feature found to the southwest of NGC 1851 extending from ${\alpha }_{2000},{\delta }_{2000}=(56.1,-50.4)$ to $(78.5,-40.0)$ at a distance modulus of $m-M\sim 17.5$ (animation of Figure 4). We expect Gaia DR2 to greatly improve the power of stellar stream searches using astrometric techniques.

Stellar streams can be used to constrain the Milky Way gravitational potential (e.g., Johnston et al. 2005; Koposov et al. 2010; Law & Majewski 2010; Gibbons et al. 2014; Bowden et al. 2015; Küpper et al. 2015; Bovy et al. 2016). Full potential modeling is beyond the scope of this work; however, we note that the streams discovered by DES span a wide range of galactocentric radii and should be able to constrain how the Milky Way's density profile and shape evolve with radius. In this context, we expect that the ATLAS stream will be especially useful since it is long and does not lie on a great circle (see Figure 10).

However, even without sophisticated modeling we can make some general observations in the context of the Milky Way potential. Erkal et al. (2016b) suggest that the connection between stream width and orbital inclination could provide an independent constraint on the symmetry axis and flattening of the Milky Way halo. In Figure 19, we plot the angular stream width (as would be observed from the Galactic center) against the galactocentric polar angle, ψ, and the galactocentric azimuthal angle, ϕ. This figure shows a large scatter, which is to be expected from a heterogeneous population of progenitors. While it is interesting to note that the streams that appear the widest are also on nearly polar orbits, interpreting any trend in this figure is subject to a number of caveats.

Zoom In Zoom Out Reset image size

In particular, inferences involving stream widths rely on the assumed mass, structural properties, orbit, and dynamical age of the stream and its progenitor. Geometric effects can cause debris in the stream plane to contribute to the perceived width as calculated by a heliocentric observer. Furthermore, the width can fluctuate along a stream owing to the existence of nodes between the progenitor plane and the planes of the stream debris. Nonetheless, it is interesting to note that Figure 19 is in general agreement with a model where the Galactic symmetry axis is perpendicular to the plane of the disk.⁶²

In Table 2 we provided estimates of the stellar mass and progenitor mass (based on the stream width; see Erkal et al. 2016b) of each stream. These can be combined to give a mass-to-light ratio for the progenitor. Doing this, we find that five of the streams (Tucana III, ATLAS, Phoenix, Willka Yaku, and Turbio) have mass-to-light ratios less than ∼30, while the other eight streams have significantly higher mass-to-light ratios. Given this seeming dichotomy, it is possible that the progenitors with low mass-to-light ratios are globular clusters while those with the higher ratios are dwarf galaxies. However, since the mass-to-light ratios are all higher than expected for a globular cluster (Phoenix stream has the lowest ratio of 10), we cannot make any firm conclusions.

These higher-than-expected mass-to-light ratios could be due to a variety of reasons. As discussed in 3.3, the progenitor mass estimate is only approximate and only works on average for streams on eccentric orbits. Furthermore, if the stream widths have fanned out owing to evolving in a nonspherical potential (e.g., Erkal et al. 2016b), the inferred progenitor masses will be overestimated. Finally, the stellar masses will be underestimated for streams that are not fully contained within the DES footprint.