EIGHT NEW MILKY WAY COMPANIONS DISCOVERED IN FIRST-YEAR DARK ENERGY SURVEY DATA

Several independent searches of the stellar density field were conducted. One approach involved direct visual inspection of coadded images. Other searches used binned stellar density maps constructed from the coadd object catalogs. As an example, we detail below how one of these maps was built and analyzed.

We began by spatially binning the stellar catalog into equal-area pixels using the HEALPix scheme (Górski et al. 2005).⁴⁶ We considered HEALPix pixel sizes of $\sim 0\buildrel{\circ}\over{.} 06$ (nside = 1024) and $\sim 0\buildrel{\circ}\over{.} 11$ (nside = 512) to optimize sensitivity to satellites possessing different angular sizes. Since the stellar density is greatly enhanced in regions of the Y1A1 footprint near the LMC and Galactic plane, we further grouped the stars into larger regions of $\sim 13\;{\mathrm{deg}}^{2}$ (nside = 16) to estimate the local field density of stars. We corrected the effective solid angle of each pixel using the survey coverage, as estimated by mangle as part of DESDM processing (Swanson et al. 2008).⁴⁷ Several conspicuous stellar over-densities were immediately apparent after this simple spatial binning procedure.

We increase our sensitivity to ultra-faint satellite galaxies by focusing our search on regions of color–magnitude space populated by old, low-metallicity stellar populations (Koposov et al. 2008; Walsh et al. 2009). As a template, we used a PARSEC isochrone corresponding to a stellar population of age $12\;\mathrm{Gyr}$ and metallicity $Z=0.0002$ (Bressan et al. 2012). Sensitivity to satellites at varying distances was enhanced by considering 20 logarithmically spaced steps in heliocentric distance ranging from 20 to $400\;\mathrm{kpc}$ (distance moduli $16.5\lt M-m\lt 23.0$). For each step in distance, all stars within $0.2\;\mathrm{mag}$ of the isochrone in magnitude–magnitude space were retained while those outside the isochrone template were discarded. We then created a significance map for each $\sim 13\;{\mathrm{deg}}^{2}$ region by computing the Poisson likelihood of finding the observed number of stars in each map pixel given a background level characterized by the local field density.

The simple approach described above is computationally efficient and easily generalizable. However, a more sensitive search can be performed by simultaneously modeling the spatial and photometric distributions of stars and incorporating detailed characteristics of the survey (variable depth, photometric uncertainty, etc.). One way to incorporate this information is through a maximum-likelihood analysis (Fisher 1925; Edwards 1972). Likelihood-based analyses have found broad applicability in studies of Milky Way satellites (e.g., Dolphin 2002; Martin et al. 2008a). Here we extend the maximum-likelihood approach to a wide-area search for Milky Way satellites. Similar strategies have been applied to create catalogs of galaxy clusters over wide-field optical surveys (e.g., Rykoff et al. 2014).

Our maximum-likelihood search begins by assuming that the stellar catalog in a small patch of sky represents a Poisson realization of (1) a field contribution including Milky Way foreground stars, mis-classified background galaxies, and imaging artifacts, and (2) a putative satellite galaxy. The unbinned Poisson log-likelihood function is given by

$$\begin{eqnarray}&&\mathrm{log}{\mathcal{L}}=-f\lambda +\underset{i}{\displaystyle \sum }\left(1-{p}_{i}\right),\end{eqnarray} \tag{ 1 }$$

where i indexes the objects in the stellar sample. The value p_i can be interpreted as the probability that star i is a member of the satellite, and is computed as

$$\begin{eqnarray}&&{p}_{i}\equiv \displaystyle \frac{\lambda {u}_{i}}{\lambda {u}_{i}+{b}_{i}}.\end{eqnarray} \tag{ 2 }$$

Here, u represents the signal probability density function (PDF) for the satellite galaxy and is normalized to unity over the spatial and magnitude domain, ${\mathcal{S}}$; specifically, ${\int }_{\mathrm{all}}u\;d{\mathcal{S}}=1$. The corresponding background density function for the field population is denoted by b.

We define the richness, λ, to be a normalization parameter representing the total number of satellite member stars with mass $\gt 0.1{M}_{\odot }$. In Equation (1), $f\equiv {\int }_{\mathrm{obs}}u\;d{\mathcal{S}}$ represents the fraction of satellite member stars that are within the observable spatial and magnitude domain of the survey, and $f\lambda$ denotes the expected number of observable satellite member stars.⁴⁸ Maximizing the likelihood with respect to the richness implies $f\lambda ={\sum }_{i}{p}_{i}$. This condition makes clear that the satellite membership probability for each star in the catalog is a natural product of the maximum-likelihood approach. These membership probabilities can be used to prioritize targeting when planning spectroscopic follow-up observations. Figure 2 highlights the use of membership probabilities to visualize a low-surface-brightness satellite galaxy candidate.

Zoom In Zoom Out Reset image size

To characterize a candidate satellite galaxy, we explore the likelihood of the data, ${\mathcal{D}}$, as a function of a set of input model parameters, ${\boldsymbol{\theta }}$. The signal PDF is assumed to be separable into two independent components,

$$\begin{eqnarray}&&u({{\mathcal{D}}}_{i}| {\boldsymbol{\theta }})={u}_{s}({{\mathcal{D}}}_{s,i}{| {\boldsymbol{\theta }}}_{s})\times {u}_{c}({{\mathcal{D}}}_{c,i}{| {\boldsymbol{\theta }}}_{c}).\end{eqnarray} \tag{ 3 }$$

The first component, ${u}_{s}$, depends only on the spatial properties, while the second component, ${u}_{c}$, depends only on the distribution in color–magnitude space.

We modeled the spatial distribution of satellite member stars with an elliptical Plummer profile (Plummer 1911), following the elliptical coordinate prescription of Martin et al. (2008a). The Plummer profile is sufficient to describe the spatial distribution of stars in known ultra-faint galaxies (Muñoz et al. 2012b). The spatial data for catalog object i consist of spatial coordinates, ${{\mathcal{D}}}_{s,i}=\{{\alpha }_{i},{\delta }_{i}\}$, while the parameters of our elliptical Plummer profile are the centroid coordinates, half-light radius, ellipticity, and position angle, ${{\boldsymbol{\theta }}}_{s}=\{{\alpha }_{0},{\delta }_{0},{r}_{{\rm{h}}},\epsilon ,\phi \}$.

We modeled the color–magnitude component of the signal PDF with a set of representative isochrones for old, metal-poor stellar populations, specifically by taking a grid of isochrones from Bressan et al. (2012) spanning $0.0001\lt Z\lt 0.001$ and $1\;\mathrm{Gyr}\lt \tau \lt 13.5\;\mathrm{Gyr}$. Our spectral data for star i consist of the magnitude and magnitude error in each of two filters, ${{\mathcal{D}}}_{c,i}=\{{g}_{i},{\sigma }_{g,i},{r}_{i},{\sigma }_{r,i}\}$, while the model parameters are composed of the distance modulus, age, and metallicity describing the isochrone, ${{\boldsymbol{\theta }}}_{c}=\{M-m,\tau ,Z\}$. To calculate the spectral signal PDF, we weight the isochrone by a Chabrier (2001) initial mass function (IMF) and densely sample in magnitude–magnitude space. We then convolve the photometric measurement PDF of each star with the PDF of the weighted isochrone. The resulting distribution represents the predicted probability of finding a star at a given position in magnitude–magnitude space given a model of the stellar system.

The background density function of the field population is empirically determined from a circular annulus surrounding each satellite candidate ($0\buildrel{\circ}\over{.} 5\lt r\lt 2\buildrel{\circ}\over{.} 0$). The inner radius of the annulus is chosen to be sufficiently large that the stellar population of the candidate satellite does not bias the estimate of the field population. Stellar objects in the background annulus are binned in color–magnitude space using a cloud-in-cells algorithm and are weighted by the inverse solid angle of the annulus. The effective solid angle of the annulus is corrected to account for regions that are masked or fall below our imposed magnitude limit of $g\lt 23\;\mathrm{mag}$. The resulting two-dimensional histogram for the field population provides the number density of stellar objects as a function of observed color and magnitude (${\mathrm{deg}}^{-2}\;{\mathrm{mag}}^{-2}$). This empirical determination of the background density incorporates contamination from unresolved galaxies and imaging artifacts.

The likelihood formalism above was applied to the Y1A1 data set via an automated analysis pipeline.⁴⁹ For the search phase of the algorithm, we used a radially symmetric Plummer model with half-light radius ${r}_{{\rm{h}}}=0\buildrel{\circ}\over{.} 1$ as the spatial kernel, and a composite isochrone model consisting of four isochrones bracketing a range of ages, $\tau =\{12,13.5\;\mathrm{Gyr}\}$, and metallicities, $Z=\{0.0001,0.0002\}$, to bound a range of possible stellar populations. We then tested for a putative satellite galaxy at each location on a three-dimensional grid of sky position (0.7 arcmin resolution; nside = 4096) and distance modulus ($16\lt M-m\lt 24$; $16-630\;\mathrm{kpc}$).

The statistical significance at each grid point can be expressed as a Test Statistic ($\mathrm{TS}$) based on the likelihood ratio between a hypothesis that includes a satellite galaxy versus a field-only hypothesis:

$$\begin{eqnarray}&&\mathrm{TS}=2\left[\mathrm{log}{\mathcal{L}}(\lambda =\hat{\lambda })-\mathrm{log}{\mathcal{L}}(\lambda =0)\right].\end{eqnarray} \tag{ 4 }$$

Here, $\hat{\lambda }$ is the value of the stellar richness that maximizes the likelihood. In the asymptotic limit, the null-hypothesis distribution of the $\mathrm{TS}$ will follow a ${\chi }^{2}/2$ distribution with one bounded degree of freedom (Chernoff 1954). We have verified that the output distribution of our implementation agrees well with the theoretical expectation by testing on simulations of the stellar field. In this case, the local statistical significance of a given stellar over-density, expressed in Gaussian standard deviations, is approximately the square root of the $\mathrm{TS}$.

Brief comments on the individual galaxy candidates are provided below, and spatial maps and color–magnitude diagrams for each candidate are provided in Figures 4–11. The rightmost panels of these Figures show the satellite membership probabilities of individual stars that are assigned by the likelihood fit using a single representative isochrone with $\tau =13.5$ Gyr and $Z=0.0001$. Stars with high membership probabilities contribute most to the statistical significance of each candidate. The constellation designation, should these candidates be confirmed as dwarf galaxies, is listed in parenthesis.

• 1.  
  DES J0335.6–5403 (Reticulum II, Figure 4): as the nearest and most significant candidate, DES J0335.6−5403 is highly conspicuous in the Y1A1 stellar density maps, with $\sim 300$ member stars brighter than $g\sim 23\;\mathrm{mag}$. In fact, an over-density of faint stars at this position is even visible in the much shallower Digitized Sky Survey images, although it was not detected by Whiting et al. (2007) and other searches of photographic material. Note that like the previously known ultra-faint dwarfs, DES J0335.6−5403 very likely contains several blue horizontal branch stars identified by the likelihood procedure, two of which are relatively far from the center of the object. Given its luminosity, radius, and ellipticity, DES J0335.6−5403 is almost certainly a dwarf galaxy rather than a globular cluster. As illustrated in Figure 3, it is significantly more extended than any known faint globular cluster, and its elongated shape would also make it an extreme outlier from the Milky Way cluster population. Among known dwarfs, DES J0335.6−5403 appears quite comparable to Ursa Major II (Zucker et al. 2006a; Muñoz et al. 2010). DES J0335.6−5403 is only $\sim 23\;\mathrm{kpc}$ from the LMC, and measurements of its radial velocity and proper motion will provide strong clues as to whether it originated as a Milky Way satellite or fell into the Milky Way halo as part of a Magellanic group.
• 2.  
  DES J0222.7–5217 (Eridanus II, Figure 5): at a distance of $\gt 330\;\mathrm{kpc}$, DES J0344.3−4331 is nearly a factor of three more distant than any known outer halo globular cluster, and its half-light radius of $\sim 170\;\mathrm{pc}$ is inconsistent with the sizes of globular clusters. It is therefore very likely that this object is a new dwarf galaxy. The color–magnitude diagram of DES J0344.3−4331 closely resembles that of another distant Milky Way satellite, Canes Venatici I, with a well-populated horizontal branch covering a wide range of colors (Zucker et al. 2006b; Martin et al. 2008b; Okamoto et al. 2012). Its large distance places DES J0344.3−4331 in a very intriguing range of parameter space for studying the quenching and loss of gas in dwarf galaxies. As has been known for many years, dwarf galaxies within $\sim 250\;\mathrm{kpc}$ of the Milky Way and M31 are almost exclusively early-type galaxies with no gas or recent star formation, while dwarfs beyond that limit often have irregular morphologies, contain gas, and/or are still forming stars (e.g., Einasto et al. 1974; Blitz & Robishaw 2000; Grcevich & Putman 2009; Spekkens et al. 2014). The next most distant Milky Way dwarf galaxy, Leo T, is gas-rich and has hosted recent star formation (de Jong et al. 2008; Ryan-Weber et al. 2008); deeper optical and H i imaging to search for neutral gas and young stars in DES J0344.3−4331 has the potential to provide new insight into how gas is stripped from low-mass dwarfs and reveal the minimum mass for maintaining star formation over many Gyr. A radial velocity measurement will also shed light on if DES J0344.3−4331 has already passed close to the Milky Way or whether it is infalling for the first time. Given its distance, it is unlikely to be associated with the Magellanic Clouds. Like DES J0335.6−5403, DES J0344.3−4331 is clearly detected in Digitized Sky Survey images dating as far back as 1976.
• 3.  
  DES J2251.2-5836 (Tucana II, Figure 6): DES J2251.2−5836 is the third new satellite with a large enough size ($120\;\mathrm{pc}$) to be tentatively identified as a dwarf galaxy with DES multi-band photometry alone. It has a similar luminosity to DES J0335.6−5403, but is a much lower surface brightness system and is a factor of $\sim 2$ farther away. DES J2251.2−5836 is $\sim 19\;\mathrm{kpc}$ from the LMC and $\sim 37\;\mathrm{kpc}$ from the SMC, making it a strong candidate for another member of the Magellanic group. In the surface density map in the upper left panel of Figure 6, the outer regions of DES J2251.2−5836 appear elongated and distorted. However, these features are likely a result of noise rather than real distortions (Martin et al. 2008a; Muñoz et al. 2010). The distribution of likely member stars in the upper right panel is much rounder. The high detection significance of this object demonstrates the power of the likelihood analysis to simultaneously combine spatial and color–magnitude information.
• 4.  
  DES J0255.4–5406 (Horologium I, Figure 7): DES J0255.4−5406, at a distance of $\sim 87\;\mathrm{kpc}$, has only a sparse population of red giant branch and horizontal-branch stars visible in the DES photometry (a hint of the main sequence turnoff may be present at the detection limit). Given the small number of stars visible in the DES data, deeper imaging and spectroscopy will be needed to characterize the system more fully. Its Plummer radius of $\sim 60\;\mathrm{pc}$ establishes it as a likely dwarf galaxy, twice as extended as the largest globular clusters with comparable luminosities. DES J0255.4−5406 (perhaps along with DES J0443.8−5017 and DES J2339.9−5424; see below) is significantly farther from the Milky Way than any previously known dwarf galaxy with ${M}_{V}\lesssim -4$, suggesting that tidal stripping may not be needed to explain the low luminosities of the faintest dwarfs. On its own, the Galactocentric distance is not necessarily a good indicator of the past importance of Galactic tides in shaping the photometric and spectroscopic properties of satellites. The most important factor is the peri-Galacticon distance, which is not yet known for the new satellites. DES J0255.4−5406 is $\sim 40\;\mathrm{kpc}$ away from the Magellanic Clouds and a factor ∼2 closer to them than to the Milky Way, making it a potential Magellanic satellite. If it is (or was) associated with the Magellanic group, it is possible that tides from the LMC could have been relevant to its evolution. A measurement of the systemic velocity will help clarify whether DES J0255.4−5406 is currently near apocenter, infalling, or associated with the Magellanic system.
• 5.  
  DES J2108.8–5109 (Indus I, Figure 8): we identify DES J2108.8−5109 with Kim 2 (Kim et al. 2015). DES J2108.8−5109 is one of the faintest (${M}_{V}\sim -2.2\;\mathrm{mag}$) and most compact (${r}_{{\rm{h}}}\sim 12$ pc) of the satellites discussed here. The object is visible in the coadded DES images. If it is a globular cluster, as argued by Kim et al. (2015), DES J2108.8−5109 is fainter and more extended than most of the other outer halo clusters, such as AM 1, Eridanus, Pal 3, Pal 4, and Pal 14. DES J2108.8−5109 is $\sim 37$ kpc from the SMC, $\sim 55$ kpc from the LMC, and $\sim 69$ kpc from the much more massive Milky Way, so it is more likely a satellite of the Milky Way than of the Magellanic Clouds.
• 6.  
  DES J0443.8-5017 (Pictoris I, Figure 9): DES J0443.8−5017 has a large enough radius to be a likely dwarf galaxy, but the uncertainty on the radius measurement is large enough to make it also consistent with the globular cluster population. DES J0443.8−5017 has a prominent blue horizontal branch and hints of an elliptical shape, but fewer member stars are detected in the DES data. More accurate measurements of size and shape from deeper imaging, and kinematics and chemical abundances from spectroscopy, will be required to determine the nature of this object. The large distance of DES J0443.8−5017 places it far enough behind the Magellanic Clouds that it is less likely to be a Magellanic satellite than many of the other new discoveries.
• 7.  
  DES J2339.9–5424 (Phoenix II, Figure 10): DES J2339.9−5424 is quite similar to DES J0443.8−5017, but slightly smaller and closer. Again, we cannot draw firm conclusions on its nature without additional data, At $\sim 43\;\mathrm{kpc}$ from the SMC and $\sim 65\;\mathrm{kpc}$ from the LMC, it is unclear whether DES J2339.9−5424 could plausibly be a Magellanic satellite.
• 8.  
  DES J0222.7–5217 (Eridanus III, Figure 11): DES J0222.7−5217 is the most compact of the newly discovered objects in both angular and physical units. Along with DES J2108.8−5109, it is the most likely of the new discoveries to be a distant globular cluster rather than a dwarf galaxy. However, they could also be consistent with an extension of the dwarf galaxy locus to fainter magnitudes and smaller sizes. Even though it is one of the lowest luminosity system identified in the DES data so far, its compactness gives it a relatively high surface brightness, and like DES J2108.8−5109 and DES J0443.8−5017, it is clearly visible in coadded images. However, only a handful of likely member stars are resolved at the depth of the Y1A1 data, and significantly deeper imaging will be needed to better constrain its physical properties and stellar population.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Given that no additional ultra-faint Milky Way satellite galaxies have been confirmed outside of the SDSS DR7 footprint, despite the large areas of sky subsequently observed by SDSS and Pan-STARRS, it is interesting that multiple candidates have been found within the comparatively small area explored by DES thus far.⁵¹ Without definite classifications, it is difficult to incorporate the DES candidates into constraints on the luminosity function of Milky Way satellite galaxies. However, it is still possible to quantify the sensitivity of the first-year DES search using simple semi-analytic estimates of the completeness.

First, we calculated the probability that each new satellite could have been detected in the Y1A1 data. We began by generating a large number of realizations of each galaxy candidate distributed uniformly over the Y1A1 footprint. The candidates were modeled using radially symmetric Plummer profiles and the realizations included shot noise due to the limited number of stars expected to be in the observable magnitude range of DES. We then applied the simple map-based detection algorithm described in Section 3.1 to evaluate the detection efficiency. To be "detected," the satellite must possess at least 10 stars brighter than our imposed magnitude limit ($g\lt 23$) and a large enough surface brightness to pass the visual search selection criteria. Specifically, we considered extraction of varying sizes and computed the Poisson probability of detecting ${n}_{\mathrm{satellite}}+{n}_{\mathrm{field}}$ stars when expecting ${n}_{\mathrm{field}}$ stars based on the local field surface density. We tested extraction regions with sizes corresponding to the pixel areas in the map-based search algorithm (${r}_{\mathrm{ext}}=\{0\buildrel{\circ}\over{.} 029,0\buildrel{\circ}\over{.} 057\}$; Section 3.1) and the kernel size from the likelihood scan (${r}_{\mathrm{ext}}=0\buildrel{\circ}\over{.} 1$; Section 3.2), as well as an extraction radius set to the angular half-light radius of the simulated satellite. When computing the local field density, we selected only stars along the isochrone at the distance of the satellite with $\tau =12\;\mathrm{Gyr}$ and $Z=0.0002$ (see Section 3.1). Table 3 summarizes the expected detection efficiencies for the DES candidates when applying a $5\sigma$ statistical significance threshold, as in our seed selection procedure for the map-based search. The results show that all of the DES candidates would have been identified over a substantial fraction of the Y1A1 footprint with non-negligible probability, and for several candidates, near certainty.

Table 3.  Expected Detection Efficiencies for Milky Way Companions in DES Y1A1

Name	MV	Distance	rh	Efficiency	Efficiency	Efficiency	Efficiency
		(kpc)	$(\mathrm{deg})$	$({r}_{\mathrm{ext}}=0\buildrel{\circ}\over{.} 029)$	$({r}_{\mathrm{ext}}=0\buildrel{\circ}\over{.} 057)$	$({r}_{\mathrm{ext}}=0\buildrel{\circ}\over{.} 1)$	$({r}_{\mathrm{ext}}={r}_{{\rm{h}}})$
DES-J0335.6–5403 (Ret II)	−3.6	32	0.100	1.00	1.00	1.00	1.00
DES-J0344.3–4331 (Eri II)	−7.4	330	0.030	1.00	1.00	1.00	1.00
DES-J2251.2–5836 (Tuc II)	−3.9	58	0.120	0.25	0.98	1.00	1.00
DES-J0255.4–5406 (Hor I)	−3.5	87	0.040	0.62	0.78	0.55	0.78
DES-J2108.8–5109 (Ind I)	−2.2	69	0.010	0.96	0.69	0.18	0.97
DES-J0443.8–5017 (Pic I)	−3.7	126	0.020	0.92	0.74	0.30	0.89
DES-J2339.9–5424 (Phe II)	−3.7	95	0.020	1.00	0.96	0.74	0.99
DES-J0222.7–5217 (Eri III)	−2.4	95	0.007	0.24	0.06	0.00	0.28
Segue 1	−1.5	23	0.073	0.72	0.99	0.99	0.99
Ursa Major II	−4.2	32	0.267	0.06	0.97	1.00	1.00
Bootes II	−2.7	42	0.070	0.94	1.00	1.00	1.00
Segue 2	−2.5	35	0.057	1.00	1.00	1.00	1.00
Willman 1	−2.7	38	0.038	1.00	1.00	1.00	1.00
Coma Berenices	−4.1	44	0.100	1.00	1.00	1.00	1.00
Bootes III	−5.8	47	1.666	0.00	0.00	0.00	0.96
Bootes I	−6.3	66	0.210	1.00	1.00	1.00	1.00
Sextans	−9.3	86	0.463	1.00	1.00	1.00	1.00
Ursa Major I	−5.5	97	0.188	0.00	0.30	0.90	0.98
Hercules	−6.6	132	0.143	0.86	1.00	1.00	1.00
Leo IV	−5.8	154	0.077	1.00	1.00	1.00	1.00
Canes Venatici II	−4.9	160	0.027	1.00	1.00	1.00	1.00
Leo V	−5.2	178	0.043	1.00	1.00	1.00	1.00
Pisces II	−5.0	182	0.018	1.00	1.00	1.00	1.00
Canes Venatici I	−8.6	218	0.148	1.00	1.00	1.00	1.00

Note. Detection efficiencies are calculated from many realizations of satellites with the properties (luminosity M_V, distance, Plummer profile angular half-light radius r_h) of a given ultra-faint galaxy/candidate as they would have been observed in DES Y1A1. The simulated satellites are uniformly distributed throughout the SPT region of the Y1A1 footprint, excluding regions of high stellar density near to the LMC, i.e., $\sim 1600\;{\mathrm{deg}}^{2}$. The rightmost columns list the detection efficiencies for extraction regions of different radii, ${r}_{\mathrm{ext}}$. Here, a detection constitutes $\gt 5\sigma$ stellar excess with $g\lt 23$ within the extraction region given the local density of the stellar field, after selecting stars that are consistent with the isochrone of an old and metal-poor stellar population at the satellite distance (i.e., following the map-based detection algorithm described in Section 3.1). The extraction region radii are choosen to reflect size scales used in the map-based search (${r}_{\mathrm{ext}}=\{0\buildrel{\circ}\over{.} 029,0\buildrel{\circ}\over{.} 057\}$), likelihood scan (${r}_{\mathrm{ext}}=0\buildrel{\circ}\over{.} 1$; Section 3.2), and matched to the size of the satellite (${r}_{\mathrm{ext}}={r}_{{\rm{h}}}$). Data for previously known satellites are taken from references compiled by McConnachie (2012b).

Download table as:  ASCIITypeset image

Table 3 also shows that the detection efficiency is sensitive to the size of the extraction region. Extended systems such as DES J2251.2−5836 are unlikely to found using the smallest extraction regions considered here, whereas the reverse is true for compact systems such as DES J0222.7−5217 and DES J2108.8−5109. This size dependence accounts for the low significance of the two most compact candidates, DES J0222.7−5217 and DES J2108.8−5109, in the likelihood scan (Table 1). After allowing their spatial extensions to be fit, the detection significances of these candidates increase to a level well above the our imposed threshold.

For comparison, Table 3 also provides detection efficiency estimates for previously known ultra-faint galaxies (assuming they were located in the SPT region of Y1A1 instead of their actual locations). We find that all of the SDSS ultra-faint galaxies, with the exception of the highly extended Boötes III, could have been readily detected in Y1A1. We attribute these high detection efficiencies to the deeper imaging of DES relative to SDSS and note that DES J2251.2−5836, DES J0255.4−5406, DES J2108.8−5109, and DES J0222.7−5217 have a substantially reduced detection probability when the magnitude limit is raised to $r\lt 22\;\mathrm{mag}$, comparable to the stellar completeness limit of SDSS.

Our Y1A1 search sensitivity can be quantified in a more general way by considering an ensemble of satellites spanning a range of luminosities, physical sizes, and heliocentric distances. Figure 12 presents the discovery potential of our Y1A1 search expressed as the detection efficiency with respect to these galaxy properties, estimated by the same method described above with ${r}_{\mathrm{ext}}={r}_{{\rm{h}}}$. Nearby, luminous, and compact objects have a high probability of being significantly detected whereas objects that are more distant, faint, and extended are less likely to be found. The detection threshold in the plane of physical size and luminosity is nearly parallel to contours of constant surface brightness, and is weakly dependent on the distance, provided that a sufficient number of stars are detected.

Zoom In Zoom Out Reset image size

Since the Y1A1 search procedure described in Section 3 is a combination of the map-based and likelihood-based search techniques, the actual completeness of our search is likely slightly higher than estimated here. We expect the likelihood method to be more sensitive to extended low surface brightness systems because it combines spatial and color–magnitude information simultaneously. As the depth of DES imaging increases (2 to 4 tilings in Y1A1 compared to 10 tilings planned after 5 years) and more advanced techniques are applied to separate stars and galaxies at faint magnitudes (e.g., Fadely et al. 2012; Soumagnac et al. 2013), we anticipate that lower surface brightness satellites will become accessible. Our present study is optimized for the detection of relatively compact (${r}_{0}\lesssim 0\buildrel{\circ}\over{.} 2$) and radially symmetric stellar over-densities. The search for extended low surface brightness features in the stellar distribution will be the focus of future work.

The discovery of eight new dwarf galaxy candidates in $\sim 1600\;{\mathrm{deg}}^{2}$ of Y1A1 not overlapping with SDSS Stripe 82 is consistent with expectations from the literature (Tollerud et al. 2008; Rossetto et al. 2011; Hargis et al. 2014; He et al. 2015). By empirically modeling the incompleteness of SDSS, Tollerud et al. 2008 predicted that 19–37 satellite galaxies could be found over the full DES footprint. More recent estimates based on high-resolution N-body simulations (Hargis et al. 2014) and semi-analytic galaxy formation models that include baryonic physics (He et al. 2015) predict $\sim 10$ new detectable satellite galaxies in DES. Large uncertainties are associated with each of these estimates due to weak constraints on the luminosity function in the ultra-faint regime. Additionally, as noted in Section 5.1, some of the DES candidates may be globular clusters or may be associated with the Magellanic Clouds. In the latter case, it becomes more challenging to directly compare our results to the predictions above, which assume an isotropic distribution of Milky Way satellite galaxies.

A number of studies, beginning with Lynden-Bell (1976), note that many Milky Way satellite galaxies appear to be distributed on the sky along a great circle, indicating a planar three-dimensional structure rather than an ellipsoidal or spherical distribution. This great circle has a polar orientation relative to the disk of the Milky Way. The discovery of most of the SDSS ultra-faint dwarfs in the north Galactic cap region increased the apparent significance of this alignment. However, since the primary region surveyed by SDSS is located in the direction of this so-called vast polar structure (Pawlowski et al. 2012), the true anisotropy of the Milky Way satellite population is not yet clear. The next generation of deep wide-field surveys should be able to address this issue with wider sky coverage.

In this context, it is interesting to consider the locations of the eight new satellites reported here, which may increase the known Milky Way dwarf galaxy population by $\sim 30\%$. The thickness of the vast polar structure defined by Pawlowski et al. (2012) is 29 kpc, and we find that the DES satellites have a dispersion of $\sim 28\;\mathrm{kpc}$ from this plane. This result is perhaps not surprising given their proximity to the Magellanic Clouds, which played a large role in defining the original Lynden-Bell plane. In fact, the entire Y1A1 search area (with the exception of Stripe 82) is located quite close to the previously known plane of satellites. Thus, any satellite galaxies identified in this data set are necessarily close to the plane, and a selection of eight random positions within this area would likely have a similar dispersion relative to the polar structure. A more quantitative characterization of the distribution of Milky Way satellites awaits the completion of the DES survey, including areas farther away from the vast polar structure, as well as future results from Pan-STARRS.