Ultra-faint Dwarf Research Articles

THE ORIGIN OF THE HEAVIEST METALS IN MOST ULTRA-FAINT DWARF GALAXIES

ABSTRACT The heaviest metals found in stars in most ultra-faint dwarf (UFD) galaxies in the Milky Way halo are generally underabundant by an order of magnitude or more when compared with stars in the halo field. Among the heavy elements produced by n-capture reactions, only Sr and Ba can be detected in red giant stars in most UFD galaxies. This limited chemical information is unable to identify the nucleosynthesis process(es) responsible for producing the heavy elements in UFD galaxies. Similar [Sr/Ba] and [Ba/Fe] ratios are found in three bright halo field stars, BD−18°5550, CS 22185–007, and CS 22891–200. Previous studies of high-quality spectra of these stars report detections of additional n-capture elements, including Eu. The [Eu/Ba] ratios in these stars span +0.41 to +0.86. These ratios and others among elements in the rare Earth domain indicate an r-process origin. These stars have some of the lowest levels of r-process enhancement known, with [Eu/H] spanning −3.95 to −3.32, and they may be considered nearby proxies for faint stars in UFD galaxies. Direct confirmation, however, must await future observations of additional heavy elements in stars in the UFD galaxies themselves.

Discovery of the most metal-poor damped Lyman-α system

We report the discovery and analysis of the most metal-poor damped Lyman-alpha (DLA) system currently known, based on observations made with the Keck HIRES spectrograph. The metal paucity of this system has only permitted the determination of three element abundances: [C/H] = -3.43 +/- 0.06, [O/H] = -3.05 +/- 0.05, and [Si/H] = -3.21 +/- 0.05, as well as an upper limit on the abundance of iron: [Fe/H] < -2.81. This DLA is among the most carbon-poor environment currently known with detectable metals. By comparing the abundance pattern of this DLA to detailed models of metal-free nucleosynthesis, we find that the chemistry of the gas is consistent with the yields of a 20.5 M_sun metal-free star that ended its life as a core-collapse supernova; the abundances we measure are inconsistent with the yields of pair-instability supernovae. Such a tight constraint on the mass of the progenitor Population III star is afforded by the well-determined C/O ratio, which we show depends almost monotonically on the progenitor mass when the kinetic energy of the supernova explosion is E_exp > 1.5x10^51 erg. We find that the DLA presented here has just crossed the critical 'transition discriminant' threshold, rendering the DLA gas now suitable for low mass star formation. We also discuss the chemistry of this system in the context of recent models that suggest some of the most metal-poor DLAs are the precursors of the 'first galaxies', and are the antecedents of the ultra-faint dwarf galaxies.

The contribution of dissolving star clusters to the population of ultra faint objects in the outer halo of the Milky Way

In the last decade, several ultra faint objects (UFOs, $M_V\gtrsim-3.5$) have been discovered in the outer halo of the Milky Way. For some of these objects it is not clear whether they are star clusters or (ultra-faint) dwarf galaxies. In this work we quantify the contribution of star clusters to the population of UFOs. We extrapolated the mass and Galactocentric radius distribution of the globular clusters using a population model, finding that the Milky Way contains about $3.3^{+7.3}_{-1.6}$ star clusters with $M_V\gtrsim-3.5$ and Galactocentric radius $\geq20\,$kpc. To understand whether dissolving clusters can appear as UFOs, we run a suite of direct $N$-body models, varying the orbit, the Galactic potential, the binary fraction and the black hole (BH) natal kick velocities. In the analyses, we consider observational biases such as: luminosity limit, field stars, and line-of-sight projection. We find that star clusters contribute to both the compact and the extended population of UFOs: clusters without BHs appear compact with radii $\sim5\,$pc, while clusters that retain their BHs after formation have radii $\gtrsim20\,$pc. The properties of the extended clusters are remarkably similar to those of dwarf galaxies: high inferred mass-to-light ratios due to binaries; binary properties mildly affected by dynamical evolution; no observable mass segregation; and flattened stellar mass function. We conclude that the slope of the stellar mass function as a function of Galactocentric radius and the presence/absence of cold streams can discriminate between DM free and DM dominated UFOs.

Gemini/GRACES spectroscopy of stars in Tri II

The chemical abundance ratios and radial velocities for two stars in the recently discovered Triangulum II faint dwarf galaxy have been determined from high resolution, medium signal-to-noise ratio spectra from the Gemini-GRACES facility. These stars have stellar parameters and metallicities similar to those derived from their photometry and medium-resolution Ca II triplet spectra, and supports that Triangulum II has a metallicity spread consistent with chemical evolution in a dwarf galaxy. The elemental abundances show that both stars have typical calcium abundances and barium upper limits for their metallicities, but low magnesium and sodium. This chemical composition resembles some stars in dwarf galaxies, attributed to inhomogeneous mixing in a low star formation environment, and/or yields from only a few supernova events. One of our targets (Star40) has an enhancement in potassium, and resembles some stars in the unusual outer halo star cluster, NGC 2419. Our other target (Star46) appears to be a binary based on a change in its radial velocity (Delta v(rad) = 24.5 +/- 2.1 km/s). This is consistent with variations found in binary stars in other dwarf galaxies. While this serves as a reminder of the high binary fraction in these ultra faint dwarf galaxies, this particular object has had little impact on the previous determination of the velocity dispersion in Triangulum II.

R-PROCESS PRODUCTION SITES AS INFERRED FROM Eu ABUNDANCES IN DWARF GALAXIES

ABSTRACT Recent observations of r-process material in ultrafaint dwarf galaxies (UFDs) shed light on the sources of these elements. Strong upper limits on the Eu mass in some UFDs, combined with detections of much larger masses in a UFD, Reticulum II, and other dwarf galaxies, imply that Eu production is dominated by rare events, and that the minimal Eu mass observed in any UFD is approximately the amount of Eu mass produced per event. This is consistent with other independent observations in the Galaxy. We estimate, using a model-independent likelihood analysis, the rate and Eu (Fe) mass produced per r-process (Fe production) event in dwarf galaxies, including classical dwarfs and UFDs. The mass and rate of the Fe production events are consistent with the normal core-collapse supernova (CCSN) scenario. The Eu mass per event is , corresponding to a total r-process mass per event of . The rate of r-process events is 2.5 × 10−4 &lt; R rp/SN &lt; 1.4 × 10−3 as compared with the CCSN rate. These values are consistent with the total Eu mass observed in our own Galaxy, suggesting that the same mechanism is behind the production of r-process events in both dwarf galaxies and the Milky Way, and that it may be the dominant mechanism for production of r-process elements in the universe. The results are consistent with neutron star merger estimates but cannot rule out other rare core-collapse scenarios, provided that they produce significant amounts of r-process material per event.

ON THE RR LYRAE STARS IN GLOBULARS. IV. ω CENTAURI OPTICAL UBVRI PHOTOMETRY*

ABSTRACT New accurate and homogeneous optical UBVRI photometry has been obtained for variable stars in the Galactic globular cluster ω Cen (NGC 5139). We secured 8202 CCD images covering a time interval of 24 years and a sky area of 84 × 48 arcmin. The current data were complemented with data available in the literature and provided new, homogeneous pulsation parameters (mean magnitudes, luminosity amplitudes, periods) for 187 candidate ω Cen RR Lyrae (RRLs). Among them we have 101 RRc (first overtone) and 85 RRab (fundamental) variables, and a single candidate RRd (double-mode) variable. Candidate Blazhko RRLs show periods and colors that are intermediate between the RRc and RRab variables, suggesting that they are transitional objects. A comparison of the period distribution and the Bailey diagram indicates that RRLs in ω Cen show a long-period tail not present in typical Oosterhoff II (OoII) globulars. The RRLs in dwarf spheroidals and in ultra-faint dwarfs have properties between Oosterhoff intermediate and OoII clusters. Metallicity plays a key role in shaping the above evidence. These findings do not support the hypothesis that ω Cen is the core remnant of a spoiled dwarf galaxy. Using optical period–Wesenheit relations that are reddening-free and minimally dependent on metallicity we find a mean distance to ω Cen of 13.71 ± 0.08 ± 0.01 mag (semi-empirical and theoretical calibrations). Finally, we invert the I-band period–luminosity–metallicity relation to estimate individual RRLs’ metal abundances. The metallicity distribution agrees quite well with spectroscopic and photometric metallicity estimates available in the literature.

A NEW MILKY WAY SATELLITE DISCOVERED IN THE SUBARU/HYPER SUPRIME-CAM SURVEY

ABSTRACT We report the discovery of a new ultra-faint dwarf satellite companion of the Milky Way (MW) based on the early survey data from the Hyper Suprime-Cam Subaru Strategic Program. This new satellite, Virgo I, which is located in the constellation of Virgo, has been identified as a statistically significant (5.5σ) spatial overdensity of star-like objects with a well-defined main sequence and red giant branch in the color–magnitude diagram. The significance of this overdensity increases to 10.8σ when the relevant isochrone filter is adopted for the search. Based on the distribution of the stars around the likely main-sequence turnoff at r ∼ 24 mag, the distance to Virgo I is estimated as 87 kpc, and its most likely absolute magnitude calculated from a Monte Carlo analysis is M V = −0.8 ± 0.9 mag. This stellar system has an extended spatial distribution with a half-light radius of pc, which clearly distinguishes it from a globular cluster with comparable luminosity. Thus, Virgo I is one of the faintest dwarf satellites known and is located beyond the reach of the Sloan Digital Sky Survey. This demonstrates the power of this survey program to identify very faint dwarf satellites. This discovery of Virgo I is based only on about 100 square degrees of data, thus a large number of faint dwarf satellites are likely to exist in the outer halo of the MW.

CHEMICAL DIVERSITY IN THE ULTRA-FAINT DWARF GALAXY TUCANA II*

ABSTRACT We present the first detailed chemical abundance study of the ultra-faint dwarf galaxy Tucana II, based on high-resolution Magellan/MIKE spectra of four red giant stars. The metallicities of these stars range from [Fe/H] = −3.2 to −2.6, and all stars are low in neutron-capture abundances ([Sr/Fe] and [Ba/Fe] &lt; −1). However, a number of anomalous chemical signatures are present. One star is relatively metal-rich ([Fe/H] = −2.6) and shows [Na, α, Sc/Fe] &lt; 0, suggesting an extended star formation history with contributions from AGB stars and SNe Ia. Two stars with [Fe/H] &lt; −3 are mildly carbon-enhanced ([C/Fe] ∼ 0.7) and may be consistent with enrichment by faint supernovae, if such supernovae can produce neutron-capture elements. A fourth star with [Fe/H] = −3 is carbon-normal, and exhibits distinct light element abundance ratios from the carbon-enhanced stars. This carbon-normal star implies that at least two distinct nucleosynthesis sources, both possibly associated with Population III stars, contributed to the early chemical enrichment of this galaxy. Despite its very low luminosity, Tucana II shows a diversity of chemical signatures that preclude it from being a simple “one-shot” first galaxy yet still provide a window into star and galaxy formation in the early universe.

A COMMON ORIGIN FOR GLOBULAR CLUSTERS AND ULTRA-FAINT DWARFS IN SIMULATIONS OF THE FIRST GALAXIES

ABSTRACT In this paper, the first in a series on galaxy formation before reionization, we focus on understanding what determines the size and morphology of stellar objects in the first low-mass galaxies, using parsec-scale cosmological simulations performed with an adaptive mesh hydrodynamics code. Although the dense gas in which stars are formed tends to have a disk structure, stars are found in spheroids with little rotation. Halos with masses between and form stars stochastically, with stellar masses in the range to . We observe, nearly independent of stellar mass, a large range of half-light radii for the stars, from a few parsecs to a few hundred parsecs and surface brightnesses and mass-to-light ratios ranging from those typical of globular clusters to ultra-faint dwarfs. In our simulations, stars form in dense stellar clusters with high gas-to-star conversion efficiencies and rather uniform metallicities. A fraction of these clusters remain bound after the gas is removed by feedback, but others are destroyed, and their stars, which typically have velocity dispersions of 20–40 km s−1, expand until they become bound by the dark matter halo. We thus speculate that the stars in ultra-faint dwarf galaxies may show kinematic and chemical signatures consistent with their origin in a few distinct stellar clusters. On the other hand, some globular clusters may form at the center of primordial dwarf galaxies and may contain dark matter, perhaps detectable in the outer parts.

The clustering of massive Primordial Black Holes as Dark Matter: Measuring their mass distribution with advanced LIGO

The recent detection by Advanced LIGO of gravitational waves (GW) from the merging of a binary black hole system sets new limits on the merging rates of massive primordial black holes (PBH) that could be a significant fraction or even the totality of the dark matter in the Universe. aLIGO opens the way to the determination of the distribution and clustering of such massive PBH. If PBH clusters have a similar density to the one observed in ultra-faint dwarf galaxies, we find merging rates comparable to aLIGO expectations. Massive PBH dark matter predicts the existence of thousands of those dwarf galaxies where star formation is unlikely because of gas accretion onto PBH, which would possibly provide a solution to the missing satellite and too-big-to-fail problems. Finally, we study the possibility of using aLIGO and future GW antennas to measure the abundance and mass distribution of PBH in the range [5 - 200] Msun to 10\% accuracy.

COMPLETE ELEMENT ABUNDANCES OF NINE STARS IN THE r-PROCESS GALAXY RETICULUM II*

ABSTRACT We present chemical abundances derived from high-resolution Magellan/Magellan Inamori Kyocera Echelle spectra of the nine brightest known red giant members of the ultra-faint dwarf galaxy Reticulum II (Ret II). These stars span the full metallicity range of Ret II (−3.5 &lt; [Fe/H] &lt; −2). Seven of the nine stars have extremely high levels of r-process material ([Eu/Fe] ∼ 1.7), in contrast to the extremely low neutron-capture element abundances found in every other ultra-faint dwarf galaxy studied to date. The other two stars are the most metal-poor stars in the system ([Fe/H] &lt; −3), and they have neutron-capture element abundance limits similar to those in other ultra-faint dwarf galaxies. We confirm that the relative abundances of Sr, Y, and Zr in these stars are similar to those found in r-process halo stars, but they are ∼0.5 dex lower than the solar r-process pattern. If the universal r-process pattern extends to those elements, the stars in Ret II display the least contaminated known r-process pattern. The abundances of lighter elements up to the iron peak are otherwise similar to abundances of stars in the halo and in other ultra-faint dwarf galaxies. However, the scatter in abundance ratios is large enough to suggest that inhomogeneous metal mixing is required to explain the chemical evolution of this galaxy. The presence of low amounts of neutron-capture elements in other ultra-faint dwarf galaxies may imply the existence of additional r-process sites besides the source of r-process elements in Ret II. Galaxies like Ret II may be the original birth sites of r-process enhanced stars now found in the halo.

Organized chaos: scatter in the relation between stellar mass and halo mass in small galaxies

We use Local Group galaxy counts together with the ELVIS N-body simulations to explore the relationship between the scatter and slope in the stellar mass vs. halo mass relation at low masses, $M_\star \simeq 10^5 - 10^8 M_\odot$. Assuming models with log-normal scatter about a median relation of the form $M_\star \propto M_\mathrm{halo}^\alpha$, the preferred log-slope steepens from $\alpha \simeq 1.8$ in the limit of zero scatter to $\alpha \simeq 2.6$ in the case of $2$ dex of scatter in $M_\star$ at fixed halo mass. We provide fitting functions for the best-fit relations as a function of scatter, including cases where the relation becomes increasingly stochastic with decreasing mass. We show that if the scatter at fixed halo mass is large enough ($\gtrsim 1$ dex) and if the median relation is steep enough ($\alpha \gtrsim 2$), then the "too-big-to-fail" problem seen in the Local Group can be self-consistently eliminated in about $\sim 5-10\%$ of realizations. This scenario requires that the most massive subhalos host unobservable ultra-faint dwarfs fairly often; we discuss potentially observable signatures of these systems. Finally, we compare our derived constraints to recent high-resolution simulations of dwarf galaxy formation in the literature. Though simulation-to-simulation scatter in $M_\star$ at fixed $M_\mathrm{halo}$ is large among separate authors ($\sim 2$ dex), individual codes produce relations with much less scatter and usually give relations that would over-produce local galaxy counts.

Ultra faint dwarf galaxies: an arena for testing dark matter versus modified gravity

The scenario consistent with a wealth of observations for the missing mass problem is that of weakly interacting dark matter particles. However, arguments or proposals for a Newtonian or relativistic modified gravity scenario continue to be made. A distinguishing characteristic between the two scenarios is that dark matter particles can produce a gravitational effect, in principle, without the need of baryons while this is not the case for the modified gravity scenario where such an effect must be correlated with the amount of baryonic matter. We consider here ultra-faint dwarf (UFD) galaxies as a promising arena to test the two scenarios based on the above assertion. We compare the correlation of the luminosity with the velocity dispersion between samples of UFD and non-UFD galaxies, finding a significant loss of correlation for UFD galaxies. For example, we find for 28 non-UFD galaxies a strong correlation coefficient of −0.688 which drops to −0.077 for the 23 UFD galaxies. Incoming and future data will determine whether the observed stochasticity for UFD galaxies is physical or due to systematics in the data. Such a loss of correlation (if it is to persist) is possible and consistent with the dark matter scenario for UFD galaxies but would constitute a new challenge for the modified gravity scenario.

NATAL KICKS AND TIME DELAYS IN MERGING NEUTRON STAR BINARIES: IMPLICATIONS FOR r-PROCESS NUCLEOSYNTHESIS IN ULTRA-FAINT DWARFS AND IN THE MILKY WAY

ABSTRACT Merging neutron star binaries are prime candidate sources for heavy r-process nucleosynthesis. The amount of heavy r-process material is consistent with the mass ejection and rates of mergers, and abundances of relic radioactive materials suggest that heavy r-process material is produced in rare events. Observations of possible macronovae provide further support for this model. Still, some concerns remain. One is the observation of heavy r-process elements in ultra-faint dwarf (UFD) galaxies. The escape velocities from UFDs are so small that the natal kicks, taking place at neutron stars’ birth, might eject such binaries from UFDs. Furthermore, the old stellar populations of UFDs require that r-process nucleosynthesis must have taken place very early on, while it may take several Gyr for compact binaries to merge. This last problem arises also within the Milky Way where heavy r-process materials have been observed in some low-metallicity stars. We show here that of neutron star binaries form with a sufficiently small proper motion to remain bound even in a UFD. Furthermore, approximately 90% of double neutron stars with an initial separation of 1011 cm merge within 300 Myr and merge in less than 100 Myr. This population of “rapid mergers” explains the appearance of heavy r-process material in both UFDs and in the early Milky Way.

The potential of the dwarf galaxy Triangulum II for dark matter indirect detection

The recently discovered object Triangulum II appears to be an ultra faint dwarf spheroidal galaxy which may be one of the most dark matter dominated objects yet known. In this work we try to estimate the potential of this object for studies of the indirect detection of self-annihilating dark matter by obtaining its astrophysical J-factor. We perform a basic estimate of the velocity gradient to look for signs of the halo being tidally disrupted but show that the observed value is statistically compatible with zero velocity gradient. We solve the spherical Jeans equation using Markov Chain Monte Carlo (MCMC) engine GreAT and the Jeans analysis part of the CLUMPY package. We find the results point towards a very large J-factor, appearing to make Triangulum II one of the best targets in the search for dark matter. However we stress that the very small number of line of sight velocities currently available for this object make follow up studies essential.

Theoretical lower limits on sizes of ultrafaint dwarf galaxies from dynamical friction

Dwarf spheroidal galaxies are the smallest known stellar systems where under Newtonian interpretations, a significant amount of dark matter is required to explain observed kinematics. In fact, they are in this sense the most heavily dark matter dominated objects known. That, plus the increasingly small sizes of the newly discovered ultra faint dwarfs, puts these systems in the regime where dynamical friction on individual stars starts to become relevant. We calculate the dynamical friction timescales for pressure supported isotropic spherical dark matter dominated stellar systems, yielding $\tau_{DF} =0.93 (r_{h}/10 pc)^{2} (\sigma/ kms^{-1}) Gyr$, { where $r_{h}$ is the half-light radius}. For a stellar velocity dispersion value of $3 km/s$, as typical for the smallest of the recently detected ultra faint dwarf spheroidals, dynamical friction timescales becomes smaller than the $10 Gyr$ typical of the stellar ages for these systems, for $r_{h}<19 pc$. Thus, this becomes a theoretical lower limit below which dark matter dominated stellar systems become unstable to dynamical friction. We present a comparison with structural parameters of the smallest ultra faint dwarf spheroidals known, showing that these are already close to the stability limit derived, any future detection of yet smaller such systems would be inconsistent with a particle dark matter hypothesis.

THE CHEMICAL EVOLUTION OF THE BOÖTES I ULTRA-FAINT DWARF GALAXY*

ABSTRACT We present chemical abundance measurements of two metal-poor red giant stars in the ultra-faint dwarf galaxy Boötes I, based on Magellan/MIKE high-resolution spectra. For Boo-980, with , we present the first elemental abundance measurements, while Boo-127, with , shows abundances in good agreement with previous measurements. Light and iron-peak element abundance ratios in the two Boötes I stars, as well as those of most other Boötes I members, collected from the literature, closely resemble those of regular metal-poor halo stars. Neutron-capture element abundances Sr and Ba are systematically lower than the main halo trend and also show a significant abundance spread. Overall, this is similar to what has been found for other ultra-faint dwarf galaxies. We apply corrections to the carbon abundances (commensurate with stellar evolutionary status) of the entire sample and find 21% of stars to be carbon-enhanced metal-poor (CEMP) stars, compared to 13% without using the carbon correction. We reassess the metallicity distribution functions for the CEMP stars and non-CEMP stars, and confirm earlier claims that CEMP stars might belong to a different, earlier population. Applying a set of abundance criteria to test to what extent Boötes I could be a surviving first galaxy suggests that it is one of the earliest assembled systems that perhaps received gas from accretion from other clouds in the system, or from swallowing a first galaxy or building block type object. This resulted in the two stellar populations observable today.

ON THEr-PROCESS ENRICHMENT OF DWARF SPHEROIDAL GALAXIES

Recent observations of Reticulum II have uncovered an overabundance of r-process elements, compared to similar ultra-faint dwarf spheroidal galaxies (UFDs). Because the metallicity and star formation history of Reticulum II appear consistent with all known UFDs, the high r-process abundance of Reticulum II suggests enrichment through a single, rare event, such as a double neutron star (NS) merger. However, we note that this scenario is extremely unlikely, as binary stellar evolution models require significant supernova natal kicks to produce NS-NS or NS-black hole mergers, and these kicks would efficiently remove compact binary systems from the weak gravitational potentials of UFDs. We examine alternative mechanisms for the production of r-process elements in UFDs, including a novel mechanism wherein NSs in regions of high dark matter density implode after accumulating a black-hole-forming mass of dark matter. We find that r-process proto-material ejection by tidal forces, when a single neutron star implodes into a black hole, can occur at a rate matching the r-process abundance of both Reticulum II and the Milky Way. Remarkably, dark matter models which collapse a single neutron star in observed UFDs also solve the missing pulsar problem in the Milky Way Galactic center. We propose tests specific to dark matter r-process production which may uncover, or rule out, this model.

Contamination of stellar-kinematic samples and uncertainty about dark matter annihilation profiles in ultrafaint dwarf galaxies: the example of Segue I

The expected gamma-ray flux coming from dark matter annihilation in dwarf spheroidal (dSph) galaxies depends on the so-called `J-factor', the integral of the squared dark matter density along the line-of-sight. We examine the degree to which estimates of J are sensitive to contamination (by foreground Milky Way stars and stellar streams) of the stellar-kinematic samples that are used to infer dark matter densities in `ultrafaint' dSphs. Applying standard kinematic analyses to hundreds of mock data sets that include varying levels of contamination, we find that mis-classified contaminants can cause J-factors to be overestimated by orders of magnitude. Stellar-kinematic data sets for which we obtain such biased estimates tend 1) to include relatively large fractions of stars with ambiguous membership status, and 2) to give estimates for J that are sensitive to specific choices about how to weight and/or to exclude stars with ambiguous status. Comparing publicly-available stellar-kinematic samples for the nearby dSphs Reticulum~II and Segue~I, we find that only the latter displays both of these characteristics. Estimates of Segue~I's J-factor should therefore be regarded with a larger degree of caution when planning and interpreting gamma-ray observations. Moreover, robust interpretations regarding dark matter annihilation in dSph galaxies in general will require explicit examination of how interlopers might affect the inferred dark matter density profile.

3FGL DEMOGRAPHICS OUTSIDE THE GALACTIC PLANE USING SUPERVISED MACHINE LEARNING: PULSAR AND DARK MATTER SUBHALO INTERPRETATIONS

ABSTRACT Nearly one-third of the sources listed in the Third Fermi Large Area Telescope (LAT) catalog (3FGL) remain unassociated. It is possible that predicted and even unanticipated gamma-ray source classes are present in these data waiting to be discovered. Taking advantage of the excellent spectral capabilities achieved by the Fermi LAT, we use machine-learning classifiers (Random Forest and XGBoost) to pinpoint potentially novel source classes in the unassociated 3FGL sample outside the Galactic plane. Here we report a total of 34 high-confidence Galactic candidates at . The currently favored standard astrophysical interpretations for these objects are pulsars or low-luminosity globular clusters hosting millisecond pulsars (MSPs). Yet these objects could also be interpreted as dark matter annihilation taking place in ultra-faint dwarf galaxies or dark matter subhalos. Unfortunately, Fermi LAT spectra are not sufficient to break degeneracies between the different scenarios. Careful visual inspection of archival optical images reveals no obvious evidence for low-luminosity globular clusters or ultra-faint dwarf galaxies inside the 95% error ellipses. If these are pulsars, this would bring the total number of MSPs at to 106, down to an energy flux ≈4.0 × 10−12 erg cm−2 s−1 between 100 MeV and 100 GeV. We find this number to be in excellent agreement with predictions from a new population synthesis of MSPs that predicts 100–126 high-latitude 3FGL MSPs depending on the choice of high-energy emission model. If, however, these are dark matter substructures, we can place upper limits on the number of Galactic subhalos surviving today and on dark matter annihilation cross sections. These limits are beginning to approach the canonical thermal relic cross section for dark matter particle masses below ∼100 GeV in the bottom quark ( ) annihilation channel.