Reticulum II: Gamma-rays, dark matter content, and implications for dark matter physics

Savvas M. Koushiappas

Based on PRL 115, 081101 (2015) 1503.02320 ApJL 808 L36 (2015) 1504.03309

With: Alex Geringer-Sameth & Matthew Walker (Carnegie-Mellon U.), Sergey Koposov, Vasily Belokurov, Gabriel Torrealba & Wyn Evans (Cambridge U.) Vincent Bonnivard, Celine Combet, David Maurin (U. Grenoble-Alpes), Mario Mateo, John Bailey (U. Michigan), Eduard Olszewski (U. Arizona) On March 8, 2015 arXiv:1503.02079 [pdf, ps, other] Beasts of the Southern Wild. Discovery of a large number of Ultra Faint satellites in the vicinity of the Magellanic Clouds Sergey E. Koposov, Vasily Belokurov, Gabriel Torrealba, N. Wyn Evans arXiv:1503.02584 [pdf, other] Eight New Milky Way Companions Discovered in First-Year Dark Energy Survey Data The DES Collaboration, K. Bechtol, A. Drlica-Wagner, E. Balbinot, A. Pieres, J. D. Simon, B. Yanny, B. Santiago, R. H. Wechsler, J. Frieman, A. R. Walker, P. Williams, E. Rozo, E. S. Rykof, A. Queiroz, E. Luque, A. Benoit-Levy, R. A. Bernstein, D. Tucker, I. Sevilla, R. A. Gruendl4, L. N. da Costa, A. Fausti Neto, M. A. G. Maia, T. Abbott, S. Allam, R. Armstrong, A. H. Bauer, G. M. Bernstein, E. Bertin, D. Brooks, E. Buckley-Geer, D. L. Burke, A. Carnero Rosell, F. J. Castander, C. B. D'Andrea, D. L. DePoy, S. Desai, H. T. Diehl, T. F. Eifler, J. Estrada, A. E. Evrard, E. Fernandez, D. A. Finley, B. Flaugher, E. Gaztanaga, D. Gerdes, L. Girardi, M. Gladders, D. Gruen, G. Gutierrez, J. Hao, K. Honscheid, B. Jain, D. James, S. Kent, R. Kron, K. Kuehn, N. Kuropatkin, O. Lahav, T. S. Li, et al. (32 additional authors not shown) Table 1 Parameters of the discovered MW satellites.

Name αδSignif m−MDist⊙ MV rmaj r1/2 r1/2 ePABF(ell) [deg] [deg] [mag] [kpc] [mag] [arcmin] [arcmin] [pc] [deg] +0.03 Reticulum 2 53.9256 −54.0492 48.5 17.4 30 -2.7±0.1 3.4±0.2 3.7±0.2 32±10.58−0.03 71±1 >1000 +0.07 Eridanus 2 56.0878 −43.5332 31.5 22.9 380 -6.6±0.1 1.2±0.1 1.6±0.1 172±12 0.39−0.07 80±61113 +0.12 Horologium 1 43.8820 −54.1188 28.4 19.5 79 -3.4±0.1 0.9±0.1 1.3±0.2 30±30.16−0.12 55±50 0.35 +0.19 Pictoris 1 70.9475 −50.2830 17.3 20.3 114 -3.1±0.3 0.7±0.2 0.9±0.2 31±70.39−0.22 79±23 1.41 +0.18 Phoenix 2 354.9975 −54.4060 13.9 19.6 83 -2.8±0.2 0.9±0.2 1.1±0.2 27±50.38−0.19 150±54 1.81 +0.16 Indus 1 317.2044 −51.1656 13.7 20.0 100 -3.5±0.2 0.9±0.3 1.4±0.4 39±11 0.22−0.16 82±50 0.46 a +0.24 Grus 1 344.1765 −50.1633 10.1 20.4 120 -3.4±0.3 1.6±0.6 2.0±0.7 70±23 0.37−0.25 48±60 1.01 +0.23 Eridanus 3 35.6897 −52.2837 10.1 19.7 87 -2.0±0.3 0.6±0.3 0.7±0.3 18±80.34−0.23 89±36 0.81 +0.16 Tucana 2 342.9664 −58.5683 8.3 19.2 69 -4.4±0.1 7.3±1.2 9.9±1.4 199±28 0.31−0.17 106±22 1.29

aAs this object is located very close to the CCD chip gap,’+ its morphological properties should be treated with caution PSF magnitudes of star-like objects are given by the SDSS gri. Consequently, the extinction coefficients MAG_PSF output of SExtractor. As an indicator of star- used are those suitable for the SDSS photometric sys- galaxy separation we use the SPREAD_MODEL parame- tem, while the dust reddening maps employed are from ter provided by SExtractor. This is a metric simi- Schlegel et al. (1998). Note that the depth of the result- lar to psfmag-modelmag used by SDSS (see Fig. 1). ing catalogues varies somewhat across the DES footprint, A sensible selection threshold for bright stars would but could be approximately estimated from the source be SPREAD_MODEL < 0.003 (Annunziatella et al. 2013), number counts in g, r, i filters. These number counts however| for faint magnitudes| this cut causes significant peak at magnitudes 23.7, 23.6, 22.9 in g, r, i correspond- incompleteness in stars. Therefore, instead we choose to ingly, indicating that the catalogues start to be signifi- require:4 cantly affected by incompleteness at somewhat brigher magnitudes g 23.5, r 23.4, i 22.7. SPREAD_MODEL < 0.003 + SPREADERR_MODEL (1) ∼ ∼ ∼ | | To illustrate the quality of the resulting catalogue, Fig- This particular cut ensures that the stellar complete- ure 3 displays the density of the Main Sequence Turn- ness remains reasonably high at faint magnitudes, while Off (MSTO; 0.20.5) which are also classified as stars 3. SEARCH FOR STELLAR OVER-DENSITIES by our cuts applied to the DES data. Similarly, con- To uncover the locations of possible satellites lurking tamination can be gleaned from the fraction of objects in the DES data, we follow the approach described in classified as galaxies by the CFHTLS but as stars by our Koposov et al. (2008); Walsh et al. (2009). In short, the DES cuts (red dashed line). It is reassuring to observe satellite detection relies on applying a matched filter to low levels of contamination all the way to the very mag- the on-sky distributions of stars selected to correspond nitude limit of the DES survey. At the same time, com- to a single stellar population at a chosen distance. The pleteness is high across a wide range of magnitudes and matched filter is simply a difference of 2D Gaussians, only drops to 60% for objects fainter than r 22. It the broader one estimating the local background density, is also worth noting∼ that the star-galaxy separation∼ cri- while the narrow one yielding the amplitude of the den- teria employed in this work may not be ideally suited for sity peak at the location of the satellite. other studies, as they may have different requirements in We start by taking a catalogue of sources classified as terms of the balance between the completeness and the stars. A sub-set of these is then carved out with either a contamination. set of colour-magnitude cuts or with an isochrone mask In the stellar catalogues built using the procedure offset to a trial distance modulus. Then a 2D on-sky described above, the magnitudes are equivalent to the density map of the selected stars is constructed, keeping the spatial pixel sufficiently small, e.g. 1′ on a side. At 4 http://1.usa.gov/1zHCdrq the next step, the density map is convolved with a set Reticulum 2

nearest of the new DES dwarfs (30 kpc)

Reticulum II in gamma-rays Gamma-rays 1-300 GeV Gamma-ray background model at 8 GeV

40 1.6 [1-300] GeV 1.5

45 1.4

1.3 50 1.2

1.1 Relative intensity 55 Galactic latitude [Deg] 1.0

60 0.9 105 100 95 90 85 80 Galactic longitude [Deg]

BasedFar on PRLaway 115, 081101 from & ApJL known 808 L36 (2015)sources Uniform background Reticulum 2

nearest of the new DES dwarfs (30 kpc)

Reticulum II in gamma-rays Gamma-rays 1-300 GeV Gamma-ray background model at 8 GeV

1.6 [1-300] GeV 40 Intensity contrast map at 8 GeV 40 1.6 1.5 45 1.5 1.4 45 1.4 1.3 50 1.3 50 1.2 1.2

1.1 Relative intensity 55 1.1 Relative intensity 55 Galactic latitude [Deg] Galactic latitude [Deg] 1.0 1.0

60 60 0.9 0.9 105 100 95 90 85 80 105 100 95 90 85 80 Galactic longitudeGalactic [Deg] longitude [Deg]

BasedFar on PRLaway 115, 081101 from & ApJL known 808 L36 (2015)sources Uniform background Reticulum 2

nearest of the new DES dwarfs (30 kpc)

Reticulum II in gamma-rays Gamma-rays 1-300 GeV Gamma-ray background model at 8 GeV

1.6 [1-300] GeV 40 5 10

] 1.5 1 sr

1 45 1.4

s 2

6 1.3 10 50

[GeV cm 1.2 Relative intensity dF/dE 1.1 2 55 Galactic latitude [Deg] E 7 10 1.0 100 101 102 Energy [GeV] 60 0.9 105 100 95 90 85 80 Galactic longitude [Deg] Fermi-LAT isotropic diffuse background model

Sum of the two Fermi-LAT Galactic diffuse background model BasedFar on PRLaway 115, 081101 from & ApJL known 808 L36 (2015)sources Uniform background Reticulum 2

nearest of the new DES dwarfs (30 kpc)

Reticulum II in gamma-rays Gamma-rays 1-300 GeV Gamma-ray background model at 8 GeV

Empirical 1.6 [1-300] GeV 40 5 10 1.5 ] 1 sr

1 45 1.4 s 2

6 1.3 10 50

[GeV cm 1.2 Relative intensity

dF/dE 1.1 2 55 Galactic latitude [Deg] E 7 10 1.0 100 101 102 Energy [GeV] 60 0.9 105 100 95 90 85 80 Galactic longitude [Deg] Fermi-LAT isotropic diffuse background model

Sum of the two Fermi-LAT Galactic diffuse background model BasedFar on PRLaway 115, 081101 from & ApJL known 808 L36 (2015)sources Uniform background Reticulum 2

nearest of the new DES dwarfs (30 kpc)

Reticulum II in gamma-rays Gamma-rays 1-300 GeV Gamma-ray background model at 8 GeV A first look at the gamma-ray signal 1.6 [1-300] GeV 40 5 10 1.5 ] 1 1

10 6 sr

1 45 13390 18 11 1.4 1 s 51 33 22

2 135.4 96.3 59.2 Events within1.3 0.5° 6 33.5 10 19.9 11.5 50 6.5 3.9 of Ret2 2.3 1.3

[GeV cm 0.6 1.2 0.3 0.2 0.1 0.1 Relative intensity dF/dE 1.1 2 55 Galactic latitude [Deg] E 7 10 1.0 100 101 102 Energy [GeV] 60 0.9 105 100 95 90 85 80 Galactic longitude [Deg]

AGS et. al. arXiv:1503.02320

BasedFar on PRLaway 115, 081101 from & ApJL known 808 L36 (2015)sources Uniform background 5 5

Reticulum II in gamma-rays

A first look at the gamma-ray signal 5

Pass8 Fermi-LAT w/hybrid Pass8/7 background Pass7R5 5 5 10 ) Drlica-Wagner et al. ApJL 809 L4 ] 1 1 1 6

s 10 sr

2 DES J0335.6-5403 1 133 11

90 18 1 6 s 51 33 22

10 2 135.4 96.3 59.2 Events within 0.5° 6 33.5 10 19.9 11.5 6.5 3.9 of Ret2 2.3 1.3

7 [GeV cm 10 3 0.6 0.3 0.2 0.1 0.1 dF/dE

Hooper2 & Linden E 7 3 4 5 Energy Flux (MeV cm 10 10 10 1503.0620910 100 101 102 Energy (MeV)Energy (MeV) Energy [GeV]

FIG. 2. Bin-by-binFIG. 2. integrated Bin-by-bin energy-flux integrated upper energy-flux limits upper at 95% limits confidence at 95% confidence level for the level eight for the DES eight dSph DES candidates dSph candidates modeled modeled as as point-like sources.point-like sources. FIG. 2. Bin-by-bin integrated energy-flux upper limits at 95% confidence level for the eight DES dSph candidates modeled as point-like sources. Pass7R AGS et. al. arXiv:1503.02320 for Galactic di↵use emission derived from an all-sky fit tially extended Navarro, Frenk and White (NFW) DM for Galactic di↵use emission derived from an4 all-skyFIG. 2: fit The fractiontially of “blank extended sky” locations with Navarro, a test Frenk and White (NFW) DM to the Pass 7 Reprocessed4 data, butstatistic withfor (TS) aGalactic larger small than di( a< given↵use value, emissiondensity as empirically derived profiles deter- from [FIG.35 an]. 3: all-skyThis The log-likelihood choice fit fittially was of Reticulum extendedmotivated II in 24 Navarro, energy by the Frenk and White (NFW) DM to the Pass 7 Reprocessed data, but with a smallmined forto ( a< the collectionPassdensity of 1905 7 Reprocessed randomly profiles selected skydata, [35 locations].4 but Thisbins with spanning choice a small 500 was MeV (< to motivated 500density GeV. The profiles upper by limits the[35 corre-]. This choice was motivated by the 10%) energy-dependent correction to accountconstrained to for lie at di a galactic↵er- latitudecurrentb > 30 and uncertainty at least in the spatial extension of the DM | | spond to 2 confidence in each energy range. The white line 10%) energy-dependentences in the correctionPass 8 instrument to account response. for1 di(55)↵10%) fromAdditionally,er- point-like energy-dependentcurrent (extended) 3FGL uncertaintyhalos sources correction of [28]. these For the to in new account thecorresponds objects. spatial for to dithe↵ bester- Previous extension fit fromcurrent a 49 studies GeV of dark uncertainty the matter have DM particle shown in the spatial extension of the DM 5 blue curve,ences no additional in the requirementsPass 8 instrument are placed on the response. blank annihilating5 Additionally, to b¯b. halos of these new objects. Previous studies have shown ences in the Passwe model 8 instrument extragalactic response. gamma-rayAdditionally, emissionsky locations. and For residual thehalos red curve, of thethese blankthat sky the locations new LAT objects. used flux limits Previous are fairly studies insensitive have shown to model- FIG. 2. Bin-by-bin integrated energy-flux upper limits at 95% confidence level for0.5 the eight DES dSph candidates modeled as we model extragalactic gamma-ray emission and residualare additionallyFIG.we model 2. requiredthat Bin-by-binextragalactic to lie the no closer LAT gamma-ray integrated than fluxfrom limits any emission energy-flux are and fairly residual upper insensitive limitsthat the at to LAT95% model- flux confidence limits are level fairly for insensitive the eight to model- DES dSph candidates modeled as charged particle contamination with ansource isotropic listed in the model BZCAT, CRATES,ing CGRaBS, dSph or targetsATNF 6.5 as years point-like of data, adopt versus the default spatially energy range, extended and point-like sources. catalogspoint-likecharged (see Sec. IV). particle sources. The shaded contamination region surrounding with each an isotropic model ing dSph targets as point-like versus spatially extended charged particle contamination with an isotropic model ing dSph targets as point-liketest the comparison versus to spatiallya 49 GeV dark matter extended model an- fit to the Pass 8 data. These models willcurve representsbefit to included the the poissonPass in errors 8 data. onsources this These determination. [ models15]. In Following will be included the¯ procedure in sources of Ackermann [15]. Followinget the al. procedure of Ackermann et al. generating this figure, we have adopted a spectral shape cor- nihilating to bb (corresponding to the best-fit value of fit to the Passthe 8 forthcomingdata. These public modelsPass will 8 bedata included release.respondingthe in to forthcoming Point-like a 49 GeVsources dark matter public [15[ particle17].Pass], Following we annihilating 8fitdata for to excess release. thetheprocedure mass gamma-ray Point-like for the spectrum of emission Ackermann[17 of], the we Galactic fit associated for Center excesset al. ex- gamma-ray with emission associated with ¯ cess [9]). In Fig. 2 we show the resulting distribution of bb (correspondingsources to from the best-fit the recentmass for the 3FGL Galactic catalog Center [34] within 15 of each target in each energy bin separately to derive flux the forthcomingsources public fromPass the recent 8 data 3FGL release. catalog Point-like [34] within 15[17 ],of we fiteach for excesstarget in gamma-ray eachour blank-sky energy emission test bin locations. separately associated While the existence to derive with of sys- flux gamma-raythe excess ROI [9]). center were also included in the ROI model. constraints that are independent of the choice of spectral forsources Galactic from dithe↵ theuse ROI recent emission center 3FGL derived were catalog also from included [34] an within all-sky in 15the fit ROIof model.tiallyeach target extendedconstraints in each Navarro, that energytematic are Frenk bin independent errors separately and in the modeling White of the to of the (NFW) derivechoice gamma-ray of flux DM spectralback- forThe Galactic spectral parameters di↵use emission of these sources derivedground were drives from fixed this at distribution an all-skymodel. far from Within fit that expected eachtially bin, extended we model the Navarro, putative Frenk dSph and White (NFW) DM the ROI centerThe were spectral also parameters included4 in of these the ROI sourcesour model. analysis were to fourfixedconstraints years at of Fermi-LATmodel. that data,3 areevaluate Within independentfrom each4 Poisson bin, variations, of we the model the choice result the is of in goodputative spectral agreement dSph to the Pass 7 Reprocessed data, but with a smalltotheir (< the 3FGLPassdensity catalog 7 Reprocessed profiles values. The [35]. fluxdata, This normalizations choicebut with was of a motivatedsource small with (< a by power-lawdensity the spectral profiles model [35 (dN/dE]. ThisE choice ) was motivated by the their 3FGL catalog values. The fluxan normalizations energy range of 300 MeV of – 100 GeVsource in 20 energy with bins a power-lawwith all previous spectral studies. Inmodel this figure, (dN/dE we show resultsE ) / 10%)The energy-dependent spectral parameters correction of these to sources account were for fixedutilizing di10%)↵theer- 8at energy Galactic energy-dependent binscurrentmodel. per di↵ decadeuse uncertaintyand Withinexcept isotropic for the eachcorrection final components bin in bin, thecorresponding we to spatial and model account 3FGL to the extension case the in forwith which putative di spectral no↵ of additionaler- the dSph index require- DMcurrent of = uncertainty 2. We show the in bin-by-bin the spatial extension of the DM ments are placed on the blank sky locations (blue),/ and their 3FGL catalogthe Galactic values. di↵use The and flux isotropic normalizations5 components(whichcatalog was of extendedand sources 3FGLsource to an within energy withwith of the 100 a GeV),10 power-law spectral and10 we ROI index spectral were of fit simul- model=5 2. We (integrateddN/dE show the energy-fluxE bin-by-bin) 95% confidence level upper lim- ences in the Pass 8 instrument response. Additionally,scanences the likelihood inhalos fits the usingPass power-law, of these 8 ratherinstrument new than⇥ dark objects.to response. when Previous the blank skyAdditionally, locations studies used have are further shown requiredhalos of these new objects. Previous studies have shown catalog sources within the 10 10 ROI weretaneously fit simul- in a binnedintegrated likelihood analysis energy-fluxto over lie no the closer 95% broad- than confidence0.5 fromits any for source each level listed/ dSphupper in the BZ-candidate lim- in Figure 2. The Poisson wethe model Galactic extragalactic di↵use and gamma-ray isotropic components emission⇥ and and residualmatter 3FGL motivated, spectralthatwith shapes. the spectral LAT In Fig. 1 index flux we show limits the of = are 2. fairly We show insensitive the bin-by-bin to model- taneously in a binned likelihood analysisdistribution overweband model the of energy the broad- TS extragalactic calculated range from inits our 500 foranalysis MeV gamma-ray each (TS toHL dSph500) GeV.CAT, candidate emissionCRATES, The fluxes CGRaBS, in and Figure orlikelihoods residual ATNF catalogs2. from The (red). eachthat Poisson This bin the were LAT combined flux limitsto form areglobal fairly insensitive to model- chargedcatalog particle sources contamination within the 10 with10 anROI isotropic were fitcompared model simul-and to normalizations that obtainedingintegrated dSph by the of Fermi-LAT targets the energy-flux background collabora- as point-like sources 95%will be confidencediscussed are versus insen- in more spatially detail levelspectral in Sec. upper likelihoods IV. extended lim- for di↵erent DM annihilation chan- band energy range from⇥ 500 MeV to 500tion GeV. (TScharged3FGL The)inthesameenergyrange.Wefindthatour fluxes particle contaminationlikelihoods from with each bin an wereisotropic combined model to forming global dSph targets as point-like versus spatially extended fittaneously to the Pass in a 8 binneddata. likelihood These models analysis will over be included theTS broad- valuessitive in are, to on the average,sourcesits inclusion for slightly each [15 (13.5%) of]. a dSph Following putative lower those candidate power-law the procedure in source Figure at of2 Ackermannnels. The and masses. Poissonet al. and normalizations of the background sourcesfit to are the insen-Pass 8 spectraldata. These likelihoods models for will di↵erent beIII. included RESULTS DM annihilation in sources chan- [15]. Following the procedure of Ackermann et al. band energy range from 500 MeV to 500 GeV. Thereported fluxesthe in the locations 3FGL.likelihoods We of attribute the DES this from primarily dSph each candidates, to the bin were as expected combined toWe form tested global for excess gamma-ray emission consistent the forthcomingsitive public to thePass inclusion 8 data of a release. putative Point-like power-lawfact thatthewhen we forthcoming normalize source there[17 is the], at no we background bright fitnels public for point by and excessfitting sourcePass masses. over gamma-ray a at 8 thedata center release. of emission the Point-like associated with[17], we fit for excess gamma-ray emission associated with 10 10 region, rather than over the entire sky. The In Fig. 3, we show the delta-log-likelihoodwith two representative (LG(L)) dark matter annihilation chan- and normalizations of the background sources are insen-⇥ spectral likelihoods for di↵erent DM annihilation chan- sources from thethe recent locations 3FGL of the catalog DES dSph [34] within candidates,dashed 15ROI. curveof as in expected Fig.each 1 represents target the best-fit in each gaussian energy of distribution bin for separately our analysis of Fermi to data derive from¯ the flux direc-+ sources from the recentWe tested 3FGL for catalog excess [ gamma-ray34] withinnels 15emission (i.e., ofbb and consistenteach⌧ ⌧ ) target and a range in each of particle energy masses bin separately to derive flux sitive to the inclusion of a putative power-law sourcethis distribution, atIn contrast withnels a mean to and Ackermann of -0.135 masses. and aet standard al. [17],tion we of modeled Reticulum the II. As in both Ref. [26] and Ref. [27], the ROI centerwhen were there also is no included bright point in the source ROI atdeviation model. the center of 0.176. constraints of the that are independent of the choicefrom 2 GeVof spectral to 10 TeV (when kinematically allowed). the locations of the DES dSph candidates, as expectedthedSph ROI candidates center as were point-likewith two also sources representative included ratherwe find an than in excess dark the spa- of events ROI matter in the model. annihilation bins covering approx-constraints chan- that are independent of the choice of spectral ROI. Secondly, we apply theWe “blank-sky” tested null-test for employed excessimately gamma-ray+ 2-10 GeV. For emission a spectralNo significant shape consistent corresponding excess gamma-ray emission was observed The spectral parameters of these sources were fixed at model. Withinnels (i.e., eachb¯b and bin,⌧ we⌧⇠ ) model and a the range putative of particle dSph¯ masses when there is no bright point source at the centerin ofpreviousThe the dwarf spectral spheroidal parameters studies. Specifically,of we se- theseto a 49 sources GeV dark matter were particlefrom fixed annihilating any at of the to DESbmodel.b (the dSph candidates Within for each any bin, of the we DM model the putative dSph In contrast to Ackermann et al. [17],lect we 1905 modeled sky locationswith the with twob > 30 representative, which are 1 re- best-fit dark mass matter for the Galactic annihilation Center excess [9]), chan- we find their 3FGL catalog values. The flux normalizations of source| | withfrom a power-law 2 GeV to 10spectral TeV (when model kinematically (dN/dE E allowed). ) ROI. movedtheir from any 3FGL 3FGL source catalog and 5 removed values. from+ any Thea value flux of TS=17.4 normalizations from Reticulummasses II, or correspondingof channelssource totested. with The a data power-law were found spectral to be model (dN/dE E ) dSph candidates as point-like sources rather4 thannels spa- (i.e., b¯b and ⌧ ⌧ ) and a range of particle/ masses the Galactic di↵use and isotropic components andextended 3FGLthehttp://fermi.gsfc.nasa.gov/ssc/data/access/lat/ 3FGL Galactic source.with In di this spectral↵ case,useNo we and significant employ index isotropic the full of excessasignificanceof3.2 components= 2. gamma-ray We show(see and Fig.well emission the 2). 3FGL described If webin-by-bin do was not by imposeobservedwith the background spectral model index with of no signif-= 2. We show the bin-by-bin/ In contrast to Ackermann et al. [17], we modeled theBackgroundModels.htmlfrom 2 GeV to 10 TeVthis (when choice of the kinematically dark matter mass, but allowed). rather allow the catalog sources within the 10 10 ROI were fit simul-5 integratedfrom energy-flux any of the 95% DESmass to dSph confidence float as candidates a free parameter,icant level for residuals the upper valueany of of observed. the lim- the TS DM We calculated the test statis- dSph candidates as point-like sources rather thancatalog spa-Standard sources LAT analyses within treat the the di↵use 10 emission 10 model ROI as being were fit simul- integrated energy-flux 95% confidence level upper lim- ⇥ defined inNo terms significant of truemasses energy, excess but or the channels model gamma-rayincreases was necessarily tested. only slightly emission de- The (to 18.1), datatic (TS) was illustrating were for observed signal thefound compat- detection to be by comparing the likelihood taneously in a binned likelihood analysis over the3 broad-MET range: 239557417its - 365467563 for each dSph candidate⇥ibility between in Figure this signal and2. that The observed Poisson from the 4 http://fermi.gsfc.nasa.gov/ssc/data/access/lat/taneouslyrived fromfrom the in measured a any binnedwell of energies the described likelihood DES of events. dSph by This analysis the candidates implies background a weakover for thevalues model any broad- both of with the with DM noits and signif- forwithout each the dSph added dSph candidate candidate in Figure 2. The Poisson band energy rangeBackgroundModels.html from 500 MeV to 500 GeV. The fluxesdependencelikelihoods of the model onfrom the instrument each bin response were functions. combinedtemplate to form (see global Equation 6 in Ackermann et al. [17]). bandThe correction energy applied range to from the di↵use 500 emission MeV model to 500 accounts GeV. for The fluxes likelihoods from each bin were combined to form global 5 Standard LAT analyses treat the di↵use emission model asmasses being oricant channels residuals tested. observed. The data We calculated were found the to test be statis- and4 normalizations of the background sources are insen-the di↵erentspectral energy dependence likelihoods of the e for↵ective di area↵erent and energy DM annihilationThe most significant chan- excess, TS = 6.8, was for http://fermi.gsfc.nasa.gov/ssc/data/access/lat/defined in terms of true energy, but the model wasand necessarily normalizationswell de- describedtic of (TS) the by for the background signal background detection sources model by comparing are with insen- no the signif- likelihoodspectral likelihoods for di↵erent DM annihilation chan- resolution between Pass 7 Reprocessed and Pass 8. DES J0335.6 5403 and a DM particle with m = sitiveBackgroundModels.html to the inclusionrived from of the a putative measured energies power-law of events. source This at implies anels weak and masses. DM 5 sitive toicant the inclusion residualsvalues of bothobserved. a putative with and We power-law without calculated the source added the test dSph at statis- candidatenels and masses. theStandard locations LAT of analysesdependence the DES treat of dSph the the model di candidates,↵use on emission the instrument modelas expected as response being functions. the locationsticWe (TS) of tested thefortemplate signal DES for excess (see detectiondSph Equation candidates, gamma-ray by 6 comparing in Ackermann as emission expected the likelihoodet consistent al. [17]). whendefined there in is terms noThe ofbright true correction energy, point applied but source the to model the at di the↵ wasuse center emission necessarily of model the de- accounts for We tested for excess gamma-ray emission consistent rived from the measuredthe di↵erent energies energy of dependence events. This of the implies e↵ective awhen weak area and therewith energyvalues is two no both bright representativeThe with pointmost and without significant source dark at the matter the excess, added center annihilation dSph TS of = the candidate 6.8, chan-with was for two representative dark matter annihilation chan- ROI.dependence of theresolution model between on the instrumentPass 7 Reprocessed responseand functions.Pass 8. ¯ + ROI. nelstemplate (i.e.,DES (seebb and J0335.6 Equation⌧ ⌧ 5403) 6 and in and Ackermann a range a DM of particleet particle al. [17 with]). massesmDM = ¯ + The correction applied to the di↵use emission model accounts for nels (i.e., bb and ⌧ ⌧ ) and a range of particle masses In contrast to Ackermann et al. [17], we modeled theIn contrastfrom 2 to GeV Ackermann to 10 TeVet al. (when[17], we kinematically modeled the allowed). dSphthe candidates di↵erent energy as dependence point-like of sources the e↵ective rather area than and energy spa- The most significant excess, TS = 6.8, was forfrom 2 GeV to 10 TeV (when kinematically allowed). resolution between Pass 7 Reprocessed and Pass 8. dSph candidatesNo significant as point-like excess gamma-ray sources rather emission than was spa- observed DES J0335.6 5403 and a DM particle with mDM No= significant excess gamma-ray emission was observed from any of the DES dSph candidates for any of the DMfrom any of the DES dSph candidates for any of the DM masses or channels tested. The data were found to be 4 http://fermi.gsfc.nasa.gov/ssc/data/access/lat/ masses or channels tested. The data were found to be 4 http://fermi.gsfc.nasa.gov/ssc/data/access/lat/well described by the background model with no signif- BackgroundModels.html well described by the background model with no signif- BackgroundModels.html 5 Standard LAT analyses treat the di↵use emission model as being icant residuals observed. We calculated the test statis- 5 Standard LAT analyses treat the di↵use emission model as being icant residuals observed. We calculated the test statis- defined in terms of true energy, but the model was necessarily de- tic (TS) for signal detection by comparing the likelihood defined in terms of true energy, but the model was necessarily de- tic (TS) for signal detection by comparing the likelihood rived from the measured energies of events. This implies a weak values both with and without the added dSph candidate rived from the measured energies of events. This implies a weak dependence of the model on the instrument response functions. values both with and without the added dSph candidate dependencetemplate of the (see model Equation on the instrument 6 in Ackermann response functions.et al. [17]). The correction applied to the di↵use emission model accounts for template (see Equation 6 in Ackermann et al. [17]). The correction applied to the di↵use emission model accounts for the di↵erent energy dependence of the e↵ective area and energy The most significant excess, TS = 6.8, was for resolution between Pass 7 Reprocessed and Pass 8. the di↵erent energy dependence of the e↵ective area and energy The most significant excess, TS = 6.8, was for resolutionDES between J0335.6Pass5403 7 Reprocessed and a DMand Pass particle 8. with mDM = DES J0335.6 5403 and a DM particle with mDM = The dark matter content of Reticulum II Draft version April 14, 2015 Preprint typeset using LATEXstyleemulateapjv.5/2/11

DARK MATTER ANNIHILATION AND DECAY PROFILES FOR THE RETICULUM II DWARF SPHEROIDAL GALAXY Vincent Bonnivard1,Celine´ Combet1,DavidMaurin1,AlexGeringer-Sameth2,SavvasM.Koushiappas3, Matthew G. Walker2,MarioMateo4,EdwardW.Olszewski5,andJohnI.BaileyIII4 Draft version April 14, 2015

ABSTRACT The dwarf spheroidal galaxies (dSph) of the Milky Way are among the most attractive targets for indirect searches of dark matter. In this work, we reconstruct the dark matter annihilation (J-factor) and decay profiles for the newly discovered dSph Reticulum II. This is done using an optimized spher- ical Jeans analysis of kinematic data obtained from the Michigan/Magellan Fiber System (M2FS). We find Reticulum II to have one of the highest J-factor when compared to the other Milky Way dSphs. We have also checked the robustness of this result against several ingredients of the analysis. Unless it su↵ers from tidal disruption or significant inflation of its velocity dispersion from binary stars, Reticulum II may provide a unique window on dark matter particle properties. Subject headings: galaxies: dwarf — galaxies: individual (Reticulum II) — dark matter — gamma rays: galaxies — methods: statistical — stars: kinematics and dynamics

1. INTRODUCTION der to set constraints on the DM particle properties. Along with the Galactic center and galaxy clusters, Here, we reconstruct the DM annihilation and decay the dwarf spheroidal galaxies (dSph) of the Milky Way profiles of Ret II from a spherical Jeans analysis ap- have been identified as promising targets for indirect plied to stellar kinematic data obtained with the Michi- dark matter (DM) searches (see recent reviews by Stri- gan/Magellan Fiber System (M2FS) (Walker et al. 2015). gari 2013; Conrad et al. 2015). Their low astrophysi- We use the optimized Jeans analysis setup from Bon- cal background, high mass-to-light ratio, and proximity nivard et al. (2015a,b), described in Section 2. From the make them compelling targets(Lake 1990; Evans et al. reconstructed DM density profiles, we then compute the 2004). About twenty-five Galactic dSphs were known as astrophysical J- and D-factors, for annihilating and de- of early 2015, and their observation by -ray telescopes caying DM respectively, and cross-check our results by has thus far shown no significant emission, leading to varying di↵erent ingredients of the analysis (Section 3). Finally, we evaluate the ranking of Ret II among the most stringent constraints on annv , the thermally-averaged DM self-annihilation cross-sectionh i (Acciari et al. 2010; promising dSphs for DM indirect detection in Section 4. Paiano et al. 2011; Abramowski et al. 2014; Geringer- 2. ASTROPHYSICAL FACTORS, JEANS ANALYSIS AND Sameth et al. 2014; Fermi-LAT Collaboration 2015). DATA SETS Recently, imaging data from the Dark Energy Survey 2.1. Astrophysical factors has led to the discovery of nine new (potential) Milky- Way satellites in the Southern sky (Koposov et al. 2015; The di↵erential -ray flux coming from DM annihila- DES Collaboration et al. 2015). The nearest object, tion (resp. decay) in a dSph galaxy is proportional to the Reticulum II (Ret II, d 32 kpc), is particularlyThe dark intrigu- matter contentso-called of ‘astrophysical’Reticulum II factor J (resp. D) (Bergstr¨om ing, as evidence of -ray⇠ emission has been detected in et al. 1998), its direction using the public Fermi-LAT data (Geringer- v 2 Sameth et al. 2015b; Hooper & Linden 2015). TheJ Fermi-h i J = ⇢DM(l, ⌦) dld⌦ resp.D= ⇢DM(l, ⌦) dld⌦ , (1) M 2 LAT collaboration simultaneously published a⇠ search for ZZ ✓ ZZ ◆ -ray emission from the newly discovered objects (Fermi- which corresponds to the integration along the line-of- LAT Collaboration et al. 2015), based on the unreleased sight (l.o.s.) of the DM density squared (resp. DM den- arXiv:1504.03309v1 [astro-ph.HE] 13 Apr 2015 PASS8 dataset, and found no significant excess. sity) and over the solid angle ⌦ =2⇡ [1 cos(↵ )], Nonetheless, and whatever the situation regarding a ⇥ int where ↵int is the integration angle. This quantity de- (non-)detection in this object might be, a robust de- pends on both the extent of the DM halo and the mass termination of Ret II’s DM content is crucial in or- density distribution, and is essential for putting con- [email protected] straints on the DM particle properties. All calculations [email protected] of astrophysical factors are done with the CLUMPY code 1 LPSC, Universit´eGrenoble-Alpes, CNRS/IN2P3, 53 avenue (Charbonnier et al. 2012), a new module of which has des Martyrs, 38026 Grenoble, France 2 been specifically developed to perform the Jeans analy- McWilliams Center for Cosmology, Department of Physics, 6 Carnegie Mellon University, Pittsburgh, PA 15213, USA sis . 3 Department of Physics, Brown University, Providence, RI 02912, USA 2.2. Jeans analysis 4 University of Michigan, 311 West Hall, 1085 S. University Ave., Ann Arbor, MI 48109, USA 5 Steward Observatory, The University of Arizona, 933 N. 6 This upgrade will be publicly available in the soon-to-be re- Cherry Ave., Tucson, AZ 85721, USA leased new version of the software (Bonnivard et al., in prep.). Draft version April 14, 2015 Preprint typeset using LATEXstyleemulateapjv.5/2/11

DARK MATTER ANNIHILATION AND DECAY PROFILES FOR THE RETICULUM II DWARF SPHEROIDAL GALAXY Vincent Bonnivard1,Celine´ Combet1,DavidMaurin1,AlexGeringer-Sameth2,SavvasM.Koushiappas3, Matthew G. Walker2,MarioMateo4,EdwardW.Olszewski5,andJohnI.BaileyIII4 Draft version April 14, 2015

ABSTRACT The dwarf spheroidal galaxies (dSph) of the Milky Way are among the most attractive targets for indirect searches of dark matter. In this work, we reconstruct the dark matter annihilation (J-factor) and decay profiles for the newly discovered dSph Reticulum II. This is done using an optimized spher- ical Jeans analysis of kinematic data obtained from the Michigan/Magellan Fiber System (M2FS). We find Reticulum II to have one of the highest J-factor when compared to the other Milky Way dSphs. We have also checked the robustness of this result against several ingredients of the analysis. Unless it su↵ers from tidal disruption or significant inflation of its velocity dispersion from binary stars, Reticulum II may provide a unique window on dark matter particle properties. Subject headings: galaxies: dwarf — galaxies: individual (Reticulum II) — dark matter — gamma rays: galaxies — methods: statistical — stars: kinematics and dynamics

1. INTRODUCTION der to set constraints on the DM particle properties. Along with the Galactic center and galaxy clusters, Here, we reconstruct the DM annihilation and decay the dwarf spheroidal galaxies (dSph) of the Milky Way profiles of Ret II from a spherical Jeans analysis ap- have been identified as promising targets for indirect plied to stellar kinematic data obtained with the Michi- dark matter (DM) searches (see recent reviews by Stri- gan/Magellan Fiber System (M2FS) (Walker et al. 2015). gari 2013; Conrad et al. 2015). Their low astrophysi- We use the optimized Jeans analysis setup from Bon- cal background, high mass-to-light ratio, and proximity nivard et al. (2015a,b), described in Section 2. From the make them compelling targets(Lake 1990; Evans et al. reconstructed DM density profiles, we then compute the 2004). About twenty-five Galactic dSphs were known as astrophysical J- and D-factors, for annihilating and de- of early 2015, and their observation by -ray telescopes caying DM respectively, and cross-check our results by has thus far shown no significant emission, leading to varying di↵erent ingredients of the analysis (Section 3). Finally, we evaluate the ranking of Ret II among the most stringent constraints on annv , the thermally-averaged DM self-annihilation cross-sectionh i (Acciari et al. 2010; promising dSphs for DM indirect detection in Section 4. Paiano et al. 2011; Abramowski et al. 2014; Geringer- 2. ASTROPHYSICAL FACTORS, JEANS ANALYSIS AND Sameth et al. 2014; Fermi-LAT Collaboration 2015). DATA SETS Recently, imaging data from the Dark Energy Survey 2.1. Astrophysical factors has led to the discovery of nine new (potential) Milky- Way satellites in the Southern sky (Koposov et al. 2015; The di↵erential -ray flux coming from DM annihila- DES Collaboration et al. 2015). The nearest object, tion (resp. decay) in a dSph galaxy is proportional to the Reticulum II (Ret II, d 32 kpc), is particularlyThe dark intrigu- matter contentso-called of ‘astrophysical’Reticulum II factor J (resp. D) (Bergstr¨om ing, as evidence of -ray⇠ emission has been detected in et al. 1998), its direction using the public Fermi-LAT data (Geringer- v 2 Sameth et al. 2015b; Hooper & Linden 2015). TheJ Fermi-h i J = ⇢DM(l, ⌦) dld⌦ resp.D= ⇢DM(l, ⌦) dld⌦ , (1) M 2 LAT collaboration simultaneously published a⇠ search for ZZ ✓ ZZ ◆ -ray emission from the newly discoveredUse objects stellar kinematics (Fermi- to reconstructwhich the corresponds gravitational potential. to the integration along the line-of- LAT Collaboration et al. 2015), based on the unreleased sight (l.o.s.) of the DM density squared (resp. DM den- arXiv:1504.03309v1 [astro-ph.HE] 13 Apr 2015 PASS8 dataset, and found no significant excess. PHYSICAL REVIEWsity) D 75, and083526 over (2007) the solid angle ⌦ =2⇡ [1 cos(↵ )], Nonetheless, and whatever the situation regarding a ⇥ int Precise constraints on the dark matterwhere content↵ ofint Milkyis the Way dwarf integration galaxies angle. This quantity de- (non-)detection in this object might be, a robust de-for gamma-raypends experiments on both the extent of the DM halo and the mass termination of Ret II’s DM content is crucial1, in or- 2,† 1,‡ 1, Louis E. Strigari, * Savvas M. Koushiappas,densityJames S. Bullock, distribution,and Manoj Kaplinghat and isx essential for putting con- 1Center for Cosmology, Department of Physics and Astronomy, University of California, Irvine, California 92697, USA [email protected] 2Theoretical Division and ISR Division, MS B227, Los Alamosstraints National Laboratory, on the Los DM Alamos, particle New Mexico 87545, properties. USA All calculations The Astrophysical Journal, 678:614–620, 2008 May(Received 10 7 December 2006; published 30 April 2007) A [email protected] # 2008. The American Astronomical Society. All rights reserved. Printed in U.S.A. of astrophysical factors are done with the CLUMPY code 1 LPSC, Universit´eGrenoble-Alpes, CNRS/IN2P3,The AstrophysicalWe examine the Journal prospects 53, 678:614–620, avenue for detecting 2008 -rays May from 10 dark matter annihilation in the six most promising A #dwarf2008. The spheroidal American Astronomical (dSph) satellite Society. galaxies All rights reserved. of the Milky Printed(Charbonnier Way.in U.S.A. We use recently measured et al. velocity 2012), dispersion a new module of which has des Martyrs, 38026 Grenoble, France profiles to provide a systematic investigation of the dark matter mass distribution of each galaxy, and show 2 been specifically developed to perform the Jeans analy- McWilliams Center for Cosmology, Departmentthat the uncertainty in of the Physics, -ray flux from mass modeling6 is less than a factor of 5 for each dSph if we THEassume MOST a smooth DARK-MATTER–DOMINATED Navarro-Frenk-White (NFW) profile.sis GALAXIES:. We show that Ursa PREDICTED Minor and Draco GAMMA-RAY are the most Carnegie Mellon University, Pittsburgh, PA 15213,SIGNALS USA FROM THE FAINTEST MILKY WAY DWARFS 3 Department of Physics, Brown University,promising dSphs Providence, for -ray detection RI with GLAST and other planned observatories. For each dSph, we investigateLouis E. the Strigari, fluxTHE enhancement MOST1 Savvas DARK-MATTER–DOMINATED resulting M. Koushiappas, from halo2 substructure,James S. Bullock, and show1 Manoj that GALAXIES: the Kaplinghat, enhancement PREDICTED1 factor GAMMA-RAY 02912, USA relative to a smooth haloJoshua flux cannot D. Simon,SIGNALS be greater3 Marla than FROM Geha,about 100.4 THEand This Beth FAINTEST enhancement Willman5 depends MILKY2.2. very WAYJeans weakly DWARFS on analysis 4 the lower mass cutoff scale of the substructure mass function. While the amplitude of the expected flux University of Michigan, 311 West Hall, 1085 S. UniversityReceived 2007 October1 12; accepted 2008 January 72 1 1 from each dSph dependsLouis sensitively E. Strigari, on the darkSavvas matter model, M. Koushiappas, we show that theJames flux ratios S. Bullock, between theManoj six Kaplinghat, Ave., Ann Arbor, MI 48109, USA 3 4 5 5 Sphs are known to within a factor of aboutJoshua 10. The flux D.6 Simon,ratios areMarla also relatively Geha, insensitiveand Beth to the Willman current Steward Observatory, The Universitytheoretical of Arizona, range of cold dark 933 matter N. halo centralABSTRACTRecei slopesvedThis and 2007 substructure October upgrade 12; fractions. accepted will 2008 be January publicly 7 available in the soon-to-be re- Cherry Ave., Tucson, AZ 85721, USAWe use kinematic data from three new nearby, extremelyleased low luminosity new version Milky Way of dwarf the galaxies software (Ursa Major (Bonnivard II, et al., in prep.). WillmanDOI: 1, and10.1103/PhysRevD.75.083526 Coma Berenices) to constrain the properties ofPACS their numbers: dark matter 95.35.+d, halos, 14.80.Ly, and from 98.35.Gi, these 98.62.Gq we make pre- dictions for the -ray flux from annihilation of dark matter particles in theseABSTRACT halos. We show that these 103 L dwarfs  6 are the most dark-matter–dominatedWe use kinematic galaxies data from known, three with new totalsystem nearby, masses contains extremely within at 100 least low pc luminosity18 that dSphs, are in which Milky excess are Way of observed 10 dwarfM . to galaxies be (Ursa Major II, I. INTRODUCTION Coupled with theirWillman relative proximity,1, and Coma their Berenices) large masses to implyconstrainlow-luminosity that thethey properties should systems have of mean their with dark-ray an extent matter fluxes that halos,kpc are. and Basedcom- from on these we make pre- Inparable the  toCDM or greatercosmologicaldictions than those for the model, of any-ray cold other flux dark knownfrom matter annihilation satellite galaxytheir ofdark stellar of the matter mass-to-light Milky particles Way. Our in ratios, these results dSphs halos. are robust contain We show to both of that order these 103 L dwarfs variations of the inner slope of the density profile and the effect of tidal interactions. The fluxes could be boosted by up  6 (CDM) comprises approximatelyare the most dark-matter–dominated one-fourth of the total galaxiesO 101 known,–102 more with mass total in masses dark matter within than 100 in pc visible that are light in excess of 10 M . to 2 orders of magnitude if we include the density enhancements caused by surviving dark matter substructure. energy density of theCoupled Universe with [1].their However, relative the proximity, nature of their[20 large] and masses thus† are imply ideal that laboratories they should for have studies mean that -ray are fluxes that are com- darkSubject matter headin remainsggs:parable unknown.cosmology: to orExtensions greater theory — than to dark the those standard matter of any othersensitive known to satellite the distribution galaxy of of the dark Milky matter. Way. Furthermore, Our results are robust to both model,Online such material: as thosevariationscolor based figures on of supersymmetry the inner slope [ of2,3 the] and densitytheir profile relative and the proximity effect of and tidal high interactions. galactic The longitude fluxes andcould be boosted by up universal extra dimensionsto 2 orders [4], of predict magnitude the existence if we include of thelatitude density makes enhancements them ideal caused for by surviving high signal-to-noise dark matter substructure. stable, weakly interactingSubject massiveheadinggs: particlescosmology: (WIMPs) theory —detection. dark matter with mass 101.1– INTRODUCTION104 GeV, which provide excellent can- Berenices,In this paper, and Ursa we consider Major II) the are prospects likely to for have -ray masses detec- that are ‰ OnlineŠ material: color figures comparable to those of their more luminous counterparts. Initial Thedidates census for of the cold Local dark Group matter. has In changed these models, dramatically WIMPs in tion from dark matter annihilation in six dSphs of the local estimates have already shown that these galaxies have mass-to- the lastinteract few years. gravitationally Prior to the turn as of well the century, as weakly, there were therefore only group. The six dSphs are selected because of both their light ratios that are similar to or larger than those of the pre-SDSS 11 knownWIMP satellite annihilation galaxies can of theproduce Milky Way-ray (MW),photons. with a dis- proximity and estimated masses, the latter of which is dwarfs (Martin et al. 2007; Simon & Geha 2007). With luminos- covery ratePresent of roughly and next-generation one new Local Group-ray1. INTRODUCTION observatories satellite per decade such as based on the most recentBerenices, measurements and Ursa of Major their II)velocity are likely to have masses that are ities that are more thancomparable 2 orders of to magnitude those of less their than more those luminous of counterparts. Initial (MateoSTACEE 1998). However, [5],The HESScensus the Sloan [6], of Digital MAGIC the Local Sky [7 Survey], Group VERITAS (SDSS) has changed has [8], dramaticallydispersion profiles. in We estimate the range of allowable been able to uncover a population of extremely low luminosity the pre-SDSS dwarfs,estimates these new satelliteshave already are interesting shown that not theseonly galaxies have mass-to- CANGAROOthe last [9], few GLAST years. [ Prior10], and to the HAWC turn of [ the11] century, will theredistributionsin the were context only of of dark galaxy matter formation that satisfy at the lowest the observed mass scales, ve- but satellite galaxies, which has roughly doubled the number of known light ratios that are similar to or larger than those of the pre-SDSS search for11 the known signatures satellite of dark galaxies matter of annihilation. the Milky Way The (MW),locityalso with for dispersion indirect a dis- dark profiles, matter and detection. deduce The the new -ray dwarfs flux are ex- very satellitesnearest (Willman location et al. to 2005; search Zucker for this et signalal. 2006; is Belokurovthe center ofet al. the pected from each dSph.dwarfs We (Martin focus et on al. quantifying 2007; Simon the & Geha 2007). With luminos- 2007; Irwin et al.covery 2007; rate Walsh of roughly et al. 2007). one new Determining Local Group how satellitefaint, per but decade they contain large amounts of dark matter and are lo- Milky Way, although uncertain backgrounds from astro- uncertaintycated quite nearby, in the which predictedities makes that fluxes are them more thatideal than comes sites 2 to orders searchfrom theof for magnitude sig- less than those of these new satellites(Mateo fit in 1998). a given However, model for the dark Sloan matter Digital and cos- Sky Survey (SDSS) has physical sources would make the clean extraction of such a darknals of matter dark matterdensity annihilation. distributionthe pre-SDSS in dwarfs, each system. these new As part satellites are interesting not only mology presentsbeen a very able exciting to uncover theoretical a population challenge. of extremely low luminosity of thisCurrent uncertainty, and future wein observatories, determine the context the of including flux galaxy contribution formationspace-based of at ex- the lowest mass scales, but Thesignal cold difficult darksatellite matter [12– (CDM) galaxies,14]. Additionally, model which predicts has there roughly the isexistence wide doubled empiri- of the number of known substructureperiments, such within as GLAST the dSph(Ritz dark et al. matter 2006), halos. as well as a suite of hundredscal uncertainty of MW satellites as to the that shape are expected of the central to host darkgalaxies matter at also for indirect dark matter detection. The new dwarfs are very satellites (Willman et al. 2005; Zucker et al. 2006; Belokurovground-basedPast work et al. in Cerenkov the literature detectors, considered such detectingas STACEE -rays (Hanna the faintdensity end ofprofile, the luminosity which may function have been (Kauffmann altered by et the al. growth 1993; faint, but they contain large amounts of dark matter and are lo- 2007; Irwin et al. 2007; Walsh et al. 2007). Determininget al. 2002), how H.E.S.S. (Hofmann 2003), MAGIC (Cortina 2005), Klypinof et a supermassive al. 1999; Moore black et al. hole 1999). [15, The16] orability any ofprocess gas to which cool from dark matter annihilationcated quite in nearby, Milky which Way-bound makes dark them ideal sites to search for sig- these new satellites fit in a given model for dark matterVERITAS and cos- (Weekes et al. 2002), CANGAROO (Yoshikoshi et al. and formcan exchange stars in these energy low-mass between dark the matter baryonic halos and depends dark matter on a matter halos: dSphsnals were of studied dark matter in [14 annihilation.,21–24], more mology presents a very exciting theoretical challenge.1999), and HAWK (Sinnis 2005), will search for the signal of numbercomponents of complex (e.g. physical [17–19 processes,]). such as supernova feed- massive galaxies in theCurrent local group and future were considered observatories, in including space-based ex- The cold dark matter (CDM) model predicts the-rays existence from ofdark matter annihilations. The prospects for -ray back, theIn photoionizing the case of dwarf background, spheroidal and galaxies mass loss (dSphs), due to astro- tidal [ 25], potentially dark subhalos were studied in [26–31], detection from dark matterperiments, in well-known such as GLAST MW satellites(Ritz et al. with 2006), as well as a suite of interactionsphysical (Dekel backgroundshundreds & Silk 1986;of and MW Cole baryonic-dark satellites et al. 1994; that matter Somerville are expected interac- & to hostand galaxies the prospects at of detecting microhalos were explored in these observatories hasground-based been the subject Cerenkov of many previous detectors, studies such as STACEE (Hanna Primack 1999; Barkanathe faint & end Loeb of 1999; the luminosity Bullock et functional. 2000;Chiu (Kauffmann[32, et33 al.]. 1993; tions are expected to be largely absent. The Milky Way (Baltz et al. 2000; Evanset al. et 2002), al. 2004; H.E.S.S. Profumo (Hofmann & Kamionkowski 2003), MAGIC (Cortina 2005), et al. 2001; BensonKlypin et al. et 2002). al. 1999; Despite Moore the etbroad al. 1999). range of The ob- ability ofIn gas comparison to cool to previous studies of dSphs, our work is VERITAS (Weekes et al. 2002), CANGAROO´ (Yoshikoshi et al. served luminosities,and form the dark stars matter in these masses low-mass for all dark of the matter pre- halosthe2006; depends first Bergstrom to on combine a & Hooper theoretical 2006; predictions Strigari et al. for 2007b; CDM Sa halonchez- *Electronic address: [email protected] Conde et al. 2007). All1999), of these and systems HAWK are interesting (Sinnis 2005), targets will not search for the signal of SDSS satellites† arenumber constrained of complex to within physical a relatively processes, narrow range, such as supernovaprofile shapes feed- and normalizations with specific dynamical Electronic address:7 [email protected] only because of their large -rays mass-to-light from dark ratios, matter but annihilations. also because The prospects for -ray approximately‡Electronicback,1 address:6 ; the10 [email protected] photoionizingM within their background, inner 600 pc (Walker and mass lossconstraints due to tidal for each observed system. Though the observed  they are expected to havedetection very low from intrinsic dark matter-ray emission. in well-known This MW satellites with et al. 2007;xElectronic Strigariinteractions address: et al. 2007a). [email protected] (Dekel Understanding & Silk 1986; this Cole strong et al. lu- 1994;velocity Somerville dispersion & profiles are equally well fit by both is in contrast to the situationthese observatories at the Galactic has center, been where the subject astro- of many previous studies minosity bias atPrimack the low-mass 1999; end Barkana is crucial & Loeb to deciphering 1999; Bullock the et al. 2000; Chiu formation of these dwarf spheroidal (dSph) galaxies, as well as to physical backgrounds(Baltz hinder the et al. prospects 2000; Evans of extracting et al. 2004;the signal Profumo & Kamionkowski et al. 2001; Benson et al. 2002). Despite the broadfrom range dark of matter ob- annihilation (Hooper & Dingus 2004). More- constraining the nature of dark matter. 2006; Bergstrom & Hooper 2006; Strigari et al. 2007b; Sa´nchez- served luminosities, the dark matter masses for allover, of the the known pre- location of the MW satellites makes a search of In1550-7998 this paper= we2007 show=75(8) that=083526(13) three new and nearby members of 083526-1 Conde© 2007 et The al. 2007). American All Physical of these Society systems are interesting targets not SDSS satellites are constrained to within a relativelydark narrow matter range, annihilation well-defined, unlike the search of com- the Local Group discovered by the SDSS (Willman 1, Coma only because of their large mass-to-light ratios, but also because approximately 1 6 ; 107 M within their inner 600pletely pc (Walker dark substructure, which would rely on serendipitous dis- they are expected to have very low intrinsic -ray emission. This 1 et al. 2007; Strigari et al. 2007a). Understanding thiscovery strong (Calcaneo-Roldan lu- & Moore 2000; Tasitsiomi & Olinto Center for Cosmology, Department of Physics and Astronomy, University is in contrast to the situation at the Galactic center, where astro- of California, Irvine,minosity CA 92697. bias at the low-mass end is crucial to deciphering2002; Stoehr the et al. 2003; Koushiappas et al. 2004; Pieri et al. 2005; 2 Theoretical Divisionformation and ISR of Division, these dwarf Los Alamos spheroidal National Laboratory,(dSph) galaxies,Koushiappas as well as to 2006; Diemandphysical et backgrounds al. 2007; Baltz hinder et al. the 2007). prospects of extracting the signal Los Alamos, NM 87545. From the mass modelingfrom dark of the matter dark annihilation matter halos, (Hooper we pro- & Dingus 2004). More- 3 constraining the nature of dark matter. Department of Astronomy, California Institute of Technology, Pasadena, vide the first determinationover, theof the known -ray location signal from of the dark MW matter satellites makes a search of CA 91125. In this paper we show that three new and nearby members of 4 National Research Council of Canada, Herzberg Institute of Astrophysics, from Ursa Major II, Willmandark matter 1, and annihilation Coma Berenices well-defined, (‘‘Coma’’ unlike the search of com- Victoria, BC V9E 2E7,the Canada. Local Group discovered by the SDSS (Willmanhereafter). 1, Coma These galaxies provide promising targets for -ray 5 pletely dark substructure, which would rely on serendipitous dis- Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138. detection for three reasons:covery (1) (Calcaneo-Roldan they are the among & theMoore closest 2000; Tasitsiomi & Olinto 1 Center for Cosmology, Department of Physics and Astronomy, University of California, Irvine, CA 92697. 614 2002; Stoehr et al. 2003; Koushiappas et al. 2004; Pieri et al. 2005; 2 Theoretical Division and ISR Division, Los Alamos National Laboratory, Koushiappas 2006; Diemand et al. 2007; Baltz et al. 2007). Los Alamos, NM 87545. From the mass modeling of the dark matter halos, we pro- 3 Department of Astronomy, California Institute of Technology, Pasadena, vide the first determination of the -ray signal from dark matter CA 91125. 4 National Research Council of Canada, Herzberg Institute of Astrophysics, from Ursa Major II, Willman 1, and Coma Berenices (‘‘Coma’’ Victoria, BC V9E 2E7, Canada. hereafter). These galaxies provide promising targets for -ray 5 Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138. detection for three reasons: (1) they are the among the closest 614 The dark matter content of Reticulum II

The Astrophysical Journal, 808:108 (14pp), 2015 August 1 doi:10.1088/0004-637X/808/2/108 © 2015. The American Astronomical Society. All rights reserved.

MAGELLAN/M2FS SPECTROSCOPY OF THE RETICULUM 2 DWARF SPHEROIDAL GALAXY*

Matthew G. Walker1, Mario Mateo2, Edward W. Olszewski3, John I. Bailey III2, Sergey E. Koposov4, Vasily Belokurov4, and N. Wyn Evans4 1 McWilliams Center for Cosmology, Department of Physics, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA; [email protected] 2 Department of Astronomy, University of Michigan, 311 West Hall, 1085 S. University Avenue, Ann Arbor, MI 48109, USA 3 Steward Observatory, The University of Arizona, 933 N. Cherry Avenue., Tucson, AZ 85721, USA The Astrophysical Journal,4 808:95Institute(14pp of Astronomy,), 2015 July University 20 of Cambridge, Madingley Road, Cambridge, CB3 0HA, UK doi:10.1088/0004-637X/808/1/95 © 2015. The American Astronomical Society. All rightsReceived reserved. 2015 April 13; accepted 2015 June 14; published 2015 July 27

ABSTRACT STELLAR KINEMATICS AND METALLICITIES IN THE ULTRA-FAINT DWARF GALAXY RETICULUM II*†‡ We present results from spectroscopic observations with the Michigan/Magellan Fiber System (M2FS) of 182 stellar targets1 along the line of sight2 LOS to3 the newly2 discovered “4ultrafaint” object5 Reticulum 2 6Ret,7 2 . For 37 of these2 2 J. D. Simon , A. Drlica-Wagner ,( T. S.) Li , B. Nord , M. Geha , K. Bechtol , E. Balbinot( , E.) Buckley-Geer , H. Lin , targets,3 the spectra are7 suf,8 ficient to provide3 simultaneous3 estimates of9 LOS,10,11 velocity (vlo2s, median random12 error 13 J. Marshall , B. Santiago−1 , L. Strigari , M. Wang , R. H. Wechsler , B. Yanny , T. Abbott , A. H. Bauer , dvlos = 1.4 km s ), effective14 temperature15,16 (Teff , dTeff 17= 478 K), surface10,11 gravity (log g, 18dlog g = 0.63 dex), and iron7,19 G. M. Bernstein , E. Bertin , D. Brooks , D. L. Burke fi , D. Capozzi , A. Carnero Rosell , abundance ([Fe20 H,21], d[Fe H] = 0.47 dex18). We use these results7,19 to con rm 173 stars as members22 of Ret 2.2 From the 2,5 M. Carrasco Kind , C. B. D’Andrea , L. N. da Costa +,1.0 D. L. DePoy−1 , S. Desai , H. T. Diehl+,1.2 S. Dodelson−1 , member10 sample we estimate2 a velocity23 dispersion of s =73.6 km s about24 a mean of áv 2 ñ= 64.3 km2 s 2,5 C. E. Cunha , J. Estrada , A. E. Evrard , A. Fausti Netovlos , E.- Fernandez0.7 , D. A. Finleylos, B. Flaugher-1.2 , J. Frieman , 13 23 −251 ,26 20,21 27,28 12 2+0.19 29 E. Gaztanagain the solar, rest D. frameGerdes(~-, D.90.9 Gruenkm s in the, R. Galactic A. Gruendl rest frame),, K. and Honscheid a metallicity dispersion, D. James of s[Fe, S. H] = Kent0.49-,0.14 K. Kuehn , 2 17 7,19+0.34 17 27,30 23,31 24 7,19 N. Kuropatkindex about, O.a mean Lahav of á,[Fe M. H]A.ñ= G. Maia-2.58-0.33, M.. These March estimates, P. Martini marginalize, C. over J. Miller possible velocity, R. Miquel and metallicity, R. Ogando , A. K. Romergradients,32, A. which Roodman are consistent10,11, E. S. with Rykoff zero.10 Our,11, results M. Sako place14, RetE. Sanchez 2 on chemodynamical33, M. Schubnell scaling23, relations I. Sevilla followed20,33, R. by C. Smith12, M. Soares-Santosthe Milky Way2, F.’s Sobreira dwarf-galactic2,7, E. satellites. Suchyta Under27,28, M.assumptions E. C. Swanson of dynamic21, G. equilibrium Tarle23, J. and Thaler negligible34, D. contamination Tucker2, V. Vikram35, from binary stars—both of which must beA. checked R. Walker with12 deeper, and imaging W. Wester and2 repeat spectroscopic observations—the estimated velocity dispersion suggests a dynamical mass of MR()»= 5 Rs 2 (2) G 2.4+1.4 ´ 105 M enclosed (The DES Collaborationhh) vlos -0.8 fl 1 Carnegie Observatories, 813 Santa Barbara St., Pasadena, CA 91101, USA +286 within projected hal ight radius Rh ~ 32 pc, with mass-to-light ratio »=2()MRh LV 467-168 in solar units. 2 Fermi National Accelerator Laboratory, P.O. Box 500, Batavia, IL 60510, USA Key3 George words: P.galaxies: and Cynthia dwarf Woods– Mitchellgalaxies: Institute individual for Fundamental(Reticulum Physics 2) and– Astronomy,galaxies: kinematics and Department and of dynamics Physics and– Astronomy, Local Group – methods: data analysisTexas– A&Mtechniques: University, spectroscopic College Station, TX 77843, USA 4 Astronomy Department, Yale University, New Haven, CT 06520, USA Supporting material:5machine-readableKavli Institute for Cosmological table Physics, University of Chicago, Chicago, IL 60637, USA 6 Department of Physics, University of Surrey, Guildford GU2 7XH, UK 7 Laboratório Interinstitucional de e-Astronomia—LIneA, Rua Gal. José Cristino 77, Rio de Janeiro, RJ-20921-400, Brazil 8 1. INTRODUCTIONInstituto de Física, UFRGS, Caixa Postal 15051,luminosities Porto Alegre, characteristic RS-91501-970, of Brazil ultrafaints. One of them, Reticu- 9 Department of Physics, Stanford University, 382 Via Pueblo Mall, Stanford, CA 94305, USA 10 lum 2 Ret 2 , has already attracted attention for several The census ofKavli Local Institute Group for Particlegalaxies Astrophysics has been & Cosmology, revised P.O. Box 2450,( Stanford) University, Stanford, CA 94305, USA 11 dramatically and repeatedly over the lastSLAC decade. National Mining Accelerator of the Laboratory,reasons. Menlo First, Park, CA Ret 94025, 2 is USA the nearest (D ~ 30 kpc) and most 12 Cerro Tololo Inter-American Observatory, National Opticaleasily Astronomy detected Observatory, of the newly Casilla 603, discovered La Serena, objects. Chile Second, Ret 2 SDSS stellar13 catalog has yielded discoveries of ∼15 low- Institut de Ciències de l’Espai, IEEC-CSIC, Campus UAB, Facultatclearly de Ciències, has Torre a C5flattened par-2, E-08193 morphology, Bellaterra, Barcelona, which may Spain indicate luminosity, dwarf-galactic14 satellitesDepartment of of thePhysics Milky and Astronomy, Way (e.g., University of Pennsylvania, Philadelphia, PA 19104, USA Willman et al. 200515 Sorbonne; Zucker Universités, et UPMCal. 2006 Univ; Paris Belokurov 06, UMR 7095,ongoing Institut d’Astrophysique tidal disruption de Paris, or F-75014, perhaps Paris, rotation. France Third, using 16 et al. 2007). The PanDASInstitut and d’ PanStarrsAstrophysique surveys de Paris, have Univ. found Pierre et Mariepublic Curie & data CNRS from UMR7095, the Fermi F-75014-LAT, Paris, France Geringer-Sameth et al. 17 Department of Physics & Astronomy, University College London,fi Gower Street, London, WC1E 6BT, UK nearly two dozen new satellites18 of M31 (McConnachie (2015) nd evidence for gamma-ray emission that is consistent Institute of Cosmology & Gravitation, University of Portsmouth, Portsmouth, PO1 3FX, UK et al. 2009; Martin et al. 201319 )Observatório. All told, Nacional, the population Rua Gal. José of Cristinowith 77, Rio dark de Janeiro, matter RJ-20921-400, annihilation Brazil in Ret 2. The Fermi-LAT known Local Group galaxies20 Department has nearly of Astronomy, tripled University(McConna- of Illinois,Collaboration 1002 W. Green Street, et al Urbana,(2015 IL) 61801,assign USA low significance to the chie 2012). Of the new members,21 National perhaps Center for the Supercomputing most intriguing Applications,gamma-ray 1205 West Clark signal St., Urbana, based IL on 61801, unreleased USA Fermi-LAT data; 22 Department of Physics, Ludwig-Maximilians-Universität, Scheinerstr.1, D-81679 München, Germany are the Milky Way’s “ultrafaint” satellites.23 These objects have however, in an independent analysis of the public data, Hooper fl Department of Physics, University of Michigan, Ann Arbor, MI 48109, USA lowered the oor of the24 Institut observed de Física galaxy d’Altes luminosity Energies, Universitat function Autònoma& de Linden Barcelona,(2015 E-08193) reproduce Bellaterra, the Barcelona, original Spain detection. In any case, 25 from MV ~-8 to MV ~-Max2, such Planck that Institute some for galaxiesExtraterrestrial’ total Physics, Giessenbachstrasse,Ret 2 is clearly D-85748 an intriguing Garching, object, Germany and the first question to luminosities can be dominated by a26 singleUniversity red Observatory giant star ( Munich,Martin Scheinerstrasse 1, D-81679 Munich, Germany 27 settle is whether Ret 2 presents chemo-dynamical evidence for et al. 2008). Moreover, theCenter structural for Cosmology parameters and Astro-Particle of ultrafaints Physics, The Ohio State University, Columbus, OH 43210, USA 28 Department of Physics, The Ohio State University,dark matter. Columbus, That is,OH is 43210, Ret USA2 a globular cluster or a galaxy? have blurred what was once an obvious29 distinction between the Australian Astronomical Observatory, NorthHere Ryde, we NSW present 2113, Australia results from an initial spectroscopic Milky Way’s dwarf-galactic satellites30 Department and its of globular Astronomy, clusters. The Ohio State“ University,reconnaissance Columbus,” of OH individual 43210, USA stellar targets along the line of fi 31 As a result, the proper classi cationDepartment of most of ultrafaint Astronomy, objects University ofsight Michigan,(LOS Ann) Arbor,to Ret MI 2. 48109, We USA identify a sample consisting of now requires spectroscopic32 Department measurements of Physics and of velocityAstronomy, disper- Pevensey Building, University of Sussex, Brighton, BN1 9QH, UK 33 17 member stars, which we use to characterize Ret 2ʼs sions, metallicities, and metallicityCentro de dispersions Investigaciones that Energéticas, can indicate Medioambientales y Tecnológicas (CIEMAT), Madrid, Spain 34 Department of Astronomy, University of Illinois,chemodynamical 1002 W. Green Street, properties. Urbana, IL 61801, Speci USAfically, we estimate the the presence of a dark matter halo.35 Argonne National Laboratory, 9700 Southmeans Cass Avenue, and dispersions Lemont, IL 60439, of velocity USA and metallicity distributions, Most recently, the Dark Energy Survey (DES) has revealed Received 2015 April 12; accepted 2015and June we 16; check published for 2015 velocity July 23 and metallicity gradients that might nine new Galactic satellites at southern latitudes (Koposov provide clues about Ret 2ʼs dynamical state and formation et al. 2015; The DES Collaboration et al. 2015, “K15” andABSTRACT “DES15” hereafter). Seven of the new objects have sizes and history. Finally, we compare Ret 2ʼs properties to those of We present Magellan/M2FS, Very Large Telescope/GIRAFFE,known and dwarf Gemini galaxies South/GMOS and globular spectroscopy clusters of in the order to * This papernewly presents discovered data gathered Milky with Way the Magellan satelliteTelescopes Reticulum at LasII. Baseddetermine on the spectra which of 25population Ret II member can claim stars Retselected 2 as from its newest -1 Campanas Observatory,Dark Energy Chile. Survey imaging, we measure a mean heliocentricmember. velocity of 62.8 0.5 km s and a velocity -1 dispersion of 3.3 0.7 km s . The mass-to-light ratio of Ret II within its half-light radius is 470 210 ML, demonstrating that it is a strongly dark matter-dominated1 system. Despite its spatial proximity to the Magellanic Clouds, the radial velocity of Ret II differs from that of the LMC and SMC by 199 and 83 km s-1, respectively, suggesting that it is not gravitationally bound to the Magellanic system. The likely member stars of Ret II span

* This paper includes data gathered with the 6.5 m Magellan Telescopes located at Las Campanas Observatory, Chile. † Based on observations obtained at the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., undera cooperative agreement with the NSF on behalf of the Gemini partnership: the National Science Foundation (United States), the National Research Council (Canada), CONICYT (Chile), the Australian Research Council (Australia), Ministério da Ciência, Tecnologia e Inovação (Brazil) and Ministerio de Ciencia, Tecnología e Innovación Productiva (Argentina). ‡ Based on data obtained from the ESO Science Archive Facility under request number 157689.

1 The Astrophysical Journal, 808:108 (14pp), 2015The August Astrophysical 1 Journal, 808:95 (14pp), 2015 July 20 Walker et al. Simon et al.

Table 4Table 2 resolution spectroscopy from VLT/GIRAFFE and 6 members The darkSummary matter of Reticulum content 2ʼs ObservedSummary of Reticulum Photometric of Properties and of ReticulumII Spectroscopic II Properties using low resolution spectroscopy from Gemini South/GMOS. -1 Quantity ValueRow Reference Quantity Value Notes Ret II has a velocity dispersion of sv =3.3 0.7 km s , a corresponding to a dynamical mass within its half-light R.A. at center aJ2000 = 03:35:42(1) R.A.K15(J2000) 03:35:41L 5 radius of 5.6´ 2.4 10 M and a mass-to-light ratio of Decl. at center dJ2000 =-54:02:57(2) Decl.K15(J2000) −54:03:00L (3) Distance (kpc) 32 470 210 ML. Galactic longitude l = 266.2958 deg K15 L The metallicity of Ret II determined from the CaT is (4) MV,0 −3.6 ± 0.1 Galactic latitude b =-49.7357 deg K15 L [Fe H]=- 2.65 0.07, consistent with that of Segue 1 Distance modulus mM-=17.4(5) 0.2 LV,0K15(L) 2360 ± 200L (Frebel et al. 2014) within the uncertainties, and 0.2 dex Distance from Sun D ~ 30(kpc6) ϵ K15 0.60+L0.10 ∼ -0.20 lower than that of any other known galaxy. We find that Ret II Absolute magnitude MV =-2.7 0.1(7()−3.6 ± 0.1) K15r12((pcDES15) ) 55 ±L5 +0.23 has an internal metallicity spread of 0.28 ± 0.09 dex, with stars Exponential scale length Re = 3.37-0.13 arcmin K15 semimajor axis -1 (8) +0.02 Vhel (kms ) 62.8 ± 0.5 spanning a total range of more than 1 dex. Even the most metal- Ellipticity eba=-1()0.59 = -0.03 K15 L 9 V -1 −92.5 0.5 rich stars in the galaxy are at [Fe H]»- 2.0, and Ret II Position angle PA =71 ( 1) deg GSRK15(kms ) ±L fl +0.21(10) σ (kms-1) 3.3 ± 0.7 contains 4 extremely metal-poor stars with [Fe H]<- 3. Both Projected hal ight radius Rh = 3.64-0.12 arcmin K15 Rh »-1.68Ree 1 5 fi fl +(1.911) Mass within the half-light 5.6´ 2.4 10 its chemical and kinematic properties con rm that Ret II is a Projected hal ight radius Rh = 32-1.1 pc K15 L +1.2 −1 radius (M) dwarf galaxy. Systemic line of sight velocity vlos = 64.3-1.2 km s this work solar rest frame (12) −1 M1/2/LV (M/L) 470 ± 210 The location of Ret II just 23 kpc away from the LMC Systemic line of sight velocity vlos =-90.9 km s this work Galactic rest frame, given solar motion measured by Schönrich et al. (2010) (13) Mean [Fe/H] −2.65 ± 0.07 suggests that it could have originated as a satellite of the Internal velocity dispersion s = 3.6+1.0 km s−1 this work L vlos -0.7(14) Metallicity dispersion (dex) 0.28 ± 0.09 Magellanic system rather than having always been associated +0.4 −1 −1 Velocity gradient kv = 0.5-0.3 km s arcmin this work◦ 2 −5 L los (15) log10J (0. 2) (GeV cm ) 18.8 ± 0.6 with the Milky Way (Bechtol et al. 2015; Koposov et al. 2015). +217 PA of velocity gradient q = -92 deg this work◦ 2 −5 L However, our measured systemic radial velocity of = vlos -(6516) log10J (0. 5) (GeV cm ) 18.9 ± 0.6 vhel +0.34 -1 Mean metallicity á[Fe H]ñ=-2.58-0.33 dex this work L 62.8 0.5 km s means that Ret II is moving away from the +0.19 -1 Metallicity dispersion s[Fe H] = 0.49Note.-0.14 dexRows (1)–(7) thisare takenwork from the DES photometric similar analysis to of median Bechtol metallicityLMC error at a minimum velocity of 199 km s . According to +0.06et al. 2015 −.1 Values in rows 8 – 16 are derived in this paper. Metallicity gradient k[Fe H] = 0.01-0.06 dex arcmin( ) this work( ) ( ) L current LMC mass estimates, this velocity likely exceeds the +1.4 5 2 escape velocity of the LMC, indicating that the two objects are Mass enclosed within Rh M (Rh ) = 2.4-0.8 ´ 10 M this work M ()RRhh» 5svlos (2) G; assumes equilibrium, negligible binary stars +286 not gravitationally bound. This result does not rule out the Mass-to-light raio ¡ = 467 obtainedML by modelingthis work this individual¡ » system2()MRh is LV ; larger assumes than equilibrium, is negligible binary stars -168 possibility that Ret II was previously a Magellanic satellite, and obtained by modeling the entire population of dSphs future proper motion measurements will shed more light on its Note. (Martinez 2015). origin. a Unless otherwise noted, K15 and DES15 report similar values. Several previously known ultra-faint dwarf galaxies possess The J-factor calculated from the internal kinematics of Ret II larger mean J-factors than Ret II, most notably Segue 1, Ursa is logJ = 18.8 0.6 GeV25 cm- within a radius of 0 ◦.2, 4.3. Scaling Relations et al. 2007; Simon & Geha 2007; Walker10 et al. 2007 . In Major II, and Coma Berenices (Ackermann et al. 2014; Conrad somewhat lower) than previously estimated based on the contrast, the stellar kinematics of globular clusters generally do K15 use photometric data to show thatet al. Ret2015 2 occupies; Geringer-Sameth a et al. 2015a). Though the velocity galaxy’s distance alone (Drlica-Wagner et al. 2015). The dispersions of Ret II andnot Segue require 1 an are internal consistent dark within matter component. In order to see region of structural parameter space that is populated by objects predicted gamma-ray flux from dark matter annihilation in uncertainties, Ret II is morethese distant trends,(3 we2 kpc plotcomparedR s2 (GL to ) against luminosity, where of ambiguous classification, somewhat intermediate between h vlos V Ret II is therefore likely to be lower than that predicted for 23 kpc) and has a larger half-lightR is hal radiusflight as radius, measureds alongis LOS the velocity dispersion, L is total well-established globular clusters and dwarf galaxies Gilmore h vlos several other MilkyV Way satellites (Drlica-Wagner et al. 2015; major axis (5(5 pc compared to 29 pc). The larger distance and fi V-band luminosity and G is Newton’s constant.Geringer-Sameth The combina- et al. 2015c; Hooper & Linden 2015). et al. 2007, cf. K15ʼs Figure 17). Speci largercally, half-lightwith projected radius imply a reduced mean J-factor relative fl tion of macroscopic observables on theSatellite vertical galaxies axis likehas Ret II provide a crucial testing ground hal ight radius Rh =32 1 pc, Ret 2 isto larger Segue than 1. In nearly comparison all to Ursa Major II, Ret5 II is at a similar dimensions of M/L and is therefore sufforficient the Λ toCDM highlight paradigm, the and accordingly, the search for ultra- globular clusters and smaller than nearlydistance, all known but galaxies. has a velocity dispersion that is smaller by different behavior of dwarf galaxiesfaint and globular galaxies clusters. has become a major theme of near-field Moreover, its absolute magnitude, MV =-roughly2.7 a factor0.1, would of two. The larger dispersion, and hence mass, Again we find that Ret 2 followscosmology. the galactic It relation. is expected that many additional Milky Way place Ret 2 among the least luminousaccounts members for of the either larger J-factor of Ursa Major II. Coma Moreover, assuming dynamic equilibriumsatellite and galaxies negligible could be found in ongoing and near-future population. In these regards, Ret 2 andBerenices many of is its more newly distant than Ret II (44 kpc compared to contamination from binary stars, Ret 2wide- has amongfield optical the highest imaging surveys (Tollerud et al. 2008; discovered siblings are similar to “ultrafaint32” kpcsatellites); however, Segue 1, the larger velocity dispersion of Coma dynamical M/Ls of any known object.Hargis The et al. crude2014 mass. The link between newly discovered stellar Segue 2, and Willman 1, as well as to theBerenices extended implies globular a slightly larger mean J-factor. ) estimator of Walker et al. (2009a) impliessystems that and the dynamicalthe dark matter halos in which they may reside is clusters Pal 14 and Crater (Belokurov et al.Since2014 Segue; Laevens 1, Ursa Major II, and Coma Berenices all mass enclosed within Ret 2 s projected halflight radius is et al. 2014, Mateo et al. 2015). possess larger J-factors than Ret II, we expect dark matterʼ established by follow-up dynamical and chemical analysis. Ret MR()»= 5 Rs 2 (2) G 2.4+1.4 ´ 10II5 isM the, andfirst the of asso- several recently reported stellar systems Our spectroscopic results provide newannihilation leverage thatto produce can a largerhh gamma-rayvlos flux from these-0.8 +286 (Bechtol et al. 2015; Kim et al. 2015; Koposov et al. 2015; settle the question of Ret 2ʼs nature. Forobjects. Galactic However, globular no gamma-rayciated M excess/L is »= has2()MR beenh associated LV 467-168 in solar units. clusters as well as the dwarf spheroidal satellites of the Milky Laevens et al. 2015; Martin et al. 2015) to be spectroscopically with any of the previously knownTable dwarf4 summarizes galaxies (Ackermann the observed propertiesfi of Ret 2, based Way and M31, the top panel of Figure 9 plotset al. mean2015 metallicity). Given comparableon the gamma-ray photometric sensitivity, results of K15 it isand thecon spectroscopicrmed as a results dark-matter-dominated Milky Way satellite against luminosity. While globular clustersunlikely show thatno obvious a dark matterpresented annihilation in this work.signal Where would photometric be galaxy. results The from spectroscopicDES15 campaign to characterize new satellite galaxy candidates represents an essential step in trend, it is well-known that dwarf galaxiesdetected follow a from luminosity/ Ret II withoutdiffer also from being those detected of K15 from(e.g., dwarf for absolute magnitude), we list ongoing tests of the standard cosmological model. metallicity relation (Mateo 1998; Tolstoygalaxies et al. 2009 with; Kirbyhigher J-factorsthe DES15(Drlica-Wagnerresults as et well. al. 2015; et al. 2013). Given the low mean metallicityGeringer-Sameth that we estimate et al. 2015c; Hooper & Linden 2015). +0.34 This publication is based upon work supported by the from the M2FS spectra, á[Fe H]ñ=-2.58-0.33, we place Ret 2 5. SUMMARY AND DISCUSSIONNational Science Foundation under grant AST-1108811. We squarely onto the galactic relation (large black square5. SUMMARY in AND CONCLUSIONS thank Dan Kelson for helpful conversations, Anna Frebel for Figure 9 . ) We have presented the firstWe spectroscopic have presented analysis results of from the ourproviding initial stellar-spectro- the MIKE spectrum of HD 122563, and Becky The lower panel of Figure 9 shows another well-known recently discovered Milkyscopic Way satellite observations Reticulum of Ret II. 2. We IntegratingCanning for 2 and hr in Jimmy below- for helpful conversations on the reduction scaling relation that distinguishes dwarf galaxies from globular fi measure the velocities ofaverage 25 Ret observing II members conditions, using highMagellanof/M2FS VLT spectra. has delivered We also thank the anonymous referee for clusters. Speci cally, the mass-to-light ratiosresolution(M/Ls) spectroscopyof dwarf from Magellan/M2FS, as well as suggestions that improved the presentation of the paper. A.C.R. galaxies are anti-correlated with luminosity, such that gravita- 5 2 fi metallicities from the CaT linesMany for popular 16 members dynamical using mass estimators high haveacknowledgesMdynµ RG h svlos nancial, where support provided by the PAPDRJ. tional potentials in the least luminous galaxies all seem to be the constant of proportionality is typically between ~-26(e.g., Richstone & dominated by dark matter (Mateo et al. 1993; Martin Tremaine 1986; Wolf et al. 2010; Amorisco12 & Evans 2011). 12 The Astrophysical Journal Letters, 808:L36 (5pp), 2015 August 1 Bonnivard et al. e.g., Battaglia et al. 2013; Oswalt & Gilmore 2013; Strigari et al. (2015a) use for “ultrafaint” dSphs (see their Section 3.1). The Astrophysical Journal Letters, 808:L36 (5pp), 2015 August 1 2013). Here, we focus on the spherical Jeans analysis,Bonnivard a widely et al.We consider a flexible Zhao–Hernquist model for the 3D used approach for the determination of astrophysical factors profile, (Strigari et al. 2007; Essig et al. 2010; Charbonnier et al. 2011; n Cholis & Salucci 2012; Bonnivard et al. 2015a; Geringer- n Zhao ()r = s ,(5) The Astrophysical Journal Letters, 808:L36 (5pp), 2015 August 1 Bonnivard()bga- et al. Sameth et al. 2015a). We refer the reader to Bonnivard et al. rrgaé1 + rr ù ()ssëê ()ûú e.g.,(2015b Battaglia) for a thoroughet al. 2013 description; Oswalt & of Gilmore the analysis2013 setup; Strigari we use et al. (2015a) use for “ultrafaint” dSphs (see their Section 3.1). fifl 2013in this). Here, work. we Here, focus we on summarize the spherical the Jeans main analysis, ingredients. a widely Wewhere consider the a veexible parameters Zhao–Hernquist are the normalization model for thens 3D, the scale usedAssuming approach steady-state, for the determination spherical of symmetry, astrophysical and factors negligible profile, radius rs , the inner power law index γ, the outer index β, and (rotationalStrigari et al.support,2007; Essig the second-order et al. 2010; Charbonnier Jeans equation, et al. 2011 obtained; the transition parameter αn. Along with an additional free Cholisfrom the & collisionless Salucci 2012; Boltzmann Bonnivard equation, et al. 2015a reads; Geringer-(Binney & n Zhao ()r = s ,(5) parameter S that represents a uniform()bga- background density, Sameth et al. 2015a). We refer the reader to Bonnivard et al. bkd gaé ù Tremaine 2008) ()rrssê1 + () rr ú (2015b) for a thorough description of the analysis setup we use these parameters then specifyë a modelû for the projected stellar ¯2 density:fi in this work.1 d Here,¯ we2 summarizebani ()rv ther mainGM ingredients.() r where the ve parameters are the normalization ns , the scale The dark matter contentAssuming of steady-state,Reticulumnvr +=-2 spherical II symmetry, and,(2) negligible ¥ n dr () r r2 radius rs , the inner power law index γn,() therr outer index β, and rotational support, the second-order Jeans equation, obtained the transitionSº parametermodel ()R α 2.ò Along with andr additional+Sbkd.(6) free from the collisionless Boltzmann equation,¯2 reads (Binney & R rR22- with n ()r the stellar number density, vrr ()the radial velocity parameter S that represents a uniform background density, Tremaine 2008) bkd The Astrophysical Journal Letters, 808:L36 (5pp), 2015 August 1 ¯¯22doi:10.1088/2041-8205/808/2/L36 theseWe parametersfit this then model specify to the a model photometric for the projected catalog stellargenerated by © 2015. The American Astronomical Society. All rights reserved. dispersion, bani ()rvvº- 1 q r the velocity anisotropy, and 7 ¯2 density:Koposov et al. (2015a), which provides positions, colors, and M(r) the1 massd ¯2 enclosedbani ()rv withinr radiusGM() rr. After solving nvr +=-2 ,(2) magnitudes of individual¥ stars detected as point sources. From Equationn(dr2) ()and projectingr along ther LOS,2 the (squared) n ()rr DARK MATTERfi ANNIHILATION AND DECAY PROFILES FOR THE Sºmodel ()R 2 fi dr +Sbkd.(6) Figure 1. Projected stellar density pro le of Ret II, derived from thevelocity dispersion at the projected radius R reads the raw catalog, weòR rst identify22 possible members of Ret II as photometric catalog of Koposov et al. (2015aRETICULUM). Overplotted II(red DWARF line) is the SPHEROIDAL GALAXY ¯2 rR- with n ()r the stellar number density, vrr ()the radial velocity point sources (selected as sources with Sextractor “spread” best-fitting model (we note that the fit is to the unbinned data), which is the Vincent Bonnivard1, Céline Combet1, David Maurin1, Alex Geringer-Sameth¥ 2æ,¯¯ Savvas22 M. Koushiappas2 ö n 3¯2, We fit this model to the photometric catalog generated by sum of contributions from Ret II itself and a constant background (see Sectiondispersion,2 bani ()rvv2 º- 1 ç q r the velocityR ÷ () anisotropy,rvr () rr and parameter <0.01 in the g-band) whose extinction-corrected Matthew G. Walker2, Mario Mateo4, Edwardsb W.p ()R Olszewski= 7 5, and Johnç1()- I. Baileyani r III4 ÷ dr,(3) 2.3). Dotted lines enclose1 68% CIs for the projection of n (r). M(r) the mass enclosedòR ç within radius2 ÷ r.22 After solving Koposovgr- etcolors al. (2015a place), which them provides within positions,0.25 dex ofcolors, the and Dartmouth LPSC, Université Grenoble-Alpes, CNRS/IN2P3, 53 avenue des Martyrs,Figure F-38026S Grenoble,3.()RMedian France;(solidèç [email protected]), 68% (dashedr ø), andrR 95%- (dash–dot) CIs of the 2 McWilliams Center for Cosmology, Department of Physics, Carnegie MellonEquation University,J-((2top) Pittsburgh,)andand D projecting PA-factors 15213,(bottom USA; [email protected]) of the Ret II,LOS, as a functionthe (squared of integration) angle,magnitudesisochrone of individual(Dotter stars et al. detected2008 as), point calculated sources. for From a stellar 3 Department of Physics, Brown University, Providence,reconstructed RI 02912, from USA our Jeans/MCMC analysis. fi 4 with the surface brightness profile. We compare the LOS the raw catalog, we rst identify possible members of Ret II as University of Michigan, 311 West Hall, 1085 S.velocity UniversityS()R dispersion Avenue, Ann Arbor, at the MI projected 48109, USA radius R reads population with age 12 Gyr, metallicity [Fe H]=- 2.5, and 5 Steward Observatory, The University of Arizona,velocities 933 N. Cherry of Avenue, the Tucson, stars AZto 85721, the projected USA velocity dispersion s , point sources (selected as sources with Sextractor “spread” ¯2 p distance modulus mM-=17.4 (Koposov et al. 2015a). To Received 2015 April 13; accepted 2015 June 9; published2 2015 July¥ æ 27 R2 ö n ()rv() rr computed2 using parametricç formsTable for÷ the 1 unknownr velocity parameterthe unbinned <0.01 in distribution the g-band) whose of projected extinction-corrected positions for the sbp ()R = Astrophysicalç1()- ani Factorsr ÷ for Ret II (d = 30dr,(3) kpc) S()R òR èç r2 ø÷fi 22 gr- colors place them within 0.25 dex of the Dartmouth ABSTRACTanisotropy bani (r) and DM density pro lerRr-DM ()r . We use the N = 12470 RGB candidates identified within 1.5° of Ret log (J (a )) log (D (aisochrone )) (Dotter et al. 2008), calculated for a stellar The dwarf spheroidal galaxies (dSph) of the Milky Wayfollowing are amonga theint likelihood most attractive function targets( forStrigari10 indirectint et searches al. 2007): 10 int II’s center, we fit 2D projections of n ()r according to the with S()R the surface brightness profi25le.- Wea compare the LOS -2 b of dark matter (DM). In this work, we reconstruct the DM annihilation(deg)(J-factor) and decay(J pro GeVfiles cm for the) newly (GeVcmD population) with age 12 Gyr, metallicity [Fe H]=- 2.5, and velocities of the stars to the projected velocity dispersion s , likelihood function: discovered dSph Reticulum II. Using an optimized spherical JeansNstars analysis of kinematic-12 data obtained++é0.5( 1.1) æ from the 2 öùp ++0.6(distance 1.0) modulus mM-=17.4 (Koposov et al. 2015a). To 0.01 (2p ) 17.1--0.5( 0.9)1 (vv- ¯) ÷ 15.7--0.3( 0.5) Michigan/Magellan Fiber System, we find Reticulum II’scomputedJ-factor to be using among parametric the largest of forms any Milkyê for the Wayç unknown dSph.i velocity÷ú N  = 0.05 1exp8.3++ê0.5(- 1.1) ç ÷ú,(4)17.0++0.5(the 1.0) unbinned distribution of projected positions for the We have checked the robustness of this result against several ingredients of the22 analysis. Unless it-- suffers0.4(fi 0.8)2 fromç s22 tidal +D ÷ --0.3( 0.6) µS R .(7) anisotropyi=1bani (sr) andR DM+D density++ proê le èçr p ()rRi. We usevi ø theú ++N 12470 RGB candidates2 identimodelfied() withini 1.5 of Ret fi fl 0.1 p ()i vi 18.8 ë0.6( 1.2) DM û 17.6 0.6( 1.1)= ° disruption or signi cant in ation of its velocity dispersionfollowing from binary likelihood stars, Reticulum function II mayStrigari-- provide0.5( et 0.8) al. a unique2007 : --0.4( 0.6) fi i=1 window on DM particle properties. ( ++1.0( 1.7) ) ++0.7(II 1.2)’s center, we t 2D projections of n ()r according to the which assumes0.5 a Gaussian distribution19.6--0.7( 1.3) of LOS stellar velocities18.8--0.7( 1.1) fi – – – ++1.2( 2.0) – ++0.8(likelihood 1.4) As in Bonnivardfunction: et al. (2015a), the t is done with the software Key words: dark matter galaxies: dwarf galaxies: individualN(stars1Reticulum II)-12gamma rays:19.8é -- galaxies0.9( 1.4)æ 2 öù 19.3--0.9( 1.4) v , centered on the(2p ) mean stellarê velocity1 ç v¯(,vv withi - ¯) a dispersion÷ú of methods: statistical – stars: kinematics and dynamics i = exp - ç ÷ ,(4) package MultiNestN (Feroz & Hobson 2008; Feroz  ê ç 22÷ú velocities (at the22 radius Ri) coming2 ç from both the÷ intrinsic et al. 2009, 2013µS, and we useR the.(7) samples from the posterior i=1 s Rfi +D ê èç sp ()Ri +Dvi øú 2 ) model ()i Notes. Forp ()vei differentvi integrationë angles, the median J û(resp D)-factors as  dispersion s (R ) and the measurement uncertainty D . i=1 fi 1. INTRODUCTION stellarwell kinematic aspi their data 68% obtained and 95% CIs with are the given. Michigan/ Note thatMagellan possible triaxialityvi of thePDFs to propagate the light pro le uncertainty into the Jeans whichProbabilityFiber assumesdSph System density galaxies a( GaussianM2FS; adds functions Walkera systematicdistribution et(PDFs al.uncertainty of2015) LOSof). of the Westellar±0.4 anisotropy use( velocitiesresp. the±0.3)( andBonnivardAs inanalysis. Bonnivard Figure et al. 1(2015ashows), the fifitt is to done the withprojected the software stellar density Along with the Galactic center and Galaxy clusters, the dwarf optimizedet al. 2015b Jeans) and analysis is not includedsetup from in the Bonnivard quoted intervals. et al. fi spheroidal galaxies (dSph) of the Milky Way have been vDMi, centered parameters on the are mean obtained stellar with velocity a Markovv¯, with Chain a dispersion Monte of Carlo packagepro le MultiNest of Ret II (dashed(Feroz red & line Hobson), with the2008 contributions; Feroz from (2015aa,12015b GeV2 )cm, and8--57=´ compute2.25 10 theM astrophysical2 kpc−5. J- and D- identified as promising targets for indirect dark matter (DM) velocities(MCMC)(atengine, the radiusandR arei) coming used to from compute both the the median intrinsic and et al.Ret2009 II, itself2013 and), and from we the use constant the samples background from the posterior(solid black and factors,b for annihilating--215 and decaying DM,−2 respectively, from searches (see, e.g., Strigari 2013; Conrad et al. 2015). Their low dispersion1 GeVcm and=´8.55 the 10 measurementM kpc . uncertainty D . fi crediblethe reconstructed intervalsspi(R ) DM(CIs density) of the profi astrophysicalles. We cross-check factors our forvi any PDFsblue to propagate lines, respectively the light) pro. le uncertainty into the Jeans astrophysical background, high mass-to-light ratio, and proximity Probabilityintegrationresults by density angle. varying functions different ingredients(PDFs) of of the analysis anisotropy and and analysis.Kinematic Figure 1 datashows— thewefi uset to the the Ret projected II stellar stellar kinematic density data set make them compelling targets (Lake 1990; Evans et al. 2004). evaluate the ranking of Ret II among the most promising dSphs fi fi About 25 Galactic dSphs were known as of early 2015, and their DMFollowing parametersWalker the are etoptimized al.obtained2015 after withJeans a exclusion Markov analysis Chain of setupRet2-142 Monte proposed) Carloas our inducialpro fromle of Ret Walker II (dashed et al. ( red2015 line),), obtained with the with contributions M2FS. It from consists of for DM indirect detection.8 observation by γ-ray telescopes has thus far shown no significant (BonnivardMCMCsetup.) engine, et al. (2015band are), used the DM to compute density the is described median and by an Ret IIprojected itself and positions from the andconstant LOS background velocities( forsolid 38 black individual and stars, emission, leading to stringent constraints onásann vñ, the thermally credible intervalsfi (CIs) of the astrophysical factors for any blueas lines, well respectively as an estimation). of their membership probability P . The fi Einasto2. ASTROPHYSICAL pro le (Merritt FACTORS, et al. 2006 JEANS), and ANALYSIS, the anisotropy and i Figureaveraged 2. Top: velocity DM annihilation dispersion cross-sectionpro le of Ret( IIAcciari and reconstructed et al. 2010 median; integrationlight profiles angle. are givenAND by DATA Baes3. SETS RESULTS & van Hese (Baes & van Kinematiclatter, obtained data—we using use the an Ret expectation II stellar kinematic maximization data set algorithm and crediblePaiano intervals et al. 2011(solid; Abramowski and dashed black et al. lines,2014; respectively Geringer-Sameth), as well as fi Following the optimized– Jeans analysis setup proposed in from(Walker et et al. al. (20092015)),, quanti obtainedfies with the probabilityM2FS. It consists that a ofgiven star best ett al.(see2014 footnote; Fermi-LAT 9; long Collaboration dashed red2015 lines)).. Bottom: distributionHese of 2007Figure) and Zhao32.1.displays AstrophysicalHernquist the J- Factors((Hernquisttop) and D1990-factors; Zhao(bottom1996))of Ret membership probabilities as a function of the projected radius R andBonnivard the et al. (2015b), the DM density is described by an projectedbelongs positions to the and dSph LOS or velocities to the Milky for 38 Way individual foreground. stars, Recently, imaging data from the Dark Energy Survey have parametrizations,The differential respectively.γ-ray flux coming The from large DM freedom annihilation allowed by departure from the mean velocity (z-axis, blue to red color) for the eighteen II, reconstructedfi from the Jeans/MCMC analysis, as a functionas well as an estimation of their membership probability P . The led to the discovery of nine new potential Milky Way satellites Einastothese(resp. parametrizations pro decayle) in(Merritt a dSph wasgalaxy et al. found is2006 proportional to), andmitigate the to the anisotropy possible so-called biases and of The top panel of Figure 2 presents the velocityi dispersion stars with Pi ¹ 0. The size of the points is proportional to the velocity offi the integration angle aint. Solid lines represent the median in the Southern sky (DES Collaboration et al. 2015; Koposov lightthe“astrophysical Jeans pro les analysis are factor given(Bonnivard” J ( byresp. BaesD; et Bergström & al. van2015b Heseet al.). Finally,1998(Baes), & the van extent latter,pro obtainedfile of Ret using II, an as expectation well as its maximization reconstruction algorithm with the Jeans uncertainty. See text for discussion. values, while dashed and dash–dot lines symbolize the 68% 9 fi et al. 2015a). The nearest object, Reticulum II (Ret II, d ~ 30 Heseof the2007 DM) haloand Zhao is computed–Hernquist using(Hernquist the tidal1990 radius; Zhao estimation1996) as (Walkeranalysis. et al. 2009The) bottom, quanti paneles the of probability Figure 2 thatshows a given the distribution star kpc), is particularly intriguing, as evidence of γ-ray emission parametrizations,JldldDldld=WW=WWandrr2 95%(, respectively. CIs ) respectively. resp. The Our large data-driven freedom (, ) allowed Jeans , (1) analysis by givesbelongs to the dSph or to the Milky Way foreground. in Bonnivard∬∬DM et al. (2015a()). DM of membership probabilities as a function of the projected has been detected in its direction using the public Fermi-LAT these parametrizationslarge statistical was uncertainties found to mitigate due possible to the smallbiases of size of theThe top panel of Figure 2 presents the velocity dispersion hintsPass at Milky 7 data. Way Geringer-Sameth foreground contamination, et al. 2015b determined which can the affect radius R and the departure from the mean velocity (color- ( ) thewhich Jeanskinematic corresponds analysis (Bonnivard sample, to the integration comparable et al. 2015b along) to the. Finally, those line of theobtained sight extent for otherprofile of Ret II, as well as its reconstruction with the Jeans the Jprobability- and D-factor of background reconstruction. processes For producing Ret II, the only observed one star 2.3. Data Set coded9 ), for stars with non-zero Pi. As pointed out in Bonnivard (LOS)“ofultrafaint the DM” densitydSphs squared by Bonnivard(resp. DM densityet al. ()2015aand over; see also Figureanalysis. The bottom panel of Figure 2 shows the distribution showsRet an II intermediate gamma-ray signalP (Ret2-142 to be between in thep catalog= 0.01% ofand Walkerof the DM halo is computed using the tidal radius estimation as i the solid4). Table angle D1 W=summarizes2pa ´ [1 - our cos( resultsint )], for with thea astrophysicalint the factorset al. (2015a), a large fraction of stars with both intermediate Pi p = 1%, depending on the background modeling. An analysis in BonnivardSurface brightness et al. (2015a data).—we fit the stellar number density of membership probabilities as a function of the projected et al. 2015, with Pi = 0.69), with a very small departure from integrationof Ret angle. II. This quantity depends on both the extent of (0.05< 0.95 (16 likely members, one less than identi ed by Bonnivard et al. 2015a) of the kinematic6 data, and nd data, finding a p value of 0.16%. specifically developed to perform the Jeans analysis. 9 The binned data and associated velocity dispersion reconstruction are only In any case, a robust determination of Ret II’s DM content is 7 The mass is dominated by DM, and we neglect the stellar component. 2 fi 3 2.2. Jeans Analysis shown for illustration purposes. The nal results are obtained with an unbinned crucial in order to constrain particle nature of DM. Ret II was 8 We use the GreAT toolkit (Putze 2011; Putze & Derome 2014). analysis. found to be a DM-dominated dSph galaxy from the Several approaches have been developed to infer the DM independent chemodynamical analyses of Walker et al. density profile of dSph galaxies from stellar kinematics (see, (2015), Simon et al. (2015), and Koposov et al. (2015b). 2 Here, we reconstruct the DM annihilation and decay emission 6 This upgrade will be publicly available in the new version of the software profiles of Ret II from a spherical Jeans analysis applied to (V. Bonnivard et al. 2015, in preparation).

1 Astrophysical factors for Ret IIAstrophysical factors for Ret II 3 3 10 10 Median Reticulum II Median Reticulum II Table 1 Table 1 68% CIs 68% CIs Astrophysical factors for Ret II (d =32kpc).ForfivediAstrophysical factors↵ forerent Ret II (d =32kpc).Forfivedi↵erent 8 8 Data Data 95% CIs 95% CIs integration angles,The dark the medianmatterJ content(respintegrationD)-factors of Reticulum angles, as well the as medianII their J (resp D)-factors as well as their Best fit Best fit 68% and 95% CIs are given. Note that68% possible and 95% triaxiality CIs are given. of the Note that possible triaxiality of the 6 6 dSph galaxies adds a systematic uncertaintydSph galaxies of adds0.4(resp. a systematic0.3) uncertainty of 0.4(resp. 0.3) (Bonnivard et al. 2015a) and is not(Bonnivard included± in et the al. quoted 2015a)± and is not included± in the quoted± [km/s] [km/s] p p Bonnivard et al. ApJL 808intervals. L36 (2015) Simon et al. ApJ 808, intervals.95 (2015) σ 4 σ 4

↵int log10(J(↵int)) log↵10int(D(↵int))log10(J(↵int)) log10(D(↵int)) 2 2 2 5 a 2 b 2 5 a 2 b [deg] [J/GeV cm ] [D/[deg]GeV cm [J/] GeV cm ] [D/GeV cm ] +0.5(+1.1) +0.5(+1.0) +0.5(+1.1) +0.5(+1.0) 0.01 16.9 0.4( 0.8) 15.0.016 0.3( 0.5) 16.9 0.4( 0.8) 15.6 0.3( 0.5) 0.05 0.1 0.15 0.05 0.2 0.1 0.15 0.2 R [kpc] R [kpc] +0.5(+1.0) 0.05+0.6(+1.0) 18.2+0.5(+1.0) 17.0+0.6(+1.0) 0.05 18.2 0.4( 0.7) 17.0 0.3( 0.5) 0.4( 0.7) 0.3( 0.5) 1 1 14 14 +0.6(+1.1) +0.6(+1.1) 0.1 18.6+0.6(+1.1) 17.0.15+0.6(+1.1) 18.6 17.5 ] ] 0.4( 0.8) 0.4( 0.6) 0.4( 0.8) 0.4( 0.6) 12 -1 0.2 18.8 0.6 12 -1 +1.0(+1.6) +0.7(+1.2) +1.±0(+1.6) +0.7(+1.2) 0.5 19.5 18.0.58 19.5 18.8 10 10 0.6( 1.3) 0.50.7( 1.1) 18.90.6(0.16.3) 0.7( 1.1) Reticulum II Reticulum II ± +1.2(+2.0) +0.9(+1.4) +1.2(+2.0) +0.9(+1.4) i i = 64.8 km/s = 64.8 km/s8 119.87 19.2119.7 0.9( 1.5) 19.2 0.9( 1.4)

P 0.9( 1.5) 0.9( 1.4) P 0.5 0.5 v 6 ∆ vi ∆ i 6 a 2 5 7 2 a 5 2 5 7 2 5

-| [km s 1GeV cm =2.25 10 M kpc 1 km/s 1 km/s -| [km s 1GeV cm =2.25 10 M kpc i

i 4 4 b 2 ⇥ 15 b 2 2 ⇥ 15 2 3 km/s 3 km/s 1GeVcm =8.55 10 M kpc1GeVcm =8.55 10 M kpc |v |v 2 ⇥2 ⇥ by Walker et al. (2015). Thereforeby Walker we do et not al. (2015). expect a Therefore we do not expect a -2 10-2 strong sensitivity to foreground contamination. In this 10 R [kpc] R [kpc] strong sensitivity to foreground contamination. In this study, and as advocated in Bonnivardstudy, and et as al. advocated (2015b), we in Bonnivard et al. (2015b), we Figure 2. Top: velocity dispersionFigure profile 2. of RetTop: IIvelocity and recon- dispersionuse profile the of data Ret withII andP recon-i > 0.95 (seventeenuse the data likely with members,Pi > 0.95 (seventeen likely members, structed median and credible intervalsstructed (solid and median dashed and black credible lines intervalsone (solid less and than dashed identified black lines by Walkerone less et al. than 2015 identified after exclu- by Walker et al. 2015 after exclu- respectively), as well as best fit (longrespectively), dashed red as lines, well shown as best for fit (long dashed red lines, shown for sion of Ret2-142) as our fiducial setup. illustration purposes only). Bottom:illustrationdistribution purposes of membership only). Bottom:siondistribution of Ret2-142) of membership as our fiducial setup. probabilities as a function of the projectedprobabilities radius asR a functionand the ofde- the projected radius R and the de- parture from the mean velocity (z-axis,parture blue from to red the color) mean for velocity the (z-axis, blue to red color) for the3. RESULTS 3. RESULTS nineteen stars with P =0.Thesizeofthepointsisproportional nineteen stars with Pi =0.Thesizeofthepointsisproportionali 6 Figure 3 displays the J- (top) and D-factors (bottom) to the velocity uncertainty.6 See textto for the discussion. velocity uncertainty. See text forFigure discussion. 3 displays the J- (top) and D-factors (bottom) of Ret II, reconstructed fromof the Ret Jeans/MCMC II, reconstructed analy- from the Jeans/MCMC analy- sis, as a function of the integrationsis, as angle a function↵int. Solid of the lines integration angle ↵int. Solid lines et al. 2009, 2013), and we use theet al. samples 2009, 2013), from the and pos- we use therepresent samples the from median the pos- values, whilerepresent dashed the and median dash-dot values, while dashed and dash-dot terior PDFs to propagate the lightterior profile PDFs uncertainty to propagate in the lightlines profile symbolize uncertainty the 68% in and 95%lines CIs symbolize respectively. the 68% Our and 95% CIs respectively. Our the Jeans analysis. Figure 1 showsthe Jeans the fit analysis. to the projected Figure 1 showsdata-driven the fit to the Jeans projected analysis givesdata-driven large statistical Jeans analysis uncer- gives large statistical uncer- stellar density profile of Ret IIstellar (solid density red line), profile with of the Ret IItainties (solid red due line), to the with small the sizetainties of the kinematic due to the sample, small size of the kinematic sample, contributions from Ret II itselfcontributions and from the from constant Ret II itselfand and reflects from theour restrictedconstant knowledgeand reflects of the our DM restricted content knowledge of the DM content background (solid black and bluebackground lines respectively). (solid black and blueof this lines object. respectively). The CIs are comparableof this object. to those The obtained CIs are comparable to those obtained for other ‘ultrafaint’ dSphs byfor Bonnivard other ‘ultrafaint’ et al. (2015b) dSphs by Bonnivard et al. (2015b) Kinematic data — We use theKinematic Ret II stellar data — kinematicWe use the(see Ret also II stellar Figure kinematic 4). Table 1 summarizes(see also Figure our results 4). Table for 1 summarizes our results for data set from Walker et al. (2015),data obtained set from Walker with M2FS. et al. (2015),the obtained astrophysical with factors M2FS. of Retthe II. astrophysical factors of Ret II. It consists of projected positionsIt consists and l.o.s. of velocities projected for positionsWe and cross-checked l.o.s. velocities our for findingsWe by cross-checked varying di↵erent our in- findings by varying di↵erent in- 38 individual stars, as well as an38 estimation individual of stars, their as mem- well as angredients estimation of of the their Jeans mem- analysis.gredients The resulting of the JeansJ-factors analysis. The resulting J-factors bership probability P . The latter, obtained using an ex- bership probability Pi. The latter, obtained using ani ex- are shown in Figure 4. First, weare ran shown the analysis in Figure using 4. First, all we ran the analysis using all pectation maximization algorithmpectation (Walker maximization et al. 2009), algorithm38 stars (Walker of the et al. sample, 2009), but weighting38 stars of the the log-likelihood sample, but weighting the log-likelihood quantifies the probability that aquantifies given star the belongs probability to the that afunction given star of belongs equation to (4) the by thefunction membership of equation probabilities (4) by the membership probabilities dSph or to the Milky Way foreground. P . Bonnivard et al. (2015b) find that a large di↵erence dSph or to the Milky Way foreground. Pi. Bonnivard et al. (2015b) findi that a large di↵erence The top panel of Figure 2 presentsThe topthe velocitypanel of disper- Figure 2 presentsbetween the a velocityP -weighted disper- and abetweenP > 0.95 a P analysisi-weighted is an- and a Pi > 0.95 analysis is an- sion profile of Ret II, as well as its reconstructioni with the otheri hint of contamination by Milky Way foreground sion profile of Ret II, as well as its reconstruction8 with the other hint of contamination by Milky Way foreground Jeans analysis8. The bottom panelJeans of analysis Figure 2. shows The bottom the panelstars. of Figure Here, the 2 shows two the analysesstars. give very Here, similar the two results, analyses give very similar results, distribution of membership probabilitiesdistribution as of a membership function probabilitiessuggesting aas clean a function sample forsuggesting Ret II. We a then clean randomly sample for Ret II. We then randomly of the projected radius R andof the the departure projected from radius theR anddivide the departure the kinematic from sample the individe two parts, the kinematic using one sample out of in two parts, using one out of mean velocity (color-coded from blue to red), for stars mean velocity (color-coded from blue to red), for stars every two stars to build the twoevery sub-samples. two stars to Applying build the two sub-samples. Applying with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- (2015b), a large fraction of stars with both intermediate (2015b), a large fraction of stars with both intermediate and D-factors, which confirmsand thatD the-factors, reconstruction which confirms of that the reconstruction of Pi (0.1

↵int log10(J(↵int)) log↵10int(D(↵int))log10(J(↵int)) log10(D(↵int)) 2 2 2 5 a 2 b 2 5 a 2 b [deg] [J/GeV cm ] [D/[deg]GeV cm [J/] GeV cm ] [D/GeV cm ] +0.5(+1.1) +0.5(+1.0) +0.5(+1.1) +0.5(+1.0) 0.01 16.9 0.4( 0.8) 15.0.016 0.3( 0.5) 16.9 0.4( 0.8) 15.6 0.3( 0.5) 0.05 0.1 0.15 0.05 0.2 0.1 0.15 0.2 R [kpc] R [kpc] +0.5(+1.0) 0.05+0.6(+1.0) 18.2+0.5(+1.0) 17.0+0.6(+1.0) 0.05 18.2 0.4( 0.7) 17.0 0.3( 0.5) 0.4( 0.7) 0.3( 0.5) 1 1 14 14 +0.6(+1.1) +0.6(+1.1) 0.1 18.6+0.6(+1.1) 17.0.15+0.6(+1.1) 18.6 17.5 ] ] 0.4( 0.8) 0.4( 0.6) 0.4( 0.8) 0.4( 0.6) 12 -1 0.2 18.8 0.6 12 -1 +1.0(+1.6) +0.7(+1.2) +1.±0(+1.6) +0.7(+1.2) 0.5 19.5 18.0.58 19.5 18.8 10 10 0.6( 1.3) 0.50.7( 1.1) 18.90.6(0.16.3) 0.7( 1.1) Reticulum II Reticulum II ± +1.2(+2.0) +0.9(+1.4) +1.2(+2.0) +0.9(+1.4) i i = 64.8 km/s = 64.8 km/s8 119.87 19.2119.7 0.9( 1.5) 19.2 0.9( 1.4)

P 0.9( 1.5) 0.9( 1.4) P 0.5 0.5 v 6 ∆ vi ∆ i 6 a 2 5 7 2 a 5 2 5 7 2 5

-| [km s 1GeV cm =2.25 10 M kpc 1 km/s 1 km/s -| [km s 1GeV cm =2.25 10 M kpc i

i 4 4 b 2 ⇥ 15 b 2 2 ⇥ 15 2 3 km/s 3 km/s 1GeVcm =8.55 10 M kpc1GeVcm =8.55 10 M kpc |v |v 2 ⇥2 ⇥ by Walker et al. (2015). Thereforeby Walker we do et not al. (2015). expect a Therefore we do not expect a -2 10-2 strong sensitivity to foreground contamination. In this 10 R [kpc] R [kpc] strong sensitivity to foreground contamination. In this study, and as advocated in Bonnivardstudy, and et as al. advocated (2015b), we in Bonnivard et al. (2015b), we Figure 2. Top: velocity dispersionFigure profile 2. of RetTop: IIvelocity and recon- dispersionuse profile the of data Ret withII andP recon-i > 0.95 (seventeenuse the data likely with members,Pi > 0.95 (seventeen likely members, structed median and credible intervalsstructed (solid and median dashed and black credible lines intervalsone (solid less and than dashed identified black lines by Walkerone less et al. than 2015 identified after exclu- by Walker et al. 2015 after exclu- respectively), as well as best fit (longrespectively), dashed red as lines, well shown as best for fit (long dashed red lines, shown for sion of Ret2-142) as our fiducial setup. illustration purposes only). Bottom:illustrationdistribution purposes of membership only). Bottom:siondistribution of Ret2-142) of membership as our fiducial setup. probabilities as a function of the projectedprobabilities radius asR a functionand the ofde- the projected radius R and the de- parture from the mean velocity (z-axis,parture blue from to red the color) mean for velocity the (z-axis, blue to red color) for the3. RESULTS 3. RESULTS nineteen stars with P =0.Thesizeofthepointsisproportional nineteen stars with Pi =0.Thesizeofthepointsisproportionali 6 Figure 3 displays the J- (top) and D-factors (bottom) to the velocity uncertainty.6 See textto for the discussion. velocity uncertainty. See text forFigure discussion. 3 displays the J- (top) and D-factors (bottom) of Ret II, reconstructed fromof the Ret Jeans/MCMC II, reconstructed analy- from the Jeans/MCMC analy- sis, as a function of the integrationsis, as angle a function↵int. Solid of the lines integration angle ↵int. Solid lines et al. 2009, 2013), and we use theet al. samples 2009, 2013), from the and pos- we use therepresent samples the from median the pos- values, whilerepresent dashed the and median dash-dot values, while dashed and dash-dot terior PDFs to propagate the lightterior profile PDFs uncertainty to propagate in the lightlines profile symbolize uncertainty the 68% in and 95%lines CIs symbolize respectively. the 68% Our and 95% CIs respectively. Our the Jeans analysis. Figure 1 showsthe Jeans the fit analysis. to the projected Figure 1 showsdata-driven the fit to the Jeans projected analysis givesdata-driven large statistical Jeans analysis uncer- gives large statistical uncer- stellar density profile of Ret IIstellar (solid density red line), profile with of the Ret IItainties (solid red due line), to the with small the sizetainties of the kinematic due to the sample, small size of the kinematic sample, contributions from Ret II itselfcontributions and from the from constant Ret II itselfand and reflects from theour restrictedconstant knowledgeand reflects of the our DM restricted content knowledge of the DM content background (solid black and bluebackground lines respectively). (solid black and blueof this lines object. respectively). The CIs are comparableof this object. to those The obtained CIs are comparable to those obtained for other ‘ultrafaint’ dSphs byfor Bonnivard other ‘ultrafaint’ et al. (2015b) dSphs by Bonnivard et al. (2015b) Kinematic data — We use theKinematic Ret II stellar data — kinematicWe use the(see Ret also II stellar Figure kinematic 4). Table 1 summarizes(see also Figure our results 4). Table for 1 summarizes our results for data set from Walker et al. (2015),data obtained set from Walker with M2FS. et al. (2015),the obtained astrophysical with factors M2FS. of Retthe II. astrophysical factors of Ret II. It consists of projected positionsIt consists and l.o.s. of velocities projected for positionsWe and cross-checked l.o.s. velocities our for findingsWe by cross-checked varying di↵erent our in- findings by varying di↵erent in- 38 individual stars, as well as an38 estimation individual of stars, their as mem- well as angredients estimation of of the their Jeans mem- analysis.gredients The resulting of the JeansJ-factors analysis. The resulting J-factors bership probability P . The latter, obtained using an ex- bership probability Pi. The latter, obtained using ani ex- are shown in Figure 4. First, weare ran shown the analysis in Figure using 4. First, all we ran the analysis using all pectation maximization algorithmpectation (Walker maximization et al. 2009), algorithm38 stars (Walker of the et al. sample, 2009), but weighting38 stars of the the log-likelihood sample, but weighting the log-likelihood quantifies the probability that aquantifies given star the belongs probability to the that afunction given star of belongs equation to (4) the by thefunction membership of equation probabilities (4) by the membership probabilities dSph or to the Milky Way foreground. P . Bonnivard et al. (2015b) find that a large di↵erence dSph or to the Milky Way foreground. Pi. Bonnivard et al. (2015b) findi that a large di↵erence The top panel of Figure 2 presentsThe topthe velocitypanel of disper- Figure 2 presentsbetween the a velocityP -weighted disper- and abetweenP > 0.95 a P analysisi-weighted is an- and a Pi > 0.95 analysis is an- sion profile of Ret II, as well as its reconstructioni with the otheri hint of contamination by Milky Way foreground sion profile of Ret II, as well as its reconstruction8 with the other hint of contamination by Milky Way foreground Jeans analysis8. The bottom panelJeans of analysis Figure 2. shows The bottom the panelstars. of Figure Here, the 2 shows two the analysesstars. give very Here, similar the two results, analyses give very similar results, distribution of membership probabilitiesdistribution as of a membership function probabilitiessuggesting aas clean a function sample forsuggesting Ret II. We a then clean randomly sample for Ret II. We then randomly of the projected radius R andof the the departure projected from radius theR anddivide the departure the kinematic from sample the individe two parts, the kinematic using one sample out of in two parts, using one out of mean velocity (color-coded from blue to red), for stars mean velocity (color-coded from blue to red), for stars every two stars to build the twoevery sub-samples. two stars to Applying build the two sub-samples. Applying with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- (2015b), a large fraction of stars with both intermediate (2015b), a large fraction of stars with both intermediate and D-factors, which confirmsand thatD the-factors, reconstruction which confirms of that the reconstruction of Pi (0.1

↵int log10(J(↵int)) log↵10int(D(↵int))log10(J(↵int)) log10(D(↵int)) 2 2 2 5 a 2 b 2 5 a 2 b [deg] [J/GeV cm ] [D/[deg]GeV cm [J/] GeV cm ] [D/GeV cm ] +0.5(+1.1) +0.5(+1.0) +0.5(+1.1) +0.5(+1.0) 0.01 16.9 0.4( 0.8) 15.0.016 0.3( 0.5) 16.9 0.4( 0.8) 15.6 0.3( 0.5) 0.05 0.1 0.15 0.05 0.2 0.1 0.15 0.2 R [kpc] R [kpc] +0.5(+1.0) 0.05+0.6(+1.0) 18.2+0.5(+1.0) 17.0+0.6(+1.0) 0.05 18.2 0.4( 0.7) 17.0 0.3( 0.5) 0.4( 0.7) 0.3( 0.5) 1 1 14 14 +0.6(+1.1) +0.6(+1.1) 0.1 18.6+0.6(+1.1) 17.0.15+0.6(+1.1) 18.6 17.5 ] ] 0.4( 0.8) 0.4( 0.6) 0.4( 0.8) 0.4( 0.6) 12 -1 0.2 18.8 0.6 12 -1 +1.0(+1.6) +0.7(+1.2) +1.±0(+1.6) +0.7(+1.2) 0.5 19.5 18.0.58 19.5 18.8 10 10 0.6( 1.3) 0.50.7( 1.1) 18.90.6(0.16.3) 0.7( 1.1) Reticulum II Reticulum II ± +1.2(+2.0) +0.9(+1.4) +1.2(+2.0) +0.9(+1.4) i i = 64.8 km/s = 64.8 km/s8 119.87 19.2119.7 0.9( 1.5) 19.2 0.9( 1.4)

P 0.9( 1.5) 0.9( 1.4) P 0.5 0.5 v 6 ∆ vi ∆ i 6 a 2 5 7 2 a 5 2 5 7 2 5

-| [km s 1GeV cm =2.25 10 M kpc 1 km/s 1 km/s -| [km s 1GeV cm =2.25 10 M kpc i

i 4 4 b 2 ⇥ 15 b 2 2 ⇥ 15 2 3 km/s 3 km/s 1GeVcm =8.55 10 M kpc1GeVcm =8.55 10 M kpc |v |v 2 ⇥2 ⇥ by Walker et al. (2015). Thereforeby Walker we do et not al. (2015). expect a Therefore we do not expect a -2 10-2 strong sensitivity to foreground contamination. In this 10 R [kpc] R [kpc] strong sensitivity to foreground contamination. In this study, and as advocated in Bonnivardstudy, and et as al. advocated (2015b), we in Bonnivard et al. (2015b), we Figure 2. Top: velocity dispersionFigure profile 2. of RetTop: IIvelocity and recon- dispersionuse profile the of data Ret withII andP recon-i > 0.95 (seventeenuse the data likely with members,Pi > 0.95 (seventeen likely members, structed median and credible intervalsstructed (solid and median dashed and black credible lines intervalsone (solid less and than dashed identified black lines by Walkerone less et al. than 2015 identified after exclu- by Walker et al. 2015 after exclu- respectively), as well as best fit (longrespectively), dashed red as lines, well shown as best for fit (long dashed red lines, shown for sion of Ret2-142) as our fiducial setup. illustration purposes only). Bottom:illustrationdistribution purposes of membership only). Bottom:siondistribution of Ret2-142) of membership as our fiducial setup. probabilities as a function of the projectedprobabilities radius asR a functionand the ofde- the projected radius R and the de- parture from the mean velocity (z-axis,parture blue from to red the color) mean for velocity the (z-axis, blue to red color) for the3. RESULTS 3. RESULTS nineteen stars with P =0.Thesizeofthepointsisproportional nineteen stars with Pi =0.Thesizeofthepointsisproportionali 6 Figure 3 displays the J- (top) and D-factors (bottom) to the velocity uncertainty.6 See textto for the discussion. velocity uncertainty. See text forFigure discussion. 3 displays the J- (top) and D-factors (bottom) of Ret II, reconstructed fromof the Ret Jeans/MCMC II, reconstructed analy- from the Jeans/MCMC analy- sis, as a function of the integrationsis, as angle a function↵int. Solid of the lines integration angle ↵int. Solid lines et al. 2009, 2013), and we use theet al. samples 2009, 2013), from the and pos- we use therepresent samples the from median the pos- values, whilerepresent dashed the and median dash-dot values, while dashed and dash-dot terior PDFs to propagate the lightterior profile PDFs uncertainty to propagate in the lightlines profile symbolize uncertainty the 68% in and 95%lines CIs symbolize respectively. the 68% Our and 95% CIs respectively. Our the Jeans analysis. Figure 1 showsthe Jeans the fit analysis. to the projected Figure 1 showsdata-driven the fit to the Jeans projected analysis givesdata-driven large statistical Jeans analysis uncer- gives large statistical uncer- stellar density profile of Ret IIstellar (solid density red line), profile with of the Ret IItainties (solid red due line), to the with small the sizetainties of the kinematic due to the sample, small size of the kinematic sample, contributions from Ret II itselfcontributions and from the from constant Ret II itselfand and reflects from theour restrictedconstant knowledgeand reflects of the our DM restricted content knowledge of the DM content background (solid black and bluebackground lines respectively). (solid black and blueof this lines object. respectively). The CIs are comparableof this object. to those The obtained CIs are comparable to those obtained for other ‘ultrafaint’ dSphs byfor Bonnivard other ‘ultrafaint’ et al. (2015b) dSphs by Bonnivard et al. (2015b) Kinematic data — We use theKinematic Ret II stellar data — kinematicWe use the(see Ret also II stellar Figure kinematic 4). Table 1 summarizes(see also Figure our results 4). Table for 1 summarizes our results for data set from Walker et al. (2015),data obtained set from Walker with M2FS. et al. (2015),the obtained astrophysical with factors M2FS. of Retthe II. astrophysical factors of Ret II. It consists of projected positionsIt consists and l.o.s. of velocities projected for positionsWe and cross-checked l.o.s. velocities our for findingsWe by cross-checked varying di↵erent our in- findings by varying di↵erent in- 38 individual stars, as well as an38 estimation individual of stars, their as mem- well as angredients estimation of of the their Jeans mem- analysis.gredients The resulting of the JeansJ-factors analysis. The resulting J-factors bership probability P . The latter, obtained using an ex- bership probability Pi. The latter, obtained using ani ex- are shown in Figure 4. First, weare ran shown the analysis in Figure using 4. First, all we ran the analysis using all pectation maximization algorithmpectation (Walker maximization et al. 2009), algorithm38 stars (Walker of the et al. sample, 2009), but weighting38 stars of the the log-likelihood sample, but weighting the log-likelihood quantifies the probability that aquantifies given star the belongs probability to the that afunction given star of belongs equation to (4) the by thefunction membership of equation probabilities (4) by the membership probabilities dSph or to the Milky Way foreground. P . Bonnivard et al. (2015b) find that a large di↵erence dSph or to the Milky Way foreground. Pi. Bonnivard et al. (2015b) findi that a large di↵erence The top panel of Figure 2 presentsThe topthe velocitypanel of disper- Figure 2 presentsbetween the a velocityP -weighted disper- and abetweenP > 0.95 a P analysisi-weighted is an- and a Pi > 0.95 analysis is an- sion profile of Ret II, as well as its reconstructioni with the otheri hint of contamination by Milky Way foreground sion profile of Ret II, as well as its reconstruction8 with the other hint of contamination by Milky Way foreground Jeans analysis8. The bottom panelJeans of analysis Figure 2. shows The bottom the panelstars. of Figure Here, the 2 shows two the analysesstars. give very Here, similar the two results, analyses give very similar results, distribution of membership probabilitiesdistribution as of a membership function probabilitiessuggesting aas clean a function sample forsuggesting Ret II. We a then clean randomly sample for Ret II. We then randomly of the projected radius R andof the the departure projected from radius theR anddivide the departure the kinematic from sample the individe two parts, the kinematic using one sample out of in two parts, using one out of mean velocity (color-coded from blue to red), for stars mean velocity (color-coded from blue to red), for stars every two stars to build the twoevery sub-samples. two stars to Applying build the two sub-samples. Applying with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- (2015b), a large fraction of stars with both intermediate (2015b), a large fraction of stars with both intermediate and D-factors, which confirmsand thatD the-factors, reconstruction which confirms of that the reconstruction of Pi (0.1

↵int log10(J(↵int)) log↵10int(D(↵int))log10(J(↵int)) log10(D(↵int)) 2 2 2 5 a 2 b 2 5 a 2 b [deg] [J/GeV cm ] [D/[deg]GeV cm [J/] GeV cm ] [D/GeV cm ] +0.5(+1.1) +0.5(+1.0) +0.5(+1.1) +0.5(+1.0) 0.01 16.9 0.4( 0.8) 15.0.016 0.3( 0.5) 16.9 0.4( 0.8) 15.6 0.3( 0.5) 0.05 0.1 0.15 0.05 0.2 0.1 0.15 0.2 R [kpc] R [kpc] +0.5(+1.0) 0.05+0.6(+1.0) 18.2+0.5(+1.0) 17.0+0.6(+1.0) 0.05 18.2 0.4( 0.7) 17.0 0.3( 0.5) 0.4( 0.7) 0.3( 0.5) 1 1 14 14 +0.6(+1.1) +0.6(+1.1) 0.1 18.6+0.6(+1.1) 17.0.15+0.6(+1.1) 18.6 17.5 ] ] 0.4( 0.8) 0.4( 0.6) 0.4( 0.8) 0.4( 0.6) 12 -1 0.2 18.8 0.6 12 -1 +1.0(+1.6) +0.7(+1.2) +1.±0(+1.6) +0.7(+1.2) 0.5 19.5 18.0.58 19.5 18.8 10 10 0.6( 1.3) 0.50.7( 1.1) 18.90.6(0.16.3) 0.7( 1.1) Reticulum II Reticulum II ± +1.2(+2.0) +0.9(+1.4) +1.2(+2.0) +0.9(+1.4) i i = 64.8 km/s = 64.8 km/s8 119.87 19.2119.7 0.9( 1.5) 19.2 0.9( 1.4)

P 0.9( 1.5) 0.9( 1.4) P 0.5 0.5 v 6 ∆ vi ∆ i 6 a 2Flexible5 profile 7 2 a 5 2 Plummer5 profile 7 2 5

-| [km s 1GeV cm =2.25 10 M kpc 1 km/s 1 km/s -| [km s 1GeV cm =2.25 10 M kpc i

i 4 4 b No artificial2 truncation⇥ 15 b 2 Truncation2 at 1⇥ kpc 15 2 3 km/s 3 km/s 1GeVcm =8.55 10 M kpc1GeVcm =8.55 10 M kpc |v |v 2 No assumption on⇥2 the distribution Gaussian approximation⇥ by WalkerJ is the et peak al. of (2015). the distribution Thereforeby Walker weJ is do etthe not al.peak (2015). expect of the Gaussian a Therefore we do not expect a -2 10-2 Error is percentiles strong sensitivityError is standard to foregrounddeviation contamination. In this 10 R [kpc] R [kpc] strong sensitivity to foreground contamination. In this study, and as advocated in Bonnivardstudy, and et as al. advocated (2015b), we in Bonnivard et al. (2015b), we Figure 2. Top: velocity dispersionFigure profile 2. of RetTop: IIvelocity and recon- dispersionuse profile the of data Ret withII andP recon-i > 0.95 (seventeenuse the data likely with members,Pi > 0.95 (seventeen likely members, structed median and credible intervalsstructed (solid and median dashed and black credible lines intervalsone (solid less and than dashed identified black lines by Walkerone less et al. than 2015 identified after exclu- by Walker et al. 2015 after exclu- respectively), as well as best fit (longrespectively), dashed red as lines, well shown as best for fit (long dashed red lines, shown for sion of Ret2-142) as our fiducial setup. illustration purposes only). Bottom:illustrationdistribution purposes of membership only). Bottom:siondistribution of Ret2-142) of membership as our fiducial setup. probabilities as a function of the projectedprobabilities radius asR a functionand the ofde- the projected radius R and the de- parture from the mean velocity (z-axis,parture blue from to red the color) mean for velocity the (z-axis, blue to red color) for the3. RESULTS 3. RESULTS nineteen stars with P =0.Thesizeofthepointsisproportional nineteen stars with Pi =0.Thesizeofthepointsisproportionali 6 Figure 3 displays the J- (top) and D-factors (bottom) to the velocity uncertainty.6 See textto for the discussion. velocity uncertainty. See text forFigure discussion. 3 displays the J- (top) and D-factors (bottom) of Ret II, reconstructed fromof the Ret Jeans/MCMC II, reconstructed analy- from the Jeans/MCMC analy- sis, as a function of the integrationsis, as angle a function↵int. Solid of the lines integration angle ↵int. Solid lines et al. 2009, 2013), and we use theet al. samples 2009, 2013), from the and pos- we use therepresent samples the from median the pos- values, whilerepresent dashed the and median dash-dot values, while dashed and dash-dot terior PDFs to propagate the lightterior profile PDFs uncertainty to propagate in the lightlines profile symbolize uncertainty the 68% in and 95%lines CIs symbolize respectively. the 68% Our and 95% CIs respectively. Our the Jeans analysis. Figure 1 showsthe Jeans the fit analysis. to the projected Figure 1 showsdata-driven the fit to the Jeans projected analysis givesdata-driven large statistical Jeans analysis uncer- gives large statistical uncer- stellar density profile of Ret IIstellar (solid density red line), profile with of the Ret IItainties (solid red due line), to the with small the sizetainties of the kinematic due to the sample, small size of the kinematic sample, contributions from Ret II itselfcontributions and from the from constant Ret II itselfand and reflects from theour restrictedconstant knowledgeand reflects of the our DM restricted content knowledge of the DM content background (solid black and bluebackground lines respectively). (solid black and blueof this lines object. respectively). The CIs are comparableof this object. to those The obtained CIs are comparable to those obtained for other ‘ultrafaint’ dSphs byfor Bonnivard other ‘ultrafaint’ et al. (2015b) dSphs by Bonnivard et al. (2015b) Kinematic data — We use theKinematic Ret II stellar data — kinematicWe use the(see Ret also II stellar Figure kinematic 4). Table 1 summarizes(see also Figure our results 4). Table for 1 summarizes our results for data set from Walker et al. (2015),data obtained set from Walker with M2FS. et al. (2015),the obtained astrophysical with factors M2FS. of Retthe II. astrophysical factors of Ret II. It consists of projected positionsIt consists and l.o.s. of velocities projected for positionsWe and cross-checked l.o.s. velocities our for findingsWe by cross-checked varying di↵erent our in- findings by varying di↵erent in- 38 individual stars, as well as an38 estimation individual of stars, their as mem- well as angredients estimation of of the their Jeans mem- analysis.gredients The resulting of the JeansJ-factors analysis. The resulting J-factors bership probability P . The latter, obtained using an ex- bership probability Pi. The latter, obtained using ani ex- are shown in Figure 4. First, weare ran shown the analysis in Figure using 4. First, all we ran the analysis using all pectation maximization algorithmpectation (Walker maximization et al. 2009), algorithm38 stars (Walker of the et al. sample, 2009), but weighting38 stars of the the log-likelihood sample, but weighting the log-likelihood quantifies the probability that aquantifies given star the belongs probability to the that afunction given star of belongs equation to (4) the by thefunction membership of equation probabilities (4) by the membership probabilities dSph or to the Milky Way foreground. P . Bonnivard et al. (2015b) find that a large di↵erence dSph or to the Milky Way foreground. Pi. Bonnivard et al. (2015b) findi that a large di↵erence The top panel of Figure 2 presentsThe topthe velocitypanel of disper- Figure 2 presentsbetween the a velocityP -weighted disper- and abetweenP > 0.95 a P analysisi-weighted is an- and a Pi > 0.95 analysis is an- sion profile of Ret II, as well as its reconstructioni with the otheri hint of contamination by Milky Way foreground sion profile of Ret II, as well as its reconstruction8 with the other hint of contamination by Milky Way foreground Jeans analysis8. The bottom panelJeans of analysis Figure 2. shows The bottom the panelstars. of Figure Here, the 2 shows two the analysesstars. give very Here, similar the two results, analyses give very similar results, distribution of membership probabilitiesdistribution as of a membership function probabilitiessuggesting aas clean a function sample forsuggesting Ret II. We a then clean randomly sample for Ret II. We then randomly of the projected radius R andof the the departure projected from radius theR anddivide the departure the kinematic from sample the individe two parts, the kinematic using one sample out of in two parts, using one out of mean velocity (color-coded from blue to red), for stars mean velocity (color-coded from blue to red), for stars every two stars to build the twoevery sub-samples. two stars to Applying build the two sub-samples. Applying with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- with non-zero Pi. As pointed out in Bonnivard et al. the analysis to the two subsets leads to very similar J- (2015b), a large fraction of stars with both intermediate (2015b), a large fraction of stars with both intermediate and D-factors, which confirmsand thatD the-factors, reconstruction which confirms of that the reconstruction of Pi (0.1

Via Lactea and Ghalo simulations

Figure thanks to Mike Boylan-Kolchin

Flexible profile Plummer profile No artificial truncation Truncation at 1 kpc No assumption on the distribution Gaussian approximation J is the peak of the distribution J is the peak of the Gaussian Error is percentiles Error is standard deviation The dark matter content of Reticulum II

Via Lactea and Ghalo simulations

Figure thanks to Mike Boylan-Kolchin

Flexible profile Plummer profile No artificial truncation Truncation at 1 kpc No assumption on the distribution Gaussian approximation J is the peak of the distribution J is the peak of the Gaussian Error is percentiles Error is standard deviation The dark matter content of Reticulum II Dark Matter Searches The dark matterin Gamma content Rays of Reticulum II

20.0 Ultra-faint dSph 19.5 Classical dSph ) 5 19.0

cm Important: Not all J values 2 shown are calculated in the 18.5 same way!

(J) (GeV 18.0 10 log 17.5

17.0 1.0 0.5 0.0 0.5 Residual From Alex Drlica-Wagnerʼs talk at Texas A&M Arrowitt symposium, May, 2015 1.0 50 100 150 200 250 300 350 Distance (kpc) 34 arXiv:1503.02632 Dark Matter Searches The dark matterin Gamma content Rays of Reticulum II

Simon et al. 20.0 ApJ 808, 95 (2015) Ultra-faint dSph 19.5 Classical dSph ) 5 19.0

cm Important: Not all J values 2 shown are calculated in the 18.5 same way!

(J) (GeV 18.0 10 log 17.5

17.0 1.0 0.5 0.0 0.5 Residual From Alex Drlica-Wagnerʼs talk at Texas A&M Arrowitt symposium, May, 2015 1.0 50 100 150 200 250 300 350 Distance (kpc) 34 arXiv:1503.02632 Dark Matter Searches The dark matterin Gamma content Rays of Reticulum II

Simon et al. 20.0 Ultra-faint dSph ApJ 808, 95 (2015) Bonnivard et al. ApJL 808 L36 19.5 Classical dSph ) 5 19.0

cm Important: Not all J values 2 shown are calculated in the 18.5 same way!

(J) (GeV 18.0 10 log 17.5

17.0 1.0 0.5 0.0 0.5 Residual From Alex Drlica-Wagnerʼs talk at Texas A&M Arrowitt symposium, May, 2015 1.0 50 100 150 200 250 300 350 Distance (kpc) 34 arXiv:1503.02632 Questions

1. Is it consistent with background?

2. Is it consistent with dark matter annihilation?

3. Is it consistent with any other source?

4. Is it something else? (e.g., instrumental/data set systematics?) (P7R vs P8)

Statistical significance of a dark matter interpretation

Background modeling 5 10

] - Diffuse 1: Fermi-LAT background averaged over 1

sr 1 degree. 1 s

2 - Diffuse 2: Fermi-LAT background averaged over

6 10 2 degrees.

[GeV cm - Empirical 1: Events in an [1-5] degree annulus from central ROI with 20% gaussian width on dF/dE 2 energy. E 7 10 100 101 102 - Empirical 2: Bin Empirical 1 events in energy. Energy [GeV]

- Background in the central 0.5 degree ROI is a Poisson random variable - Background is isotropic - Energies are drawn from a given spectrum

Based on PRL 115, 081101 & ApJL 808 L36 (2015) Statistical significance of a dark matter interpretation

See Geringer-Sameth, Koushiappas & Walker, PRD 91, 083535 (2015) for details on the methodology 33

5 5 + ⌧ ⌧ 4 4

3 3

2 2 Significance Significance

1 + + 1 µ µ qq¯ W W + Di↵use 1 Empirical 1 ⌧ ⌧ b¯b hh 0 Di↵use 2 Empirical 2 101 102 103 0 101 102 103 Mass [GeV] Mass [GeV]

FIG. 2: Significance of the -ray excess in the direction of ReticulumLocal 2 p-value as a function = of0.0000068 dark matter particle (4.4 mass.sigma)Left: Curves FIG.correspond 2: Significance to the result of the of-ray the search excess in in various the direction channels of (i.e.Reticulum using di 2↵ aserent a function ways of of weighting dark matter events) particle using mass. backgroundLeft: Curves model correspond to the result of the+ search in various+ channels (i.e. using di↵+erent ways of weighting events) using background model Di↵use 1. The curve for+e e is similar to +µ µ, ZZ is similar to W+ W ,andq represents u, d, c, s quarks and gluons. Right: Di↵use 1. The curve for+ e e is similar to µ µ, ZZ is similarGlobal to W W p-value,andq represents = 0.000042u, d, c, s (quarks3.7 sigma and gluons.) Right: SignificanceBased on inPRL the 115,+⌧ ⌧081101 channel & ApJL for 808 four L36 di↵ (2015)erent background models (see text). Significance in the ⌧ ⌧ channel for four di↵erent background models (see text).

However, it is important to consider a “trials factor” to A trials factor for the model-independent approach is However,account for it isthe important fact that to we consider are searching a “trials for factor” dark mat- to foundA trials by counting factor for the the number model-independent of background ROIs approach which is accountter particles for the of fact any that mass. we As are shown searching in Fig. for 6 dark of [41], mat- the foundhave T byvalues counting among the numberthe top n offor background any mass ROIs considered which tersearch particles is not of any particularly mass. As sensitive shown in to Fig. the 6 particle of [41], mass the have(n isT thevalues rank among of the central the top ROIn for at any the massmost significant considered + searchused is in not the particularly weight function: sensitive3 to trial the masses particle su massce if (mass).n is the For rank annihilation of the central into ROI⌧ ⌧ at, n the= 9most and significant there are ⇠ + usedthe in true the mass weight is between function: 10 GeV3 trial and masses 1 TeV forsu thece ifb¯b mass).32 such For ROIs, annihilation giving a global into p⌧-value⌧ , n of= 32/ 93306 and = there 0.0097 are + ⇠ ¯ theand true⌧ mass⌧ channels. is between Nonetheless, 10 GeV and we 1quantify TeV for the the trialsbb 32(2. such3). ROIs, The same giving global a global significancep-value is of found 32/3306 by comput-= 0.0097 + andfactor⌧ ⌧ bychannels. simulating Nonetheless, large numbers we quantify of ROIs the under trials the (2ing.3 what). The fraction same global of simulated significance Poisson is backgroundfound by comput- ROIs factorDi↵use by 1 simulating model. A largep-value numbers is found of at ROIs each under trial mass the inghave whatp-value fraction less than of simulated 8/3306. Poisson background ROIs Di↵anduse the 1 model. minimum A p of-value these ispm foundis recorded at each for trial each mass sim- haveThep-value application less than of this 8/3306. model-independent procedure andulated the minimum ROI. The of “global” these ppm-valueis recordedpglobal foris the each fraction sim- toThe Ret2 application reveals its of fundamental this model-independent limitation: a strong procedure sig- ulatedof simulated ROI. The ROIs “global” with p-valuem lessp thanglobal thatis the observed fraction in tonal Ret2 necessarily reveals its implies fundamental that very limitation: few background a strong ROIs sig- ofRet2. simulated Simulating ROIs with30 millionpm less background than that ROIs, observed we find in nalhave necessarilyT larger than implies that that of the very object few of background interest. Thus, ROIs ⇠5 5 Ret2.pglobal Simulating=9.8 1030 millionfor b¯b and backgroundpglobal =4 ROIs,.2 10 we findfor havepoorT samplinglarger than of the that large- of theT tail object prevents of interest. a robust Thus, cal- + ⇥ ⇠5 ⇥ 5 pglobal⌧ ⌧ =9. Note.8 that10 thefor trialsb¯b and factorpglobal may=4 have.2 a10 more for sig- poorculation sampling of significance of the large- forT thetail Ret2 prevents signal. a Forrobust exam- cal- + ⇥ ⇥ ⌧ nificant⌧ . Note e↵ect that for the a lighter trials factor final state may (e.g., have electrons). a more sig- culationple, had of we significance used a 5 background for the Ret2 region signal. instead For of exam- 10, nificantFollowing e↵ect for [11, a 38, lighter 41], we final also state consider (e.g., an electrons). entirely dif- ple,zero hadbackground we used ROIs a 5 background would have given region a insteadT value of larger 10, Followingferent procedure [11, 38, for 41], computing we also consider significance. an entirely Under dif- this zerothanbackground Ret2. In any ROIs case, would this have procedure given a clearlyT value identi- larger ferentsecond procedure procedure, for computing we construct significance. the PDF of UnderT due this to thanfies Ret2’s Ret2. as In the any most case, tantalizing this procedure-ray signal clearly from identi- any secondbackground procedure, by making we construct a histogram the of PDFT values of T fordue ROIs to fiesknown Ret2’s dwarf as the galaxy most (left-hand tantalizing panel-ray of Fig. signal 3). from any backgrounddistributed by over making the region a histogram surrounding of T values the dwarf. for ROIs This knownIf the dwarf-ray galaxy signal (left-hand is interpreted panel as of dark Fig. matter,3). we distributedprocedure over is model-independent the region surrounding and automatically the dwarf. This ac- performIf the a-ray simple signal exploration is interpreted of the allowed as dark particle matter, pa- we procedurecounts for is non-Poisson model-independent background and processes automatically (e.g., ac- due performrameter a space. simple As exploration shown in [41], of the for allowedthe two parametersparticle pa- to unresolved sources), an e↵ect examined by several M and v , the likelihood ratio is related to T : counts for non-Poisson background processes (e.g., due rameterh space.i As shown in [41], for the two parameters togroups unresolved [11, 19, sources), 40, 41, 50–52]. an e↵ect examined by several M and v , the likelihood ratio is related to T : Lh(datai (M, v ) + bg) groupsThe [11, left-hand 19, 40, 41, panel 50–52]. of Fig. 3 shows the significance log | h i = T s(E,✓), (2) of Ret2’s signal as calculated following the model- L(data bg) E,✓ The left-hand panel of Fig. 3 shows the significance L(data (M, |v ) + bg) Z independent procedure. Compared with the Poisson- log | h i = T s(E,✓), (2) of Ret2’s signal as calculated following the model- L(data bg) E,✓ process model for background (see above), this proce- where the integral is| the expected numberZ of events in independent procedure. Compared with the Poisson- the ROI due to dark matter annihilation. We denote the dure assigns less significance to Ret2’s -ray signal (in where the integral is the expected number of events in process model for background (see above), this proce- right-hand side as (M, v ). Maximizing (M, v ) accord with [19, 40, 41]). For example, when searching the ROI due to dark matterh annihilation.i We denoteh thei dure assigns less significance to Ret2’s+-ray signal (in yields the maximum likelihood estimate M, v .The for a 25 GeV particle annihilating to ⌧ ⌧ , eight of 3306 right-hand side as (M, v ). Maximizing h(M,i v ) accordbackground with [19, ROIs 40, have 41]).T -values For example, larger than when Ret2’s searching (2.8; di↵erence 2(M, v ) h2(iM, v ) is distributedh as ai + yields the maximumh i likelihood h estimatei M, v .The forother a 25 GeV channels particle show annihilating similar reductions to ⌧ ⌧ in, significance). eight of 3306 2 variable with 2 degrees of freedom [53]c whenhdM,i v h i background ROIs have T -values larger than Ret2’s (2.8; di↵erence 2(M,c dv ) 2(M, v ) is distributed as a other channels show similar reductions in significance). 2 variable with 2h degreesi of freedomh i [53]c whendM, v h i c d Statistical significance of a dark matter interpretation

See Geringer-Sameth, Koushiappas & Walker, PRD 91, 083535 (2015) for details on the methodology

4

3 40 1.6 Ret2 2 Seg1 1.5 45 1.4 1 1.3 0 50 1.2

Significance 1 1.1 Relative intensity 55 Galactic latitude [Deg] 2 1.0 ⌧ +⌧ 60 0.9 3 105 100 95 90 85 80 101 102 103 Galactic longitude [Deg] Mass [GeV]

Local+ p-value = 0.0024 (2.8 sigma) FIG. 3: Left:Significanceof-ray detection for annihilation into ⌧ ⌧ for various masses, calculated using the model- independent procedure of [41]. Solid and dashed blue lines correspond to Ret2 and Seg1 (another attractive nearby target). Gray curves correspond to the collection of dwarfs used in [41] as wellGlobal as the 8p-value other newly = discovered0.0097 ( DES2.3 dwarfs. sigmaRight:) The FermiBased isotropic+di on PRL 115,↵ 081101use model & ApJL intensity 808 L36 near (2015) Ret2. The color corresponds to intensity normalized to the value in the direction of Ret2 (at an energy of 8 GeV — other energies are similar). A 0.5 ROI is shown at the center and the small dots show the centers of the ROIs used for the empirical background estimation. White ’s mark the locations of known -ray sources. Green circles are the ROIs which have a test statistic larger than that in the central⇥ ROI (when searching for a 25 GeV particle + annihilating to ⌧ ⌧ ). are the true values of the mass and cross section. There- fore, regions of (M, v ) space where this di↵erence is less than 2.3, 6.2, andh 11.8i constitute 68.2%, 95.4%, and 99.7% confidence regions. The 2 behavior holds only for large sample sizes and it is not clear if that assumption is valid here. In particular, for annihilation into electrons or muons, where low masses are preferred, there are very few events above 1 GeV but below the dark matter mass. Figure 4 shows the derived constraints on the prod- uct J v for a number of representative channels. Al- thoughh wei cannot make a direct measurement of the cross section, the constraints on J v , combined with exist- ing upper limits on v , allowh usi to make a prediction for the dark matterh contenti of Ret2 which must hold if the -ray emission is due to annihilating dark matter. + In the ⌧ ⌧ channel, for example, the limits from [41] yield log J 19.6 0.3 (compare with Seg1, which has 10 & ± FIG. 4: An exploration of a dark matter interpretation log10 J = 19.3 0.3 [47]). of the observed -ray excess for four representative anni- ± 19 2 5 hilation channels. J = J19 10 GeV cm and v = While Ret2’s -ray signal is tantalizing, it would 26 3 1 h i v 26 10 cm sec . Currently the data constrain only be premature to conclude it has a dark matter ori- theh i product of J v since the dark matter content of Retic- gin. Among alternative explanations, perhaps the most ulum 2 is currentlyh unknown.i Contours represent 68%, 95%, mundane is the possibility that an extragalactic source and 99.7% confidence regions. Note that this figure does not lies in the same direction. Searching the BZCAT [54] quantify which annihilation channel is preferred by the data, and CRATES [55] catalogs reveals a CRATES quasar i.e. which channel provides the best fit to the -ray spectrum. (J033553-543026) that is 0.46 from Ret2. Further work must be done to determine whether this particular source contributes to the emission, though we note that flat millisecond pulsars [24, 26, 57–61]. In the case of Ret2, spectrum radio quasars rarely have a spectral index less it is the high-energy behavior which disfavors a pulsar than 2 [56]. One of the much-discussed astrophysical model, as millisecond pulsars exhibit an exponential cut- explanations for the apparent Galactic Center excess is o↵ at around 2.5 to 4 GeV [26, 30, 61–64]. Alternatively, Statistical significance of a dark matter interpretation

Pass7 Pass8

Empirical background

Local p-value = 0.0024 (2.8 sigma) p-values comparable with what found in Drlica-Wagner et al. ApJL 809 L4. Global p-value = 0.0097 (2.3 sigma)

Poisson background

Local p-value = 0.0000068 (4.4 sigma)

Global p-value = 0.000042 (3.7 sigma)

Based on PRL 115, 081101 & ApJL 808 L36 (2015) 4

+ FIG. 3: Left:Significanceof-ray detection for annihilation into ⌧ ⌧ for various masses, calculated using the model- independent procedure of [41]. Solid and dashed blue lines correspond to Ret2 and Seg1 (another attractive nearby target). Gray curves correspond to the collection of dwarfs used in [41] as well as the 8 other newly discovered DES dwarfs. Right: The Fermi isotropic+di↵use model intensity near Ret2. The color corresponds to intensity normalized to the value in the direction of Ret2 (at an energy of 8 GeV — other energies are similar). A 0.5 ROI is shown at the center and the small dots show the centers of the ROIs used for the empirical background estimation.Statistical White ’s marksignificance the locations of of known a dark-ray sources. matter interpretation Green circles are the ROIs which have a test statistic larger than that in the central⇥ ROI (when searching for a 25 GeV particle + annihilating to ⌧ ⌧ ).

Based on PRL 115, 081101 & ApJL 808 L36 (2015) are the true values of the mass and cross section. There- 103 fore, regions of (M, v ) space where this di↵erence is less than 2.3, 6.2, andh 11.8i constitute 68.2%, 95.4%, and 99.7% confidence regions. The 2 behavior holds only for 102 large sample sizes and it is not clear if that assumption is 26 valid here. In particular, for annihilation into electrons i or muons, where low masses are preferred, there are very v Assuming a J value as in 1 h 10 few events above 1 GeV but below the dark matter mass. Bonnivard et al. ApJL 808 L36 (2015) 19 Figure 4 shows the derived constraints on the prod- J uct J v for a number of representative channels. Al- 0 h i 10 though we cannot make a direct measurement of the cross b¯b hh section, the constraints on J v , combined with exist- + + h i ⌧ ⌧ µ µ ing upper limits on v , allow us to make a prediction 1 h i 10 for the dark matter content of Ret2 which must hold if 101 102 103 the -ray emission is due to annihilating dark matter. Mass [GeV] + In the ⌧ ⌧ channel, for example, the limits from [41] What about consistency checks with the Galactic center and other dwarfs? yield log10 J & 19.6 0.3 (compare with Seg1, which has FIG. 4: An exploration of a dark matter interpretation ± (see e.g., Abazajian & Keeley 1510.06424) log10 J = 19.3 0.3 [47]). of the observed -ray excess for four representative anni- ± 19 2 5 hilation channels. J = J19 10 GeV cm and v = While Ret2’s -ray signal is tantalizing, it wouldDoes the data prefer26 one3 explanation1 (channel) over somethingh i else? What can the LHC tell us? v 26 10 cm sec . Currently the data constrain only be premature to conclude it has a dark matter ori-(see e.g.,theh i productFan, Koushiappas of J v since & Landsberg, the dark matter1507.06993) content of Retic- gin. Among alternative explanations, perhaps the most ulum 2 is currentlyh unknown.i Contours represent 68%, 95%, mundane is the possibility that an extragalactic source and 99.7% confidence regions. Note that this figure does not lies in the same direction. Searching the BZCAT [54] quantify which annihilation channel is preferred by the data, and CRATES [55] catalogs reveals a CRATES quasar i.e. which channel provides the best fit to the -ray spectrum. (J033553-543026) that is 0.46 from Ret2. Further work must be done to determine whether this particular source contributes to the emission, though we note that flat millisecond pulsars [24, 26, 57–61]. In the case of Ret2, spectrum radio quasars rarely have a spectral index less it is the high-energy behavior which disfavors a pulsar than 2 [56]. One of the much-discussed astrophysical model, as millisecond pulsars exhibit an exponential cut- explanations for the apparent Galactic Center excess is o↵ at around 2.5 to 4 GeV [26, 30, 61–64]. Alternatively, Questions

1. Is it consistent with background?

2. Is it consistent with dark matter annihilation?

3. Is it consistent with any other source?

4. Is it something else? (e.g., instrumental/data set systematics?) (P7R vs P8)

Questions

1. Is it consistent with background?

2. Is it consistent with dark matter annihilation?

3. Is it consistent with any other source?

4. Is it something else? (e.g., instrumental/data set systematics?) (P7R vs P8)

Realities:

- At this point any dark matter discovery will come at the detection threshold.

- We only have one dataset to work with (no possibility of independent cross-check).

- Each and every photon counts (this is important for any source at the threshold). From Pass7R to Pass 8 From Pass7R to Pass 8

Positional differences between Pass7R and Pass8 in the common dataset

26% > 0.5 degrees 30 degrees from RetII 10% > 1 degree 0.1% > 10 degrees From Pass7R to Pass 8

Photon energy differences between Pass7R and Pass8 in the common dataset From Pass7R to Pass 8 MeV

MeV In conclusion

Given that this is the very first time we have a [fill in your favorite word] of gamma-rays along the line of sight to a dwarf galaxy it is important we understand Reticulum II as much as the data allows as it is a massive nearby dwarf galaxy ̶ a prime target in the search for a non-gravitational signature of dark matter.

More importantly: We need more data!