Legacy analysis of dark matter annihilation from the Milky Way dwarf spheroidal galaxies with 14 years of $\mathit{Fermi}$-LAT data

The Milky Way dwarf spheroidal satellite galaxies (dSphs) are particularly intriguing targets to search for gamma rays from weakly interacting massive particle dark matter (DM) annihilation or decay. They are nearby, DM-dominated, and lack significant emission from standard astrophysical processes. Previous studies using the Fermi-Large Area Telescope (LAT) of DM-induced emission from dSphs have provided the most robust and stringent constraints on the DM annihilation cross section and mass. We report here an analysis of the Milky Way dSphs using over 14 years of LAT data along with an updated census of dSphs and $J$-factor estimates. While no individual dSphs are significantly detected, we do find slight excesses with respect to background at the $\gtrsim 2\sigma$ local significance level in both tested annihilation channels ($b\overline{b}$, ${\tau }^{+}{\tau }^{-}$) for seven of the dSphs. We do not find a significant DM signal from the combined likelihood analysis of the dSphs ($s_{\text{global}}\sim 0.5\sigma$), yet a marginal local excess relative to background at a $2–3\sigma$ level is observed at a DM mass of $M_{\chi }=150–230\mathrm{GeV}$ ($M_{\chi }=30–50\mathrm{GeV}$) for DM annihilation into $b\overline{b}$ (${\tau }^{+}{\tau }^{-}$). Given the lack of a significant detection, we place updated constraints on the $b\overline{b}$ and ${\tau }^{+}{\tau }^{-}$ annihilation channels that are generally consistent with previous recent results. As in past studies, tension is found with some weakly interacting massive particle DM interpretations of the Galactic Center Excess, though the limits are consistent with other interpretations given the uncertainties of the Galactic DM density profile and Galactic Center Excess systematics. Based on conservative assumptions of improved sensitivity with increased Fermi-LAT exposure time and moderate increases in the sample of Milky Way dSphs, we project that the local $\sim 2\sigma$ signal, if real, could approach the $\sim 4\sigma$ local confidence level with additional $\sim 10\mathrm{years}$ of observation.

Figure 1

$J$ factors and uncertainties for all 50 dSphs considered in this work. Marker shapes and colors indicate whether the $J$ factor is determined by measurement (blue circles), the kinematic scaling relation of Eq. (3) (green squares), or the photometric scaling relation of Eq. (4) (orange triangles). The special cases are outlined in red. Also indicated in this figure are the different subsets into which each dSph falls.

Figure 2

TS distribution for combined blank fields corresponding to each of our subsamples. For reference, we show the distribution of ${\chi }^2/2$ with two degrees of freedom, as well as vertical lines comparing the TS value corresponding to $3\sigma$ for the Chernoff’s theorem approach (red dashed) vs the empirical method used in this analysis (blue dashed) for a given subsample.

Figure 3

Maximum TS as a function of $M_{\chi }$ over all cross-section values for the individual dSphs in comparison with the individual blank fields. The dSphs with a local significance $\ge 2\sigma$ are shown as colored lines. The yellow and green bands show the 97.5% and 84% containment regions for the individual blank fields for the $b\overline{b}$ (left) and ${\tau }^{+}{\tau }^{-}$ (right) annihilation channels.

Figure 4

Flux (left) and cross section (right) upper limits vs $J$ factor for the individual dSphs. The upper limits on the cross section are calculated assuming $M_{\chi }=100\mathrm{GeV}$ and $b\overline{b}$ annihilation channel. The dSphs with local significances $\ge 2\sigma$ for both annihilation channels are outlined in black and labeled. Colors and shapes are defined the same as Fig. 1, except that special cases are not highlighted here for visual clarity. Error bars are not shown for the dSphs without measured $J$ factors where an uncertainty of 0.6 dex is assumed.

Figure 5

Top: maximum TS as a function of $M_{\chi }$ over all cross section values for each dSph sample; shaded regions are the 97.5% containment region for the combined blank fields. Bottom: $1-p$ for each sample with respect to the combined blank fields. Solid lines show the local values ($1-p_{\text{local}}$) while dashed lines show the global values ($1-p_{\text{global}}$). Horizontal dotted lines indicate the 1, 2, and $3\sigma$ levels.

Figure 6

Constraints on the $b\overline{b}$ (left) and ${\tau }^{+}{\tau }^{-}$ (right) annihilation channels derived in this work compared to previous dSph constraints as well as GCE models from [17, 46, 47, 48, 49, 51]. The contour definitions in the top left panel legend apply to each figure. The yellow and green bands show the 95% and 68% containment regions, respectively, of the upper limits derived from the combined blank fields and the gray line is the thermal relic cross section from [54]. Top: comparison of the upper limits from the Measured sample with previous results from the dSphs analysis of [11]. Bottom: comparison of the upper limits from the Benchmark sample with previous results from [14, 17].

Figure 7

Left: comparison of sensitivity between the measured (30 dSphs) and benchmark (42 dSphs) samples of this analysis with the sensitivity limits from [11, 14] as determined by the median of the blank-fields limits. Light and dark green dashed curves are the predicted sensitivity limits from [131] for 15 total years of Fermi-LAT exposure and different dSph sample sizes. The contours and data points are the GCE models as defined in Fig. 6. Right: projected local signal significance relative to background for the benchmark sample given further Fermi-LAT exposure and with additional dSphs in the sample.

Figure 8

Constraints on the $⟨\sigma v⟩-M_{\chi }$ parameter space for an expanded selection of DM annihilation channels for each of the samples considered in this work. The gray horizontal line is the thermal relic cross section from [54].

Figure 9

TS vs $M_{\chi }$ for the standard likelihood analysis (solid lines) and the weighted likelihood analysis (dashed lines) for the individual high TS dSphs (left), and the combined dSph samples (right).

Figure 10

Maximum TS as a function of $M_{\chi }$ over all cross section values for each dSph sample using the weighted likelihood method; shaded regions are the 97.5% containment region for the combined blank fields.