A Tentative Gamma-Ray Line from Dark Matter Annihilation at the Fermi Large Area Telescope

TL;DR

This paper reports a refined search for monochromatic gamma-ray lines from DM annihilation in Fermi-LAT data (20–300 GeV), leveraging 43 months and a data-driven target-region optimization tied to DM halo profiles. Using a sliding energy window and a robust likelihood framework, the authors identify a line-like feature near 130 GeV with a pre-trial significance of 4.6σ and a post-trial significance of 3.2σ, corresponding to a DM mass around 130 GeV and a non-negligible branching ratio to γγ for certain DM profiles. The best-fit annihilation cross-sections vary with profile (e.g., Einasto vs. NFW), and the signal comprises roughly 46–88 photons; however, the authors stress the need for more data and caution about potential systematics. Methodological advances include region optimization to maximize SNR, thorough MC/subsampling validation, and careful treatment of instrument responses, which collectively strengthen the search for a potential DM gamma-ray line, with important implications for DM models and gamma-ray astronomy if confirmed.

Abstract

The observation of a gamma-ray line in the cosmic-ray fluxes would be a smoking-gun signature for dark matter annihilation or decay in the Universe. We present an improved search for such signatures in the data of the Fermi Large Area Telescope (LAT), concentrating on energies between 20 and 300 GeV. Besides updating to 43 months of data, we use a new data-driven technique to select optimized target regions depending on the profile of the Galactic dark matter halo. In regions close to the Galactic center, we find a 4.6 sigma indication for a gamma-ray line at 130 GeV. When taking into account the look-elsewhere effect the significance of the observed excess is 3.2 sigma. If interpreted in terms of dark matter particles annihilating into a photon pair, the observations imply a dark matter mass of 129.8\pm2.4^{+7}_{-13} GeV and a partial annihilation cross-section of <σv> = 1.27\pm0.32^{+0.18}_{-0.28} x 10^-27 cm^3 s^-1 when using the Einasto dark matter profile. The evidence for the signal is based on about 50 photons; it will take a few years of additional data to clarify its existence.

Figures (9)

• Figure 1: Left panel: The black lines show the target regions that are used in the present analysis in case of the SOURCE event class (the ULTRACLEAN regions are very similar). From top to bottom, they are respectively optimized for the cored isothermal, the NFW (with $\alpha=1$), the Einasto and the contracted (with $\alpha=1.15$, $1.3$) DM profiles. The colors indicate the signal-to-background ratio with arbitrary but common normalization; in Reg2 to Reg5 they are respectively downscaled by factors (1.6, 3.0, 4.3, 18.8) for better visibility. Right panel: From top to bottom, the panels show the $20$--$300\ \text{GeV}$ gamma-ray (+ residual CR) spectra as observed in Reg1 to Reg5 with statistical error bars. The SOURCE and ULTRACLEAN events are shown in black and magenta, respectively. Dotted lines show power-laws with the indicated slopes; dashed lines show the EGBG + residual CRs. The vertical gray line indicates $E=129.0\ \text{GeV}$.
• Figure 2: Energy windows that we use for our spectral line search. In red we indicate the window that enters the fit at $E_\gamma=129.0\ \text{GeV}$.
• Figure 3: TS value as function of the line energy $E_\gamma$, obtained by analysing the energy spectra from the different target regions in Fig. \ref{['fig:reg']}. Left and right panels show the results for the SOURCE and ULTRACLEAN event classes, respectively. The inset shows a zoom into the relevant region. The horizontal gray dotted lines show respectively from bottom to top the $1\sigma$ to 3$\sigma$ levels after correcting for trials (without trial correction the significance is given by $\sqrt{TS}\sigma$). In the right panel, the gray crosses show the TS values that we obtain when instead adopting the target region and energy windows from Refs. Fermi:Symp2011TalkEdmonds:2011PhD with 43 months of data.
• Figure 4: Upper sub-panels: the measured events with statistical errors are plotted in black. The horizontal bars show the best-fit models with (red) and without DM (green), the blue dotted line indicates the corresponding line flux component alone. In the lower sub-panel we show residuals after subtracting the model with line contribution. Note that we rebinned the data to fewer bins after performing the fits in order to produce the plots and calculate the $p$-value and the reduced $\chi_r^2\equiv\chi^2/\text{dof}$. The counts are listed in Tabs. \ref{['tab:tablesReg2']}, \ref{['tab:tablesReg3']} and \ref{['tab:tablesReg4']}.
• Figure 5: Best-fit values for the annihilation cross-section into a photon pair, as obtained for different DM halo profiles in the different target regions of Fig. \ref{['fig:reg']}, together with their 68.2% CL and 95.5% CL errors. Previous upper 95% CL limits are shown for comparison Fermi:Symp2011TalkEdmonds:2011PhD. All values are derived assuming that $m_\chi=129.0\ \text{GeV}$.