A model-independent analysis of the Fermi Large Area Telescope gamma-ray data from the Milky Way dwarf galaxies and halo to constrain dark matter scenarios

We implemented a novel technique to perform the collective spectral analysis of sets of multiple gamma-ray point sources using the data collected by the Large Area Telescope onboard the Fermi satellite. The energy spectra of the sources are reconstructed starting from the photon counts and without assuming any spectral model for both the sources and the background. In case of faint sources, upper limits on their fluxes are evaluated with a Bayesian approach. This analysis technique is very useful when several sources with similar spectral features are studied, such as sources of gamma rays from annihilation of dark matter particles. We present the results obtained by applying this analysis to a sample of dwarf spheroidal galaxies and to the Milky Way dark matter halo. The analysis of dwarf spheroidal galaxies yields upper limits on the product of the dark matter pair annihilation cross section and the relative velocity of annihilating particles that are well below those predicted by the canonical thermal relic scenario in a mass range from a few GeV to a few tens of GeV for some annihilation channels.

💡 Research Summary

This paper presents a model‑independent analysis of three years of Fermi‑LAT gamma‑ray data (562 MeV–562 GeV) aimed at constraining dark‑matter (DM) annihilation in Milky Way dwarf spheroidal galaxies (dSphs) and the Galactic halo. Instead of fitting pre‑defined spectral templates, the authors employ the FermiUnfolding package to reconstruct the energy spectra directly from photon counts, using a smearing matrix derived from full Monte‑Carlo simulations of the LAT instrument response. The analysis proceeds in three stages: (1) individual source analysis, (2) stacking of multiple sources, and (3) a composite treatment that treats the whole ensemble as a single effective source.

For each source, a signal region (circular cone) and a background region (annulus) are defined. Photon counts n (signal) and m (background) are obtained in twelve logarithmically spaced energy bins. A Bayesian approach with uniform priors on the unknown signal s and background b yields the posterior probability density p(s|n,m) (Eqs. 6‑7). The 95 % credibility upper limit sᵤ is found by numerically integrating the posterior. These count limits are converted to flux limits via the unfolding procedure, which accounts for energy dispersion and exposure variations.

The particle‑physics factor Φ_PP(E) is related to the observable flux through the astrophysical J‑factor (line‑of‑sight integral of ρ²). Using published J‑factors for each dSph (including their uncertainties), the flux limits are translated into limits on ⟨σv⟩·⟨v⟩ for specific annihilation channels (b b̄, τ⁺τ⁻, etc.) using the DMFIT package (based on DarkSUSY). In the stacking analysis, signal and background counts from all dSphs are summed, and a combined c‑factor is defined either as a simple solid‑angle ratio (Eq. 11) or, more rigorously, as a weighted average that incorporates individual background counts (Eq. 12). The authors find negligible differences between the two definitions. The effective J‑factor for the stack is computed as an exposure‑weighted average of the individual J‑factors; variations across energy bins are <1 %, justifying the use of a single J‑factor for the whole energy range.

Results: For individual dSphs, the 95 % upper limits on ⟨σv⟩·⟨v⟩ fall below the canonical thermal relic cross‑section (≈3 × 10⁻²⁶ cm³ s⁻¹) for WIMP masses between ~5 GeV and ~30 GeV, depending on the channel. Stacking improves the limits by roughly a factor of two, reaching ⟨σv⟩·⟨v⟩ ≲ 1–2 × 10⁻²⁶ cm³ s⁻¹ in the same mass range. The Galactic halo analysis yields weaker constraints due to larger background uncertainties and less well‑determined J‑factors, but still provides limits competitive with previous studies.

The paper highlights several methodological advantages: (i) no reliance on assumed spectral shapes eliminates model bias; (ii) the Bayesian framework naturally incorporates systematic uncertainties such as J‑factor errors and effective area variations; (iii) the stacking procedure treats each source uniformly while preserving exposure differences. Potential drawbacks include sensitivity of the Bayesian upper limits to the choice of prior when photon counts are extremely low, and the possibility that averaging J‑factors may dilute source‑specific information.

In conclusion, the authors demonstrate that a model‑independent unfolding combined with Bayesian upper‑limit estimation offers a robust and flexible tool for indirect DM searches with Fermi‑LAT. The technique yields limits that are comparable to or stronger than those obtained with traditional template‑fitting methods, and it can be readily extended to larger data sets, improved instrument response functions, and newly discovered dwarf galaxies, thereby sharpening constraints on the WIMP parameter space.