Collider, direct and indirect searches for dark matter have typically little or no sensitivity to weakly interacting massive particles (WIMPs) with masses above a few TeV. This rather unexplored regime can however be probed through the search for distinctive gamma-ray spectral features produced by the annihilation of WIMPs at very high energies. Here we present a dedicated search for gamma-ray boxes -- sharp spectral features that cannot be mimicked by astrophysical sources -- with the upcoming Cherenkov Telescope Array (CTA). Using realistic projections for the instrument performance and detailed background modelling, a profile likelihood analysis is implemented to derive the expected upper limits and sensitivity reach after 100 h of observations towards a $2^\circ\times2^\circ$ region around the Galactic centre. Our results show that CTA will be able to probe gamma-ray boxes down to annihilation cross sections of $10^{-27}-10^{-26}\,\text{cm}^3\text{/s}$ up to tens of TeV. We also identify a number of concrete particle physics models providing thermal dark matter candidates that can be used as target benchmarks in future search campaigns. This constitutes a golden opportunity for CTA to either discover or rule out multi-TeV thermal dark matter in a corner of parameter space where all other experimental efforts are basically insensitive.

We compute the gamma-ray output of axion-mediated dark matter and derive the corresponding constraints set by recent data. In such scenarios the dark matter candidate is a Dirac fermion that pair-annihilates into axions and/or scalars. Provided that the axion decays (at least partly) into photons, these models naturally give rise to a box-shaped gamma-ray spectrum that may present two distinct phenomenological behaviours: a narrow box, resembling a line at half the dark matter mass, or a wide box, spanning an extensive energy range up to the dark matter mass. Remarkably, we find that in both cases a sizable gamma-ray flux is predicted for a thermal relic without fine-tuning the model parameters nor invoking boost factors. This large output is in line with recent Fermi-LAT observations towards the Galactic centre region and is on the verge of being excluded. We then make use of the Fermi-LAT and H.E.S.S. data to derive robust, model-independent upper limits on the dark matter annihilation cross section for the narrow and wide box scenarios. H.E.S.S. constraints, in particular, turn out to match the ones from Fermi-LAT at hundreds of GeV and extend to multi-TeV masses. Future Cherenkov telescopes will likely probe gamma-ray boxes from thermal dark matter relics in the whole multi-TeV range, a region hardly accessible to direct detection, collider searches and other indirect detection strategies.

The observation of a sharp spectral feature in the gamma-ray sky would be one of the cleanest ways to identify dark matter and pinpoint its properties. Over the years a lot of attention has been paid to two specific features, namely gamma-ray lines and internal bremsstrahlung. Here, we explore a third class of spectral signatures, box-shaped gamma-ray spectra, that naturally arise in dark matter cascade annihilations or decays into intermediate particles that in turn decay into photons. Using Fermi-LAT data, we derive constraints on the dark matter parameter space for both annihilating and decaying dark matter, and show explicitly that our limits are competitive to strategies employing standard spectral features. More importantly, we find robust limits even in the case of non-degenerate dark matter and intermediate particle masses. This result is particularly relevant in constraining dark matter frameworks with gamma-ray data. We conclude by illustrating the power of box-shaped gamma-ray constraints on concrete particle physics scenarios.