Black‑hole images made by the Event Horizon Telescope can be altered either by matter sitting near the hole or by changes to gravity coming from extra dimensions. The new study called DARK‑HIDE asks a concrete question: if an observed image differs from the standard Kerr black hole, can we tell whether the change comes from a dark‑matter cloud or from an effective extra‑dimensional correction? The authors compare the two possibilities on the same footing and test how well current image measurements can tell them apart.

The team builds two families of rotating black‑hole models. The “dark‑matter” branch changes the hole’s mass as a function of radius using two standard density shapes: an Einasto profile and a cored Navarro–Frenk–White (cNFW) profile. The strength of that extra mass is set by a parameter labeled ε (examples use ε/M = 0.025, where M is the black‑hole mass). The “hidden‑dimension” branch uses a rotating braneworld metric with a tidal charge q (measured as q/M^2). This q is not an electric charge: it is an effective parameter that can be positive or negative and encodes how a higher‑dimensional bulk would project onto four‑dimensional gravity. For each model the authors ray‑trace light rays and compare key optical features: the photon region (where light can orbit the hole), the critical curve (the outline of the shadow seen by a distant observer), a controlled image morphology proxy, and local escape cones for a nearby observer with zero angular momentum (ZAMO).

Their main technical result is that a strong negative tidal charge is relatively easy to distinguish from the Kerr case and from the two dark‑matter benchmarks. But the more challenging situation arises when the tidal charge is continuously tuned to mimic the dark‑matter critical curve and image proxy. For the chosen benchmark amplitude ε/M = 0.025, the best matches occur at q/M^2 = −0.01917 for the Einasto model and q/M^2 = −0.01117 for the cNFW model. The standardized separations of these best matches are small (0.084 and 0.051 in the authors’ metric), meaning the two branches can be nearly degenerate in the observables they used. A separate ray‑bundle caustic test, intended to check for additional structure, failed the required convergence and topology checks and was therefore excluded from the inference.