Enhancing Dark Matter searches at the LHC with Graph Neural Networks

Dark Matter remains one of the most profound open questions in contemporary physics. A promising strategy to search for Dark Matter particles is through their production at the Large Hadron Collider (LHC). In particular, the monojet channel, characterised by a small number of high-energy jets recoiling against significant missing transverse momentum and the absence of isolated leptons, can be used for Dark Matter discovery. However, the analysis of this channel is challenging for classical methods due to the requirement for precise analysis of jet substructures. In this work, we present a novel approach based on Graph Neural Networks to enhance the sensitivity of monojet searches. We demonstrate the effectiveness of our method using a simplified supersymmetric model and provide projections for the discovery potential at the High-Luminosity LHC.