Speaker
Description
This talk presents a novel approach to detect a light fermiophobic Higgs boson (hf) with mass in the range of 1-10 GeV at the HL-LHC, focusing on the golden channel p p → hf H → γγγγℓν. Due to the highly collimated nature of photon pairs from hf decay, traditional detection methods face significant challenges as these photons manifest as single jets rather than isolated photons. We introduce an innovative strategy utilizing Convolutional Neural Networks (CNN) to identify these distinctive diphoton jets and distinguish them from QCD backgrounds. Our method employs the Delphes framework's EFlow objects, combined with a hybrid pileup subtraction technique that merges charged hadron subtraction with SoftKiller. The analysis demonstrates remarkable results across 18 benchmark points, achieving signal significance above 5σ at an integrated luminosity of 3 ab⁻¹. For challenging scenarios involving heavy charged Higgs bosons, our CNN-based approach significantly enhances detection capability, nearly doubling the significance compared to traditional cut-based methods. This study opens new possibilities for exploring previously challenging parameter spaces in BSM physics through machine learning techniques.
