Impact of neutral fluxes and signal significance optimization on semi-exclusive pp → tt̄ production via deep learning training

Photon flux benchmark models together with recent experimental estimations for pomeron energy fluxes and structure functions are implemented within Monte Carlo simulation, to determine their impact on physical observables and on the signal strength uncertainty of the yet-to-be observed semi-exclusive tt̄ t t ¯ production in proton–proton ( pp ) collisions at Large Hadron Collider (LHC) and Future Circular Collider (FCC) energies. Expected cross-section rises by a factor of ~ 50 ∼ 50 and ~ 22 ∼ 22 for pomeron-induced and photon-induced processes respectively from LHC to FCC energy regime. TensorFlow deep neural networks were implemented to discriminate semi-exclusive processes against non-peripheral tt̄X t t ¯ X background, achieving an Area Under the Curve (AUC) test performance of 0.9917 ± 0.0002 0.9917 ± 0.0002 for the photon-induced signal. At detector level, following the geometry of the CMS detector at CERN, the minimum Hadronic Forward energy observable was identified as the most effective discriminator against non-peripheral tt̄X t t ¯ X background. In the context of low pileup pp data and based on the Asimov dataset, by considering the statistical effects and the systematic contribution from pomeron/photon schemes to the total uncertainty a 5- σ σ significance is expected for pomeron-induced and photon-induced tt̄ t t ¯ production modes with 1 fb ⁻¹ - 1 and 4.7 fb ⁻¹ - 1 integrated luminosity datasets respectively. The systematic contribution to the total uncertainty is 25.1% and 7.6% respectively, highlighting the potential for experimental observation using Run 2 and Run 3 LHC data, allowing further studies within the LHC forward physics program.