sPHENIX Electron Identification for Prompt J/ψ Decays using DNN and CNN Algorithms

Accurate measurements of J/ψ production are crucial for advancing the heavy flavor program in sPHENIX. During Run 2024, sPHENIX collected 13 pb⁻¹ of data, using both tracking and calorimetry detectors, with a 1.5 mrad crossing angle and |z_vertex| < 10 cm in 200 GeV p + p collisions. Experimentally, measuring J/ψ particles remains challenging due to their low production rate and the high single-particle background at low pT, which limits electron identification capabilities. To address this, we propose training a Convolutional Neural Network (CNN), specialized in image recognition, on sPHENIX 200 GeV p + p simulation data to improve electron identification capabilities and extend electron identification from prompt J/ψ decays down to 1 GeV. This approach has the potential to increase the measurable J/ψ sample by up to a factor of five in simulations compared to traditional techniques effective primarily above 2 GeV. We will present the implementation and training of both Deep Neural Network (DNN) and CNN models, along with their performance evaluated using confusion matrices on both simulation and real sPHENIX data.