Precision Hadron Production Measurements from EMPHATIC

Precision neutrino oscillation experiments such as DUNE and SBN rely on accurate neutrino flux predictions. A leading source of uncertainty in these predictions arises from hadron production modeling, especially for sub-10 GeV protons on nuclear targets, where uncertainties can reach 5–15 %. The EMPHATIC experiment (Experiment to Measure the Production of Hadrons At Testbeam in Chicagoland) addresses this by collecting high-precision hadron production data at the Fermilab Test Beam Facility.

This work focuses on the Aerogel Ring Imaging Cherenkov detector (ARICH), a key component of EMPHATIC used to identify hadronic final states. We develop and compare particle identification techniques, including likelihood-based methods using expected Cherenkov angle distributions, machine learning models such as convolutional neural networks and U-Net trained on ARICH hit patterns and boosted decision trees trained using likelihood values. These approaches are evaluated in terms of identification efficiency, purity, and robustness to noise, with the goal of enabling precise cross-section measurements and reducing flux uncertainties in current and future neutrino experiments.