Background subtraction in inelastic scattering measurements using machine learning

Abstract

Identifying, isolating, and subtracting background from the signal of interest is vital for nuclear physics experiments. These backgrounds introduce unwanted uncertainties that must be accounted for properly to extract accurate results from the signals. In nuclear reaction measurements, the typical contaminants are carbon and oxygen, contributing to background signals, and complicating the measurement of the light ejectiles. For instance, in the inelastic scattering measurement of a 20.9-MeV proton beam on ⁹⁶Mo, the ⁹⁶Mo target was contaminated with carbon and oxygen. We used random forest, a machine learning algorithm commonly used for classification and regression tasks, to separate the inelastic scattering on the carbon and oxygen contaminants from the data of interest resulting from ⁹⁶Mo($p,p^{\prime }$).