Gray Selby
Faculty Sponsor: Michelle Kuchera
Nuclear physicists strive to describe the structure of the atomic nucleus. This nuclear scattering experiment to study 22Mg alpha-proton events was conducted to learn about the shell structure of exotic nuclei to expand our understanding of nuclear interactions and test our theoretical models. This project focused on identifying differences in this data set and adapting our method using convolutional neutral networks to separate alpha-proton events from multitude of nuclear interactions. This experiment was conducted at the National Superconducting Cyclotron Laboratory using the AT-TPC, a gas filled detector that acts as both the detector and target for high-efficiency detection of low-intensity, exotic nuclear reactions. Prior to this 22Mg alpha-proton experiment, the AT-TPC was modified so that charge was normalized across pad plane sections. This prevented the accurate use of the pytpc analysis framework developed for these experiments. With the data adjusted for the changes to the data acquisition, our convolutional neural network developed using Google’s automatic differentiation package TensorFlow, was trained using two-dimensional plots form this data set. This convolutional neutral network was used to classify alpha-proton events from the multitude of data.