Predicting Reduction of Plasma Instabilities Using a Neural Network

The KATRIN experiment is designed to find the neutrino mass from analyzing the kinematics of tritium beta decay. Tritium beta decay produces positive ions and low energy electrons, which create a plasma. This plasma is found at the source section of the KATRIN experimental setup and it determines the starting potential of higher energy beta electrons. Variations in this plasma affect the tritium beta spectrum that is analyzed to determine the mass of the neutrino. Therefore, an optimal voltage must be applied at the rear wall in the source section to minimize instabilities in plasma potential. For this purpose, rear-wall current scans are carried out. These scans observe the current at the rear wall when the rear wall voltage changes and determine an optimal rear wall voltage at the time of the scan. I will report progress on a neural network that will utilize data from these rear-wall current scans to obtain the optimal rear wall voltage at times between scans. This will lower systematic uncertainties that arise from plasma instabilities in the KATRIN experiment.

This work was partially supported by the DOE Nuclear Physics Award No. DE-SC00193204.