Research on Pulse Discrimination for Alpha and Beta Measurement Based on Auto-Encoder

Abstract Pulse shape discrimination is an important method for measuring both gross alpha and gross beta in a single measurement. When the quenching level of the sample changes significantly, the percent spillover of the alpha events into the beta count and the beta events into the alpha count will change greatly. In order to ensure the accuracy of the measurement results, it is necessary to re-optimize the pulse shape discrimination parameter. For each sample measurement, a spiked sample with the same quenching level needs to be established for pulse shape discrimination parameter optimization. In this paper, we proposed a pulse discrimination method for alpha and beta decay based on auto-encoder, which only needs to be optimized once, and can adapt to the significant changes in the quenching level of the sample. The pulse shape discrimination model was trained on the measurement data of Am241 spiked samples with different quenching levels, and verified by the measurement data of Sr90 spiked samples with different quenching levels. Finally, the optimal parameter for pulse shape discrimination was determined. The model was tested using the measurement data of Am241 and Sr90 spiked samples with different quenching levels. These samples quenching level range from 306 to 433 which represented by transformed spectral index of the external standard. The percentage of alpha pulse misclassification is 2.21%–3.35%, and the percentage of beta pulse misclassification is 2.37%–2.82%.