Energy Measurement in EXO-200 using Boosted Regression Trees

The EXO-200 experiment uses a Liquid Xenon (LXe) time projection chamber (TPC) to search for neutrinoless-double beta decay(0νββ), an extremely rare hypothetical decay that would indicate the Majorana nature of neutrinos.[1, 2, 3] The EXO-200 experiment has been taking data for over 2 years and has published one of the most sensitive limits on 0νββ half-life. Events deposit energy in the LXe through both scintillation light (175nm) and free ionization charge. The scintillation light is detected at either end of the EXO-200 detector by large area avalanche photodiodes (APDs). The ionized charge is drifted along the z-axis of the detector where it first passes a shielding/induction wire grid (V-wires) and is than collected by a second wire grid of collection wires (U-wires). Each wire is 3mm in pitch and wires are ganged into groups of three before being readout and saved. The total charge energy of an event is then calculated by determining the sum amplitude of all channels which collected charge. In order to accurately reconstruct the energy of the waveform signals have to be accurately identified as either collection or induction signals. This is currently done by performing a χ2 fit to both a collection and induction signal and than classifying the waveform based on the ratio of these scores. Waveforms classified as induction are flagged and then not included into the sum when determining event energy. This current technique achieves reasonable efficiency at identifying collection and induction signals but energy deposits spanning multiple channels present a slight challenge because waveforms will contain both collection and induction signals. In addition, the waveforms are shaped before being saved to disk making identification and energy estimation somewhat more complicated. In this study an alternative technique for reconstruction of event energy in EXO-200 using Boosted Regression Trees from sklearn is explored.[4]