Sep 12 – 17, 2021
University of Birmingham
Europe/London timezone

The Development of Novel Pulse Shape Analysis Algorithms for the Advanced Gamma Tracking Array (AGATA)

Not scheduled
20m
Teaching and Learning Building (University of Birmingham)

Teaching and Learning Building

University of Birmingham

Edgbaston Campus University of Birmingham B15 2TT UK
poster Applications in Nuclear Physics and Nuclear Industry

Speaker

Mr Fraser Holloway

Description

The Development of Novel Pulse Shape Analysis Algorithms for the Advanced Gamma Tracking Array (AGATA)

F. Holloway$^1$, LJ. Harkness-Brennan$^1$, D. Judson$^1$, V. Kurlin$^1$
$^1$The University of Liverpool, UK

I. INTRODUCTION

Standing at the forefront of Gamma-Ray Spectroscopy, the Advanced GAmma Tracking Array (AGATA) provides insight into a wide variety of Nuclear Physics, by employing Gamma-Ray Tracking (GRT) AGATA provides significant improvements to efficiency, doppler correction and position resolution.

As all tracked photons occur solely within the germanium volume without the need of additional ancillaries a critical component of GRT the field of Pulse Shape Analysis (PSA) is required. PSA uses characteristics of the measured signals from the segmented electrodes of each crystal to directly infer the positions of the gamma-ray interactions.

II. DEVELOPMENT OF NEW PSA TECHNIQUES

The use of Convolutional Neural Networks for regression was investigated and offers a viable solution on experimental data offering reasonable prediction accuracy and operating at 3kHz.

Utilising graph accelerated k-Nearest Neighbour algorithms for Fold-1 interaction prediction combined with Manifold-Learning assisted dimensionality reduction has allowed for a significant improvement in processing rate with little loss to prediction accuracy. In particular the algorithms, FAISS, HNSWLIB, MKS and Nanoflann were profiled on various embeddedings.

Adaptions were made to extend these methods to work on High-Fold data utilising hierachical restrictive geometry. Precomputation of these scenarios into hyper-efficient hierarchical structures provides a robust, dynamic and efficient approach to High-Fold PSA.

III. EXPERIMENTAL CHARACTERISATION

In order to profile the performance of these algorithms the algorithms were assessed blindly using experimental data collected at Liverpool and IPHC Strasbourg. Data from $^{137}$Cs, $^{241}$Am & $^{152}$Eu sources was used to evaluate PSA performance. The full position and energy dependence of these PSA methods is evaluated and compared against the industry standard.

Nationality British
Institute The University of Liverpool
Your name Fraser Holloway
Title Dr
email f.holloway@student.liverpool.ac.uk

Primary authors

Dr Daniel Judson (The University of Liverpool) Mr Fraser Holloway Prof. Laura Harkness-Brennan (The University of Liverpool) Dr Vitaliy Kurlin (The University of Liverpool)

Presentation materials

There are no materials yet.