Accelerating the development of novel technologies for nuclear physics
Lead Research Organisation:
University of Liverpool
Department Name: Physics
Abstract
Within the field of Nuclear Instrumentation, the utilisation of Pulse Shape Analysis (PSA) for gamma-ray localisation has proven immensely useful in pushing the envelope of Nuclear Physics both in Nuclear Structure and by improving the tracking and imaging capabilities of systems for industrial and medical applications. As the demand from the nuclear community for these tracking detectors increases, the requirement for efficient PSA algorithms and accurate simulations of novel detector geometries is paramount to ensuring that the UK Nuclear Physics community remains at the forefront of innovation. This project focuses on using high performance computing systems to accelerate the development of techniques primarily for one of the state-of-the-art nuclear spectrometers in Europe, the AGATA array. The techniques will process large data sets to determine where the gamma rays interact within the detector, underpinned by a theoretical understanding of how we expect the AGATA detectors to respond.
The move from conventional CPU-based algorithms to hyper-parallelised GPU-based algorithms is one avenue of research that shows promise. These techniques show great performance scaling with GPU cores with some algorithms capable of running several orders of magnitude faster (~450x) than conventional PSA techniques used in AGATA whilst providing comparable (and sometimes superior) accuracy. GPU-based algorithms provide a marked benefit for large-scale projects like AGATA, not only are they able to meet the processing requirements for online-PSA and tracking but they also free up resources for more complex analyses to be performed. Current resources have been sufficient to prototype the use of these techniques on simple cases, such as when a gamma-ray interact just once within an AGATA detector. However, when the gamma-ray interacts many times, as is likely when the gamma-ray is very energetic, it becomes much more challenging to identify where those interactions have taken place. This project will use high performance computing equipment to accelerate the development of GPU-based algorithms that can unpick the complexity of multiple-interaction events at a local level. In the long term, the computing equipment can also be used to accelerate other research tasks that are important to the UK nuclear physics community, such as running simulations to design new radiation sensors for use in nuclear structure physics experiments.
The move from conventional CPU-based algorithms to hyper-parallelised GPU-based algorithms is one avenue of research that shows promise. These techniques show great performance scaling with GPU cores with some algorithms capable of running several orders of magnitude faster (~450x) than conventional PSA techniques used in AGATA whilst providing comparable (and sometimes superior) accuracy. GPU-based algorithms provide a marked benefit for large-scale projects like AGATA, not only are they able to meet the processing requirements for online-PSA and tracking but they also free up resources for more complex analyses to be performed. Current resources have been sufficient to prototype the use of these techniques on simple cases, such as when a gamma-ray interact just once within an AGATA detector. However, when the gamma-ray interacts many times, as is likely when the gamma-ray is very energetic, it becomes much more challenging to identify where those interactions have taken place. This project will use high performance computing equipment to accelerate the development of GPU-based algorithms that can unpick the complexity of multiple-interaction events at a local level. In the long term, the computing equipment can also be used to accelerate other research tasks that are important to the UK nuclear physics community, such as running simulations to design new radiation sensors for use in nuclear structure physics experiments.