Advanced electronic-signal data analytics is opening up new opportunities for physical science research, ranging from nuclear processes in the Big Bang  and neutron-star mergers, to how we can create radionuclides for medical diagnosis and treatment through gamma-induced fission and gamma-induced charged-particle emission.
The Laser and Gamma-beam facility: Extreme-Light Infrastructure - Nuclear Physics (ELI-NP) is currently under construction in Bucharest, Romania. Here, a world-leading gamma-beam facility for nuclear astrophysics, nuclear structure and medical science is underway. As part of these developments, the present project aims to develop novel charged-particle detection facilities offering unique capabilities for embedded electronic signal analysis (pulse-shape analysis and pulse-shape discrimination), with on-line identification of the emitted particles. Specifically, the present project will focus on the development and implementation of a novel pulse shape analysis of signals from silicon semiconductor detectors [2,3]. This will be developed through application of machine learning tools and artificial intelligence (AI) techniques, and will be integrated directly into the FPGA-based analysing electronics enabling on-board data analysis in real time [4,5].
The primary outcome of the project will be the development of a new identification method for particles in gamma-beam experiments, developed through artificial intelligence and machine learning (ML) methods, based on extensive experimental data. The focus of the project spans the following areas:
Assessment of detector performance and response: (a) assessment and analysis of existing data on silicon detector pulse shape response; (b) recording of pulse-shape data from new silicon semiconductor detectors; (c) simulation of detector-response for radiation types of interest;
Development of AI- and ML-optimised electronic particle identification: (a) preparation of available data for machine learning input (b) development of novel pulse shape analysis implementations through machine learning; and (c) integration of these algorithms directly into Field Programmable Gate Arrays (FPGAs), programmed in VHDL.
Demonstration of the capability of the detector through in-beam experiments at the new international gamma-beam facility: Extreme-Light Infrastructure - Nuclear Physics in Bucharest, Romania.
Methodologies and Techniques:
In-lab benchmarking will include work with silicon detectors and fast (100MHz-1GHz) digital electronics. ROOT (CERN analysis package, written in C++) may be utilised for detailed off-line analysis of observed pulse shapes; with detector response simulations in GEANT4; and an on-line FPGA based implementation of the developed routines through VHDL.
 Measurement of the 7Li(γ,t)4 He ground-state cross section between Eγ=4.4 and 10 MeV; M. Munch et al., Phys. Rev. C 101 (2020) 055801
 Particle identification using digital pulse shape discrimination in a nTD silicon detector with a 1 GHz sampling digitizer; K.Mahata et al., Nuc. Inst. Methods. A 894 (2018) 20-24;
 Digital pulse-shape analysis with a TRACE early silicon prototype; D. Mengoni, et al.,
Nucl. Instrum. Meth. A 764 (2014) 241-246
 Approximate Multiply-Accumulate Array for Convolutional Neural Networks on FPGA
Wang, Z., Trefzer, M. A., Bale, S. J. & Tyrrell, A. M., 2019 14th International Symposium on Reconfigurable Communication-centric Systems-on-Chip (ReCoSoC).
 Adaptive Integer Quantisation for Convolutional Neural Networks through Evolutionary Algorithms Wang, Z., Trefzer, M. A., Bale, S. J. & Tyrrell, A. M., 2021 Int. IEEE Symposium Series on Computational Intelligence SSCI/ICES.
Visit our webpages for more information