Speaker
Description
PANDORA (Plasmas for Astrophysics, Nuclear decay Observation and Radiation for Archaeometry) is a multidisciplinary project investigating β decays in stellar-like Electron Cyclotron Resonance (ECR) plasmas, offering a breakthrough perspective for fundamental studies on weak interactions in astrophysical contexts [1,4]. Magnetized plasmas in compact traps become novel platforms for basic research and for applications in ion source technology, accelerator physics, materials science, and beyond. The facility features an advanced diagnostic system that enables more efficient electromagnetic heating and non-invasive measurements of plasma properties, significantly enhancing ECRIS performance. Indeed, plasma parameters critically influence the extracted beam’s current, charge state, emittance, and stability [1].
We have developed an innovative algorithm for X-ray imaging in Single-Photon Counting (SPhC) mode [2,3], enabling space-resolved soft-X-ray spectroscopy and magneto-plasma diagnostics (local thermodynamics, confinement dynamics, structure) via an X-ray pinhole camera. This work presents its further development and optimization through an AI-based machine-learning model implemented in MATLAB [5].
By processing data collected under varying plasma conditions, individual photon events were analysed and grouped based on their spatial and intensity features. We employed K-means clustering–based AI tool to group similar events, revealing parameters that discriminate valid from spurious signals. This approach allowed to distinguish meaningful signals from noise and pile up signals and ultimately to build a more refined dataset for training neural networks capable of identify and discard spurious signals, reducing signal distortions and enhancing data quality. [5] The overall goal is to extract plasma emission spectra more effectively, achieving higher resolution and improved accuracy in the characterization of soft X-ray fluorescence and bremsstrahlung emissions from such plasmas.
Furthermore, achieving higher resolution and an improved signal-to-noise ratio can ultimately enable fast and direct identification of meaningful events, allowing them to be processed in real time. At the same time, spurious contributions can be promptly identified and filtered out from the start—potentially stored for later inspection—thus paving the way for efficient online analysis.
References
[1] D. Mascali et al. Universe 8, 80 (2022)
[2] E. Naselli et al., Condens. Matter, 7(1), 5 (2022)
[3] G. Finocchiaro et al., Phys. Plasmas 31, 062506(2024)
[4] D. Mascali, et al. Condens. Matter 2024,9, 28
[5] B. Peri et al., AI Tools for Plasma Diagnostics by X-ray Imaging and Spectroscopy in the PANDORA Project Frame, EPS 51st Conference on Plasma Physics Proceeding, 2025