SPINE: A machine-learning program is on the trail of ghost particles
Neutrinos are among the most abundant particles in the universe, but they are also among the most elusive. Nearly massless, they are produced by nuclear reactions in the sun, exploding stars, particle accelerators and even the decay of radioactive elements on Earth. They interact so weakly with ordinary matter that trillions pass through each person every second without leaving a trace. Studying…
Neutrinos, the nearly massless particles produced by various cosmic events, are notoriously difficult to study due to their weak interactions with ordinary matter. This has made neutrinos one of the biggest enigmas in physics, potentially shedding light on dark matter and the matter-antimatter asymmetry. To detect these elusive particles, researchers construct massive detectors filled with liquid argon, which reacts to neutrino collisions by emitting light and charged particles.
However, the resulting data is highly complex, containing multiple overlapping signals and background noise. To tackle this challenge, SLAC National Accelerator Laboratory has developed a machine-learning program called SPINE, short for Scalable Particle Imaging with Neural Embeddings. SPINE analyzes liquid argon time-projection chamber data to identify particles, trace their paths, and reconstruct their interactions.
Unlike previous reconstruction methods that relied on a series of specialized algorithms, SPINE uses multiple machine-learning models working together in a hierarchical manner. This approach makes SPINE more transparent and allows researchers to understand how it arrives at its conclusions. SPINE has already been deployed on eight different detectors across 25 institutions worldwide, and recently passed a significant milestone with the first neutrino physics result from the ICARUS detector.
This analysis, published on the arXiv preprint server, demonstrated that SPINE could successfully process real experimental data, recovering a higher percentage of usable neutrino interactions than conventional methods, with improved energy resolution and reduced systematic uncertainties.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
