Science

ETH Zurich prototype tracks invisible particles in 3D

Researchers have built and tested a detector that captures the paths of elusive particles inside a single block of scintillator using a light-field camera, sensitive photon sensors and artificial intelligence. The advance points toward simpler, more scalable instruments for fundamental physics experiments.
Listen
AI-generated image: ETH Zurich prototype tracks invisible particles in 3D
AI-generated image for illustrative purposes.
Intelligent summary
  • PLATON uses a light-field camera and SPAD sensor to capture photon direction and intensity inside unsegmented scintillator.
  • Artificial intelligence based on Transformer architecture reconstructs particle tracks in three dimensions.
  • Laboratory tests achieved usable resolution down to five detected photons; simulations predict sub-millimetre accuracy in small volumes for neutrino detection.

Researchers at ETH Zurich have developed a prototype detector that can reconstruct particle tracks in three dimensions inside a large, unsegmented block of scintillator material. The system, called PLATON, replaces the millions of separate detection elements found in conventional designs with a single light-field camera and advanced reconstruction algorithms.

Traditional scintillator detectors segment their active volume into vast arrays of small cubes or fibres. The T2K neutrino experiment, for instance, relied on roughly two million such units linked by sixty thousand optical fibres. Scaling those designs to the volumes required for next-generation neutrino, dark matter or collider experiments quickly becomes impractical and expensive. PLATON offers a different route.

The prototype combines a micro-lens array with a highly sensitive SPAD imaging sensor known as SwissSPAD2. When a particle interacts with the scintillator it produces a faint shower of photons. The light-field camera records both the intensity and the direction of those photons, supplying the depth information needed for three-dimensional reconstruction even when only a handful of photons are detected.

A neural network based on Transformer architecture then reconstructs the particle paths. Laboratory tests exposed a plastic scintillator block to electrons from a strontium-90 source. The measured spatial resolution remained usable down to five detected photons and matched detailed simulations.

The work, funded by the Swiss National Science Foundation, appeared in Nature Communications earlier this year. Simulations of an upgraded version suggest that spatial resolution below one millimetre could be achieved inside a ten-centimetre cube for neutrino detection, with high purity and efficiency. For a full cubic metre of unsegmented scintillator the projected resolution relaxes to a few millimetres.