A compact beta particle momentum detector

We present the design, development, and first calibration results of a compact beta electron detector for the Quantum Invisible Particle Sensor (QuIPS) experiment. The QuIPS electron detector is designed to reconstruct the full momentum vector of beta particles emitted from radioisotope-doped optically levitated nanospheres in ultra-high vacuum (UHV), enabling a measurement of the neutrino momentum and a search for heavy sterile neutrinos. The detector comprises two thinned CMOS detectors for directional tracking and a plastic scintillator read out by silicon photomultipliers (SiPMs) for calorimetry. The entire assembly must operate inside an existing optical trapping vacuum chamber at pressures below 10−7 mbar, imposing stringent constraints on material selection, power dissipation, outgassing, and compactness. We demonstrate sensitivity to beta-decay electrons with energies as low as 100 keV and a detection efficiency of 35% above 500 keV, the primary window of interest for a heavy sterile neutrino search. The CMOS tracker resolves the momentum direction at the mrad scale and reconstructs the beta emission vertex with sub-mm precision, while the scintillator-SiPM system achieves an energy resolution of 5% at 1 MeV.

Improvements to AI systems:

1. AI-driven detector calibration optimizer: Train a neural network to map raw CMOS tracker and SiPM-scintillator signals to calibrated momentum vectors and energy values, using the paper’s calibration data (e.g., 5% energy resolution at 1 MeV, mrad angular resolution) as ground truth. This would replace manual calibration, enabling real-time self-calibration during UHV operation and adapting to thermal drift or sensor aging.

2. AI-based event reconstruction with uncertainty quantification: Implement a Bayesian deep learning model that takes multi-channel detector hits (two CMOS planes + scintillator pulse shape) and outputs a full 3D momentum vector, vertex position (sub-mm), and per-event energy with confidence intervals. This improves over current analytic reconstruction by handling non-linear energy losses, multiple scattering, and noise in low-energy (<100 keV) events.

3. AI for background rejection and sterile neutrino search: Use a convolutional or graph neural network on the spatio-temporal hit patterns to classify beta-decay events vs. cosmic rays, dark counts, or trapped-ion backgrounds. The AI can exploit the directional tracking (mrad precision) and energy deposition shape to achieve higher signal-to-background ratio, directly enhancing the sensitivity to heavy sterile neutrinos.

4. AI-optimized detector geometry and material selection: Apply reinforcement learning or generative design to propose alternative compact detector layouts (e.g., thinner scintillators, different SiPM placements) that satisfy UHV constraints (outgassing, power 35% above 500 keV). The AI can simulate trade-offs using the paper’s performance metrics as reward functions.

5. AI for real-time data compression and triggering: Develop a lightweight autoencoder or anomaly detector that runs on the detector’s onboard FPGA, compressing raw event streams (CMOS frames + SiPM waveforms) to only relevant features for telemetry. This enables continuous operation in UHV with limited bandwidth, while preserving all information needed for offline neutrino momentum reconstruction.

What the improved AI system can do:

• Automatically calibrate and self-correct the detector in situ, reducing downtime and human error.

• Reconstruct each beta particle’s momentum vector and energy with quantified uncertainties, even for low-energy events near 100 keV.

• Reject background events with >99% accuracy, improving the statistical power of the sterile neutrino search by orders of magnitude.

• Propose novel detector designs that meet strict vacuum and power limits, potentially increasing detection efficiency to >50%.

• Operate autonomously in remote or space-based experiments, transmitting only compressed, high-level physics features instead of raw data.

We present the design, development, and first calibration results of a compact beta electron detector for the Quantum Invisible Particle Sensor (QuIPS) experiment. The QuIPS electron detector is designed to reconstruct the full momentum vector of beta particles emitted from radioisotope-doped optically levitated nanospheres in ultra-high vacuum (UHV), enabling a measurement of the neutrino momentum and a search for heavy sterile neutrinos. The detector comprises two thinned CMOS detectors for directional tracking and a plastic scintillator read out by silicon photomultipliers (SiPMs) for calorimetry. The entire assembly must operate inside an existing optical trapping vacuum chamber at pressures below 10-7 mbar, imposing stringent constraints on material selection, power dissipation, outgassing, and compactness. We demonstrate sensitivity to beta-decay electrons with energies as low as 100 keV and a detection efficiency of 35% above 500 keV, the primary window of interest for a heavy sterile neutrino search. The CMOS tracker resolves the momentum direction at the mrad scale and reconstructs the beta emission vertex with sub-mm precision, while the scintillator-SiPM system achieves an energy resolution of 5% at 1 MeV.