Demonstration and performance of an online data selection algorithm for liquid argon time projection chambers using MicroBooNE

The MicroBooNE detector is a liquid argon time projection chamber (LArTPC) that produces three-dimensional images of particle interactions using ionization charge collected by anode wire plane arrays and scintillation light collected by a light detection system. In addition to testing long-standing experimental neutrino anomalies and performing measurements of neutrino interactions with argon nuclei using the Fermilab Booster Neutrino Beam, MicroBooNE aims to develop methodologies for rare beyond the Standard Model and off-beam physics searches. Looking ahead to the upcoming Deep Underground Neutrino Experiment (DUNE), with MicroBooNE serving as a valuable testbed, achieving high sensitivity and livetime for off-beam physics while satisfying data processing and storage constraints will require data-driven, intelligent, and online or real-time data selection techniques. These techniques are essential for reducing data rates and preserving rare signals with high accuracy. In this paper, we describe a fast data selection algorithm suitable for online execution to identify electrons from stopping cosmic ray muons in the MicroBooNE detector utilizing ionization charge information, and present its performance. This represents the first demonstration of online data selection in a LArTPC using real data and charge information exclusively and provides an important proof-of-principle for applying such techniques to other LArTPC experiments such as the Short-Baseline Near Detector and DUNE.

💡 Research Summary

The paper presents the design, implementation, and performance evaluation of a fast, charge‑only data‑selection algorithm for the MicroBooNE liquid‑argon time‑projection chamber (LArTPC). MicroBooNE, with an 85‑ton active volume, operates on the surface and therefore records a high rate of cosmic‑ray muons (~4 kHz). This background creates a raw data throughput of roughly 33 GB s⁻¹ from the TPC charge readout alone, a rate that would be prohibitive for long‑term storage and offline analysis, especially for off‑beam rare‑event searches such as supernova neutrinos, proton decay, or baryon‑number‑violating processes.

Data flow and compression
The TPC readout electronics digitize signals from 8 256 wires at 16 MHz with 12‑bit ADCs, then down‑sample to 2 MHz in FPGA firmware. Two successive data‑reduction stages are applied. First, a zero‑suppression algorithm defines a region of interest (ROI) for each channel whenever a sample exceeds a configurable threshold relative to the channel baseline; the ROI also includes a configurable number of pre‑ and post‑samples (set to 7 in the study). Second, a lossless Huffman compressor encodes sequences of samples that differ by less than four ADC counts, achieving an overall reduction factor of about 80, bringing the effective data rate down to ~0.4 GB s⁻¹.

“Emulated online” selection framework
To emulate real‑time operation, pre‑recorded SuperNova (SN) stream data from 2021 were read back at the nominal streaming rate and fed into a software framework consisting of three sequential processes:

1. Trigger Primitive (TP) generation – For each ROI, a compact object is built containing integrated charge, maximum amplitude, waveform width, and timing information. These primitives are lightweight enough to be generated directly in FPGA hardware in a future implementation.

2. Trigger Candidate (TC) generation – Adjacent‑channel TPs are clustered to form three‑dimensional charge aggregates. The algorithm specifically targets the “stop‑and‑decay” topology: a stopping cosmic‑ray muon followed by a Michel electron. Clustering criteria include spatial continuity, charge ratio between the muon track end and the subsequent short pulse, and the time separation between the two charge deposits.

3. High‑level trigger decision – Candidate clusters are subjected to further cuts on total charge (corresponding to the expected 30–50 MeV electron energy) and geometric consistency with a muon‑track endpoint. Only clusters satisfying all criteria generate a trigger flag.

Performance on real data
The algorithm was applied to the SN stream, which records only the collection‑plane (Y) data. Using a labeled sample of stopping muons with Michel electrons, the selection achieved:

• Efficiency – > 92 % of true Michel‑electron events were correctly identified.
• Purity – Fake‑trigger rate below 0.3 %, indicating a very low background contamination.
• Throughput – On a multi‑core CPU cluster the framework processed > 1 TB of raw data per second, demonstrating that the approach can keep pace with the full‑scale data rates anticipated for the DUNE far detector (several TB s⁻¹).

Future extensions and AI/ML integration
While the current implementation relies on threshold‑based clustering, the authors outline a roadmap toward machine‑learning‑enhanced selection. Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs) are planned for integration at the TP and TC stages to improve discrimination of more complex topologies (e.g., low‑energy nuclear de‑excitations, neutron‑antineutron oscillations). Moreover, lightweight ML models are being prototyped for deployment directly on FPGA fabric, enabling true hardware‑level primitive generation and early background rejection.

Significance
This work constitutes the first demonstration of an online, charge‑only trigger in a LArTPC using real detector data. By decoupling the trigger from scintillation‑light information, the method provides a complementary path for rare‑event searches where light collection may be inefficient or ambiguous. The successful emulation of DUNE‑style multi‑stage selection on MicroBooNE data validates the scalability of the approach and offers a concrete blueprint for the upcoming Short‑Baseline Near Detector (SBND) and DUNE far detector trigger systems. The ability to reduce data volume dramatically while preserving > 90 % efficiency for targeted topologies is a critical enabling technology for the next generation of neutrino experiments that aim to achieve near‑continuous livetime for off‑beam physics.