An atmospheric-neutrino simulation predicts approximately 2,907 neutral-current (NC) and 2,296 charged-current (CC) single-signal events for a 200 kiloton-year exposure in a large liquid-scintillator detector. The study's central finding is not just the rate. Among the tested alternatives, the predicted prompt-energy spectrum changes most when researchers vary the model used for the initial neutrino-nucleus interaction.
NC and CC are the paper's two interaction categories. A single is an event with a prompt deposit that survives the delayed-signal vetoes used in the selection. NC singles dominate at low prompt energies, especially below about 100 MeV, while CC singles dominate at higher energies. Prompt energy means the visible energy deposited in the first 1008 nanoseconds, with a threshold of 0.7 MeV.
What the simulation assumes
The forecast assumes a 20-kiloton target made of 88% carbon-12 and 12% hydrogen by mass. It combines the atmospheric-neutrino flux before oscillation, oscillation probabilities and neutrino-nucleus cross sections, then applies detector response and event selection. The rate calculation covers total neutrino energies from 0.1 to 20 GeV.
The baseline interaction setup is a GENIE calculation, configuration G18 10b 02 11b, version 3.04.02, using a local Fermi gas nuclear model. The authors compare it with alternative choices for the event generator, nuclear model, the treatment of interactions after the first collision, de-excitation and secondary interactions.
To keep events with only one visible signal, the analysis requires full containment and vetoes neutron-capture signals in the 1.9 to 2.5 MeV or 4.4 to 5.5 MeV bands between 1008 nanoseconds and 1000 microseconds after the prompt. It also vetoes Michel-electron signals from 6 to 53 MeV during the first 10 microseconds.
The detector response is idealized. It omits position-dependent energy reconstruction, energy-resolution smearing, trigger efficiency and delayed-signal tagging efficiencies.
The biggest uncertainty starts at the first collision
Once the inputs are varied, primary neutrino-nucleus interaction models produce the largest tested differences in the singles spectra. In the NC category, changes in the nuclear model are somewhat more pronounced than changes in interactions after the first collision, and the GENIE-versus-NuWro local-Fermi-gas comparison is the largest. In the CC category, the main differences come from the Valencia-versus-Llewellyn-Smith choices.
Another stage, known as de-excitation, changes what the residual nucleus emits after the interaction. An event without an outgoing neutron can acquire one, while emitted protons add low-energy visible energy. TALYS and GEMINI++4ν give different neutron-emission estimates, but their disagreement is smaller for neutron-free events than for events with one neutron. Selected singles are dominated by neutron-free topologies, and added proton energy is concentrated below about 20 MeV.
Secondary-interaction models are a subdominant but non-negligible uncertainty. BIC, INCL++ and BERT+PC differ by only a few percent from nominal BERT. Above 100 MeV, however, their spread can be comparable to or larger than the difference between TALYS and GEMINI++4ν.
The uncertainty estimate uses each model class's prediction envelope: the largest prediction minus the smallest, divided by their sum. The flux, primary-interaction, de-excitation and secondary-interaction terms are then combined by adding their squared uncertainties and taking the square root, with the sources treated as independent. Only TALYS versus GEMINI++4ν enters the de-excitation uncertainty; the no-de-excitation case is excluded.
For NC singles, primary-interaction uncertainty is about 22% from 15 to under 100 MeV and about 31% from 0.7 to under 15 MeV. That compares with flux uncertainties of about 18% and 23% and statistical uncertainties of about 3% and 5%, respectively, making interaction modeling the largest listed low-energy systematic.
The energy windows tell different stories
Over 10 years in a 20-kiloton detector, the predicted NC counts are 453 from 0.7 to under 15 MeV, 1,337 from 15 to under 100 MeV and 1,066 from 100 to 1,000 MeV. The corresponding CC counts are no more than approximately 0.01 in the lowest band, followed by 21 and 1,750.
A separate low-energy calculation estimates approximately 78 singles below 15 MeV for a 200 kiloton-year exposure. It restricts the calculation to elastic scattering of atmospheric neutrinos with energies below 100 MeV on free protons, and it does not evaluate systematic uncertainty for that contribution. The figure therefore describes a restricted component, not a complete sub-100-MeV prediction.
The study also asks whether the prompt-energy shape, apart from the total count, changes across models in the 15 to 100 MeV range. After the predictions are normalized to the same total, GENIE LFG+G4B shows the largest tested departure from nominal, with a covariance-weighted chi-square per degree of freedom of 2.35. The corresponding values are 0.02 for GENIE RFG+hN, 0.20 for NuWro LFG+Csc and 0.17 for NuWro SF+Csc. The comparison includes the stated statistical and systematic uncertainties.
A background to watch, and a possible calibration sample
Near 2.5 MeV, in the rare-event search window the paper calls the 0νββ region of interest, the estimated NC-induced background is about 3 × 10−7 events per keV-ton-year of liquid scintillator. The paper presents PSD as a possible way to reduce that background, but the interpretation depends on detector-level PSD performance.
At energies below 100 MeV, the singles sample is proton-recoil enriched and is proposed as a control sample for timing validation.
The authors frame the sample as a possible probe of neutrino-nucleus interaction models as well as a source of calibration and rare-event background information.
The preprint's boundary
The detector response is idealized and leaves out position-dependent reconstruction, energy-resolution smearing, trigger efficiency and delayed-signal tagging efficiencies. No detector-level sensitivity estimate is provided.
The document is a preprint identified as arXiv:2608.28095v1, dated 28 Aug 2026. Its uncertainty estimates apply only to the configurations tested in the study.
Paper data and sources
Original title: Characterizing Single-Signal Events from Atmospheric-Neutrino Neutral-Current Interactions in Large Liquid Scintillator Detectors
Authors: Zhenning Qu, Jie Cheng, Wan-lei Guo et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-28
DOI: Not available
Original paper · Full text