Sub-GeV Dark Matter Detection with Dark Rates in Liquid Scintillators

It was recently shown that standard sub-GeV dark matter candidates can be effectively probed by large neutrino observatories via annual modulation of the total photomultiplier hit rate. That work focused on the production of light by the excitation of scintillator molecules and considered the JUNO detector, surpassing limits from dedicated dark-matter detectors and reaching theoretical targets. Here, we significantly generalize that work, now also taking into account ionization channels and extending the analysis to other liquid-scintillator detectors, including SNO+, Daya Bay, Borexino, and KamLAND. Last, we present a call to action: with multiple detectors achieving competitive sensitivity, there is an opportunity to validate this new technique across experiments and to refine it using each detector’s strengths.

💡 Research Summary

The paper expands on the recently proposed method of detecting sub‑GeV dark‑matter (DM) particles by exploiting the annual modulation of the total photomultiplier‑tube (PMT) dark‑count rate in large liquid‑scintillator (LSc) neutrino detectors. The original study focused on the Jiangmen Underground Neutrino Observatory (JUNO) and considered only the fluorescence (excitation) channel, where DM excites bound‑state electronic transitions in the scintillator molecules, producing a modest amount of light that can be summed over many DM‑induced events. This work makes two major advances. First, it adds the ionization channel, in which a DM particle with sufficient kinetic energy ionizes a molecule, leading to a cascade of scintillation photons. Including ionization dramatically improves sensitivity for heavier DM masses, especially when the interaction is mediated by a heavy particle (momentum‑independent form factor). Second, the analysis is generalized to four additional LSc detectors—SNO+, Daya Bay, Borexino, and KamLAND—providing a multi‑experiment perspective.

The authors adopt a spin‑independent DM–electron interaction framework, with a local DM density of 0.4 GeV cm⁻³ and a standard Maxwell‑Boltzmann velocity distribution (v₀ = 230 km s⁻¹, Earth velocity v_E = 240 km s⁻¹, escape speed v_esc = 600 km s⁻¹). For the excitation channel they model the target electrons using the benzene π→π* transition (≈5 eV) and count only the six delocalized π‑electrons per molecule, yielding a conservative number of active electrons N_exc = 6 N_molecules. The molecular form factor is taken from semi‑empirical Hartree‑Fock calculations calibrated to UV absorption data. The ionization channel is treated analogously to free‑electron scattering, integrating the differential rate dR_ion/dln E_er over recoil energies. Both channels are multiplied by detector‑specific efficiency factors (ξ_exc, ξ_ion) that encode the probability that the emitted photons are collected and registered.

Two benchmark mediator models are considered: a heavy mediator (F_DM = 1) and a light mediator (F_DM ∝ 1/q²). The heavy case yields a q‑independent cross‑section, while the light case enhances low‑momentum transfer, improving sensitivity to low‑mass DM. The total event rate is R_tot = R_exc + R_ion, and the observable is the fractional annual modulation of the total dark‑count rate, driven by the Earth’s motion through the Galactic halo (≈±5 % modulation). Although the absolute dark‑count rates are large (10⁶–10⁸ Hz), the statistical uncertainty scales as √N, so with multi‑year exposures (up to 10 years for JUNO) the modulation can be detected at >5σ significance.

Detector parameters are summarized in Table I. JUNO contains 20 kt of linear alkylbenzene (LAB), 78 % PMT coverage, and a total dark rate of ~6×10⁸ Hz, providing the highest statistical power. SNO+ uses 780 t LAB with 54 % coverage and ~10⁷ Hz dark rate; a future tellurium phase will slightly reduce light yield. Daya Bay comprises eight antineutrino detectors with a combined 320 t LAB+Gd target, 12 % coverage, and ~2.3×10⁶ Hz dark rate. Borexino employs 278 t pseudo‑cumene (PC) with 34 % coverage and a low dark rate of 8.8×10⁵ Hz, benefiting from a high light yield (~11.5 γ keV⁻¹). KamLAND contains 1 kt of an 80 % dodecane / 20 % PC mixture, 34 % coverage, and ~1.3×10⁷ Hz dark rate.

Sensitivity projections show that all five detectors can surpass existing direct‑detection limits in the sub‑GeV regime. For a heavy mediator, JUNO reaches σ̄_e ≈ 10⁻³⁸ cm² at m_χ ≈ 10 MeV, while Borexino achieves comparable limits at lower masses due to its low background. For a light mediator, the reach improves to σ̄_e ≈ 10⁻⁴⁰ cm². The ionization channel dominates the reach for m_χ ≳ 10 MeV, whereas excitation dominates below a few MeV.

A key message is the power of cross‑checking results across detectors with different locations, target chemistries, and PMT configurations. Independent systematic uncertainties allow a combined analysis to tighten limits and to verify any potential signal. The authors outline future work: precise measurements of scintillation spectra, time‑profile studies of ionization‑induced bursts, and detailed modeling of PMT dark‑count temperature and voltage dependencies. Implementing these refinements could transform large neutrino observatories into competitive, multi‑experiment platforms for probing sub‑GeV dark matter, complementing dedicated low‑threshold experiments and opening a new avenue for dark‑matter discovery.