XENONnT Extends Light Dark Matter Limits With S2-Only Data

Conventional dual-phase xenon searches use both prompt scintillation (S1) and delayed ionization (S2) signals, but requiring S1 loses sensitivity when a recoil is too small to produce a usable prompt pulse. XENONnT instead analyzes S2-only events from 579 days across three science runs, accepting the harder task of separating genuine low-energy recoils from cathode events, delayed electrons, accidental electrons, and other detector backgrounds. The paper reports no significant excess and turns that null result into new constraints on several light-dark-matter interactions.

Core Contribution

The central advance is a complete S2-only background model for XENONnT coupled to dedicated event-quality and background-suppression selections. The authors use it in a blinded search with 7.83 tonne-years of exposure, targeting nuclear recoils from 0.5 to 5.0 keVnr and electronic recoils from 0.04 to 0.7 keVee. This is not simply a lower-threshold version of the collaboration’s S1–S2 analysis: removing S1 sacrifices direct depth information and increases sensitivity to single- and few-electron detector activity, so the analysis depends on whether the background components can be distinguished from signal in the remaining S2 observables.

The paper’s result is therefore as much a detector-analysis result as a limit-setting result. It demonstrates that an ionization-only selection can be calibrated and modeled across the full XENONnT science exposure well enough to make an inference in the low-mass region where conventional xenon analyses lose acceptance.

Technical Approach

The S2-only signal efficiency combines the S1 region-of-interest requirement, the S2 region of interest, and an S2 quality selection. Figure 1 compares the resulting exposure-weighted efficiency across SR0, SR1, and SR2 with the previous S1–S2 analysis, alongside predicted spin-independent dark-matter recoil spectra for 3, 6, and 8 GeV/c² masses and the solar ⁸B coherent-elastic-neutrino-scattering spectrum. That comparison makes the trade-off visible: the S2-only channel retains access to smaller ionization signals, but its utility rests on background rejection rather than on a clean two-signal event reconstruction.

For background validation, the authors fit and compare modeled components in four dimensions: corrected S2 size (cS2), an S2 CNF score, a cathode BDT score, and S2 width. The enlarged-region comparison includes modeled cathode, delayed-electron, accidental-electron, and another background contribution, with statistical and systematic uncertainty bands on the projections. Separate ²²⁰Rn and ²²²Rn calibrations test the model. The ²²⁰Rn comparison shows a mismatch in the S2 CNF score, which the paper attributes to its much higher gamma activity relative to the science data; the lower-rate ²²²Rn calibration agrees in all four dimensions. That distinction is useful evidence, but it also identifies an analysis feature whose transfer between calibration and science conditions must be handled carefully.

Results and Analysis

After all selections, the science-region data have a background shape similar to the signal shape, and the inference is performed in cS2. Figure 3 shows the combined science data and best-fit background projections; the reported outcome is no significant excess. The analysis sets 90% confidence upper limits for spin-independent dark-matter–nucleon and spin-dependent dark-matter–neutron scattering between 3 and 8 GeV/c², as well as for dark-matter–electron scattering, axion-like particles, and dark photons.

The clearest numerical benchmark is the spin-independent limit: cross sections above 6.0×10⁻⁴⁵ cm² are excluded at a dark-matter mass of 5 GeV/c². Figure 4 places this result alongside prior XENON10, XENON1T, XENONnT, LZ, PandaX, DarkSide, and SuperCDMS constraints and shows the approach toward the neutrino-fog region. The practical significance is narrow but real. At 5 GeV/c², the analysis improves constraints in a mass interval where recovering low-energy events matters more than maximizing discrimination from S1–S2 topology. It does not establish a signal; it establishes that the background model and exposure support a stronger exclusion.

Limits in Practice

The evidence is strong for the stated exclusion within the modeled region: the search is blinded, spans three science runs, uses a large exposure, validates its model with calibration data, and compares data to a fitted multi-component background. The more constrained interpretation is that sensitivity is now closer to an irreducible neutrino background for spin-independent scattering, not that this analysis can resolve that background or explore arbitrary low-mass models. Its reach is bounded by the stated recoil-energy windows, the 3–8 GeV/c² nuclear-scattering mass range, and modeling choices needed to control S2-only backgrounds.