The LUX-ZEPLIN (LZ) experiment reported one unusually high-energy nuclear recoil at 248 ± 23(stat) ± 23(sys) keV in 2.84 tonne-years of data. This single event has drawn interest because a simple form of dark matter called inelastic (or endothermic) dark matter can naturally suppress low-energy recoils and push signals to higher energies. The paper performs a model-independent check of whether inelastic, spin-independent dark matter scattering could explain that lone event without predicting too many events in nearby energy ranges.

Rather than commit to a specific particle model, the authors form two “sideband-to-signal” ratios: the predicted counts in a low-energy window (14–225 keV) and a high-energy window (271–800 keV) divided by the prediction in the signal window (225–271 keV). Because those ratios cancel the overall scattering strength, the test depends only on the spectral shape set by the kinematics of inelastic scattering, the assumed dark-matter speed distribution (the Standard Halo Model, a truncated Maxwellian), and the xenon nuclear response modeled by the Helm form factor. This lets their conclusions apply broadly to coherent spin-independent interactions with a momentum-independent normalization.

A key finding is that the two sideband ratios move in opposite directions as you change the inelastic mass splitting δ (the extra energy the dark matter must absorb). There is a unique balanced splitting δ* where the low- and high-sideband predictions are equal and the joint expected sideband count is minimized. For a 1 TeV dark-matter mass they find δ* ≈ 339 keV and a predicted count of about 1.77 events per sideband, which gives a combined Poisson tension of roughly 1.9 standard deviations with the observed zero events in those sidebands. Scanning mass from 500 GeV to 3 TeV, the balanced tension grows from ≲1.3σ to ≈2.5σ, so heavier masses are more constrained under these assumptions.