LZ’s 248 keV Event: Why One Xenon Recoil Is Not Yet a Dark Matter Discovery

LUX-ZEPLIN (LZ) has reported a single unusual interaction in its liquid-xenon detector at the Sanford Underground Research Facility in South Dakota. Recorded on June 16, 2023, the event is consistent with a xenon nuclear recoil of 248 ± 23 keV (statistical) ± 23 keV (systematic)—an energy near the upper end of LZ’s extended search window. The collaboration found it in a region where its modeled background expectation was low.

That makes the event scientifically interesting, not conclusive. LZ’s own statistical test gives a global significance of 2.6σ after accounting for the search across multiple models; its largest model-dependent local significance was 3.4σ.

What LZ observed

The result comes from a search for interactions that make xenon nuclei recoil. LZ analyzed 2.84 tonne-years of exposure and extended its nuclear-recoil search to roughly 270 keV, beyond the range emphasized in many conventional WIMP searches.

After the analysis selections were applied and artificial “salt” events were removed, one science-data event remained conspicuous in the nuclear-recoil band. LZ labels it LZ230616. Its light and charge signals were consistent with a 248 keV elastic nuclear recoil.

The reconstructed interaction was well inside the detector: about 26.4 cm above the cathode and 26.9 cm from the time-projection-chamber wall. That location is relevant because events near detector boundaries can be harder to reconstruct reliably.

Why a nuclear recoil matters for dark matter

A particle striking a xenon nucleus is one of the signatures direct-detection experiments look for. The event’s unusually high reconstructed recoil energy is especially relevant to models in which the high-energy portion of the recoil spectrum is more prominent than in standard spin-independent WIMP scattering.

LZ designed the extended-energy analysis to probe effective-field-theory and inelastic dark-matter scenarios that can produce a larger fraction of high-energy nuclear recoils. The event therefore has prompted theoretical proposals involving several nonstandard dark-matter interactions. Those proposals are possibilities, however—not evidence that any one model is correct. 1

There is another equally important possibility: the event could arise from a rare or imperfectly modeled background. A single candidate cannot distinguish a new particle from a statistical fluctuation or an overlooked detector or environmental process.

Why it is not a discovery

The central limitation is simple: there is only one event.

LZ reports a 2.6σ global tension with the background-only hypothesis once look-elsewhere effects are included. This is below the roughly 5σ standard normally used to claim a particle-physics discovery.

The distinction between local and global significance also matters. A particular dark-matter model can make the event look more striking at one point in parameter space, producing a maximum local significance of 3.4σ. But the analysis tested multiple models, so the global figure accounts for the opportunity to find an apparent excess somewhere among them.

The collaboration also characterizes the high-energy analysis as non-blind. That does not erase the observation, but it makes an independent, pre-specified follow-up analysis especially important: analysts should define selections and background tests before opening the relevant signal region in new data.

How synthetic events help test the analysis

LZ uses artificial signal-like events, often called salt, to reduce the risk that analysts unconsciously tune an analysis around an apparent candidate. In this high-energy search, the collaboration also constructed manufactured accidental events by combining isolated prompt-light (S1) and charge (S2) waveforms to study how selection cuts handle accidental backgrounds.

These checks do not prove that LZ230616 is dark matter. Their value is methodological: they test whether the event-selection procedure rejects background-like coincidences as expected and help prevent decisions from being tailored to one real outlier.

What would turn the anomaly into evidence

The decisive test is recurrence in independent data under a locked-down analysis. If a dark-matter model explains LZ230616, it should predict not only another isolated event but a rate and recoil-energy distribution that can be tested as exposure grows.

A persuasive case would require several pieces to line up:

• additional high-energy nuclear-recoil candidates in new LZ data;
• an energy spectrum and event rate consistent with one physical interaction model;
• continued success of detector and background checks; and
• compatible results from independent experiments.

Failure to see a compatible population of events would weaken a dark-matter interpretation. Identification of a specific rare background would provide a more ordinary explanation.

How XENONnT and PandaX-4T could check it

XENONnT and PandaX-4T are also liquid-xenon time-projection-chamber experiments, so they offer an important independent test of a xenon-scattering explanation. A meaningful comparison would need searches that cover the relevant high-recoil range and have enough sensitivity to the event rate implied by a proposed model.

Confirmation would not mean merely seeing one unexplained event elsewhere. The experiments would need to observe mutually consistent recoil energies, rates, and interaction parameters. Conversely, sufficiently sensitive null results in the same high-energy regime could rule out the parameter space needed to interpret LZ230616 as dark matter.

Existing PandaX-4T results found no significant nuclear-recoil excess in an approximately 5–100 keV window, but that range is below LZ230616’s reconstructed 248 keV energy. Those results therefore do not, by themselves, directly test this particular high-energy candidate.

The bottom line

LZ has reported its strongest high-energy dark-matter hint to date: one well-reconstructed event in a low-background region, consistent with a 248 keV xenon nuclear recoil. It is worth close scrutiny because high-energy recoil searches can access dark-matter scenarios that standard searches are less sensitive to. But the evidence remains preliminary: at 2.6σ and with only one event, the result is an anomaly to test—not a detection of dark matter.