At 21:22:39 UTC on 16 June 2023, two tiny flashes appeared inside the liquid xenon at the heart of the LUX-ZEPLIN experiment. Together they marked one interaction, reconstructed deep within the detector and carrying far more recoil energy than the events emphasized in conventional searches for weakly interacting massive particles.
The event did not arrive with a label saying dark matter. What made it unusual was the failure of ordinary explanations. Its position, energy and two light signals looked consistent with a xenon nucleus recoiling after being struck. The experiment’s surrounding veto detectors did not identify an accompanying particle, and the estimated background was extremely low in that part of the data.
In a new LUX-ZEPLIN collaboration paper, the event drives several heavy-dark-matter models to local significances above three standard deviations, or sigma. The largest local value is 3.4 sigma. Once researchers account for the many models searched, however, the global result falls to 2.6 sigma.
That is the balance the finding demands. It is LZ’s most compelling dark-matter candidate so far, according to the experiment’s Berkeley Lab announcement. It is also one event, unable on its own to distinguish a new particle from a rare background that the model has not captured.
A dark-matter detector built around two flashes
LZ operates at the 4,850-foot level of the Sanford Underground Research Facility in Lead, South Dakota. Nearly a mile of rock filters out most cosmic rays before they can reach the experiment. Inside, the full instrument contains about 10 tonnes of ultrapure xenon, with seven tonnes serving as the active central target.
The detector is a dual-phase time projection chamber. When a particle deposits energy in liquid xenon, it creates an immediate pulse of scintillation light known as S1. The impact also frees electrons. An electric field drifts them upward into a layer of xenon gas, where they produce a second pulse, S2.
The delay between S1 and S2 reveals how deep the event occurred. The pattern of light across photomultiplier tubes at the top gives its horizontal position. Most importantly for this search, the relative sizes of the signals help separate two broad classes of interaction.
Gamma rays and beta particles tend to scatter from electrons, producing electron recoils. A WIMP should instead strike a xenon nucleus, producing a nuclear recoil. Neutrons and neutrinos can also move a nucleus, so even an event in the nuclear-recoil band is not automatically dark matter.
LZ therefore surrounds its xenon chamber with additional xenon, liquid scintillator and 229 tonnes of ultrapure water. Those layers help tag gamma rays, neutrons and muons that might imitate a signal. The depth underground, material screening and concentric vetoes are all parts of the measurement rather than mere infrastructure.
Space Daily previously examined why experiments searching for extremely rare events care about metals with exceptionally low natural radioactivity. LZ used specially selected titanium for its cryostat. When one unexplained interaction matters, a few extra decays from the apparatus can change the conclusion.
One event survived in a newly opened energy window
The collaboration reanalysed 220 live days collected between 27 March 2023 and 1 April 2024. After accounting for the detector’s working mass and time, the exposure was 2.84 tonne-years.
An earlier search through the same data concentrated on nuclear recoils up to about 55 keV. That range is well suited to the simplest spin-independent and spin-dependent WIMP models, which predict that most recoils will occur at low energies. The new work expanded the window to almost 270 keV to test less conventional interactions that could leave a harder spectrum.
After the selection criteria were applied, 1,710 events remained across the broad search region. Almost all belonged to recognizable populations of electron recoils from radon-chain beta decays, neutrinos, xenon isotopes and other sources. One point stood out from those bands at high energy.
The collaboration reconstructed it as an elastic nuclear recoil of 248 keV, with statistical and systematic uncertainties of 23 keV each. It occurred 26.4 centimetres above the cathode and more than 20 centimetres inward from the detector wall. Its S2 shape was consistent with a point-like interaction at that depth, and the distribution of S1 light between the upper and lower photomultiplier arrays agreed with the reconstructed position.
The event sat 1.5 sigma below the middle of the modelled nuclear-recoil band but 6.7 sigma below the middle of the electron-recoil band. In the narrow high-energy projection containing it, the paper estimates only about 0.0106 background events. That does not mean the background probability is exactly 1.06 per cent; the final statistical result uses the event’s complete position in signal space, uncertainties and many tested models.
There is another restraint in the paper that deserves equal attention. The shape of the first light pulse could not conclusively classify the event as a nuclear recoil rather than an electron recoil. The full evidence favours the nuclear-recoil interpretation, but the interaction type was inferred, not directly observed.
Researchers tried to give the event an ordinary cause
An unexplained event in a rare-event detector often stops being mysterious once investigators examine it closely. It may sit near the wall, coincide with activity in an outer veto, contain more than one scatter or combine two unrelated pulses by accident.
The LZ team modelled those possibilities along with beta and gamma radiation, activated xenon and iodine isotopes, detector radioactivity, solar and atmospheric neutrinos, and neutrons produced in materials. The neutron veto efficiency for relevant radioactive processes was estimated at 92 per cent. Wall-related events and interactions that lose part of their electrical charge were also included.
None supplied a likely explanation for this particular event. It looked valid in timing, location and pulse structure, and it lacked the usual telltale accompaniment of a tagged background. That is the reason one point can be scientifically interesting inside a dataset containing thousands.
“No likely background explanation” is nevertheless different from “no possible background explanation.” A process can be absent from a model because it is extremely rare, poorly measured or not yet imagined. With one event, researchers cannot infer a population and compare its distribution against a candidate mechanism.
The analysis also could not use a fully successful blind procedure at the highest energies. Artificial signal-like events, inserted to reduce analyst bias, did not adequately cover that region. The collaboration fixed its selections and likelihood models before revealing the remaining artificial events, but it describes the final work as a non-blind analysis. That does not invalidate the result; it is part of judging its strength.
If it was dark matter, it was not the simplest kind
Weakly interacting massive particles, or WIMPs, are a broad family of hypothetical particles rather than one prediction with one mass. The simplest versions scatter elastically from nuclei without a strong dependence on transferred momentum. Decades of searches have pushed the allowed interaction rates for those models to very low levels.
The new analysis tested a wider effective-field-theory framework. Different operators alter how an interaction depends on nuclear spin, velocity and momentum transfer. It also included inelastic models in which an incoming dark-matter particle changes into a heavier state during the collision. Such models can suppress low-energy recoils and make an event near 248 keV more plausible.
If one of those dark-matter models caused the event, the WIMP would probably have a mass of at least 200 GeV/c², or more than 200 times the proton’s mass. Some of the best-fitting examples use masses near 1,000 GeV/c². The event does not select one operator, mass or inelastic splitting uniquely; several quite different models can be adjusted to resemble one point.
This is direct-detection particle physics, distinct from the astronomical evidence that unseen mass shapes galaxies and the large-scale universe. Space Daily has previously explained why ordinary matter accounts for only a small part of the cosmic inventory. LZ is asking the harder laboratory question: what particle, if any, produces the gravitational dark matter inferred from the sky?
Why 3.4 sigma became 2.6 sigma
The distinction between local and global significance is essential. Suppose researchers test one precisely defined WIMP model. A fluctuation fitting that model unusually well can be assigned a local significance. In this search, the most favourable combinations reached 3.4 sigma.
But the collaboration did not test only one combination. It scanned many interaction operators, WIMP masses and, for inelastic cases, mass splittings. The more possibilities inspected, the greater the chance that ordinary random data will happen to resemble at least one of them. This is the look-elsewhere effect.
Using simulated datasets to account for that search, the collaboration obtained a global significance of 2.6 sigma. Berkeley Lab translates the result as roughly a 0.5 per cent chance of obtaining the effect from known backgrounds under the statistical framework. The exact tail probability depends on convention, and it should not be read as a 99.5 per cent probability that the event is dark matter.
A significance calculation is conditional on a background model and analysis procedure. Unknown or mismodelled backgrounds are not magically included. Particle physics usually demands five sigma before using the word discovery, partly because repeated experience shows that interesting two- and three-sigma effects often weaken when more data arrive.
Even five sigma would not by itself identify a particle. Researchers would still need a coherent recoil spectrum, checks against time and position, compatibility with other experiments and an interaction model that survives independent tests.
The next dataset can strengthen the hint or dissolve it
LZ continues to collect data and is working toward a total exposure of 1,000 live days. If the candidate belongs to a real dark-matter population, more exposure should eventually produce additional events with energies and detector positions consistent with the same underlying model. The statistical preference should grow in a predictable way.
If the point was a rare ordinary interaction, the picture may change differently. It could remain alone, become less significant as clean exposure accumulates or be joined by events that reveal a previously missed background population. New calibration work may also sharpen the boundary between high-energy nuclear and electron recoils.
For now, the detector has done exactly what it was designed to do: suppress familiar noise so thoroughly that one stubborn event becomes worth showing to the wider physics community. The result is more interesting than another exclusion limit, but it remains far less than a detection.
One xenon nucleus may have been struck by a particle that fills the galaxy while passing invisibly through Earth. It may instead record an extraordinarily scarce process inside an extraordinarily quiet machine. At 2.6 sigma, both possibilities remain open, and the most important measurement is the one LZ has not collected yet.