On June 16, 2023, the LUX-ZEPLIN detector, known as LZ, in South Dakota recorded two associated light signals. Their ratio is more consistent with a collision involving an atomic nucleus than one involving an electron. The LZ collaboration published the event on September 1, 2026. What caused it remains unknown.

One possible explanation would be a dark matter particle colliding with a xenon nucleus. But an unusual signal is not enough to support that conclusion: Known particles or an inadequately characterized detector effect could also play a role.

After accounting for the many models tested, the anomaly reaches a statistical significance of 2.6 sigma. This puts it below the thresholds commonly used in particle physics for evidence or a discovery.

According to the Berkeley Lab News Center, Richard Gaitskell, a professor at Brown University and spokesperson for LUX-ZEPLIN, explained that the collaboration was not claiming to have seen dark matter. It had, however, observed something interesting and wanted feedback from the scientific community.

The new LZ analysis and the three theoretical interpretations discussed here are available as preprints, meaning scientific papers released before review by independent peers. The material examined does not include assessments by independent experts outside the research groups involved.


The event at a glance

Measurement: June 16, 2023, approximately 1.5 kilometers underground at the Sanford Underground Research Facility in South Dakota.

Energy: 248 kiloelectronvolts (keV), provided the event is interpreted as a nuclear recoil from elastic scattering. The statistical and systematic uncertainties are each ± 23 keV.

Anomaly: The ratio of the two light signals is much closer to the expected distribution for nuclear recoils than to that for electron recoils.

Statistics: A local significance of 3.4 sigma and a global significance of 2.6 sigma after accounting for the models tested.

Open question: So far, no convincing explanation involving known background processes has been found. This is not yet evidence of dark matter.


How LZ analyzes two light signals

LUX-ZEPLIN is located in the former Homestake gold mine. The rock above the experiment reduces the flux of cosmic particles. Additional detection systems help identify background events.

The main detector contains approximately seven metric tons of liquid xenon, a heavy noble gas. For the new analysis, the researchers used an inner measurement region containing 4.71 ± 0.08 metric tons of xenon. Its distance from the walls reduces the influence of radioactive emissions from surrounding materials.

Among the targets are WIMPs: hypothetical weakly interacting massive particles considered candidates for dark matter. Such a particle could transfer energy to a xenon nucleus in a collision.

An energy deposit in the xenon produces two measurable signals. First, a brief flash of light is generated, called S1. Electrons are also released from the atoms. An electric field draws them upward into a gas layer, where they produce a second flash, S2.

The time between the signals reveals the depth at which the event occurred. The distribution of the second flash across the upper sensors provides the other spatial coordinates. The ratio of S1 to S2 also helps indicate which type of collision the event resembles.

In an electron recoil, energy is transferred to an electron; in a nuclear recoil, it is transferred to a xenon nucleus. At the same S1 strength, S2 is generally smaller for nuclear recoils. However, the corresponding measurement distributions overlap. An electron recoil can therefore appear in the nuclear recoil region. Neutrons can also strike xenon nuclei.

The detector thus provides clues about the type of collision. A nuclear recoil alone does not identify a dark matter particle.


What makes the 248-keV event unusual

For the new analysis, LZ extended the energy range examined for nuclear recoils: The upper limit increased from about 55 to 270 keV. This brought events outside the previous main analysis into consideration.

The data cover 220 live days between March 27, 2023, and April 1, 2024. After all selection criteria were applied, 1,710 events remained. The model combining background and a possible signal, fitted to the data, yielded 1,713 ± 39 events. There was no appreciable excess in the total count.

What stands out instead is the position of a single event in the plot of the two light signals. Interpreting it as a nuclear recoil from elastic scattering gives an energy of 248 keV. In this type of scattering, the particles involved exchange kinetic energy without changing their internal states.

The energy alone does not reveal the mass of the incoming particle. Its speed and the scattering angle also play a role.

The key feature here is the comparatively weak second light signal. After corrections for position-dependent effects, S2 at the measured S1 value lay 6.7 standard deviations below the center of the electron recoil distribution. By contrast, it was only 1.5 standard deviations from the center of the nuclear recoil distribution. A standard deviation describes how widely measurements typically scatter around their mean. The signal therefore resembles a nuclear recoil much more closely.

For the narrow S1 region containing the event, the model of known background processes predicts 0.0106 ± 0.0008 events. This corresponds, on average, to roughly one event in a hundred comparable datasets. The number describes a particular region of the model. It is neither the probability of dark matter nor the statistical assessment of the entire search.


Why the anomaly has a global significance of 2.6 sigma

The statistical question is: How often would the assumed background alone produce a deviation at least as large?

LZ tested 616 models of interactions between xenon nuclei and hypothetical particles of different masses. The largest deviation from the background-only hypothesis initially reached a local significance of 3.4 sigma.

Examining many possibilities, however, makes it easier to find an anomaly by chance. The researchers accounted for this so-called look-elsewhere effect in a further step. They grouped models with nearly identical expected event distributions, leaving 293 distinguishable cases. They then repeatedly simulated datasets containing only background and ran the entire search on each one.

In about five out of a thousand simulated experiments, a deviation at least as large as the one in the real data occurred. This corresponds to a p-value of around 0.5 percent, or a global significance of 2.6 sigma.

The p-value does not indicate how likely dark matter is to be the cause. It describes how unusual the result would be under the background assumption used.

In particle physics, 3 sigma is commonly treated as the threshold for statistical evidence and 5 sigma as the threshold for a discovery. The LZ anomaly reaches neither threshold. The 6.7 standard deviations from the center of the electron recoil distribution are not a discovery significance either: They describe the position of a measurement within a particular distribution, rather than an assessment of all possible backgrounds and the models tested.


What other causes could be involved


When electrons are lost along the way

One possible explanation is charge loss. A gamma-ray photon, a high-energy particle of light, can deposit energy at several neighboring points in the xenon. The nearly simultaneous light from these interactions can merge into a single S1 pulse.

If some of the released electrons originate near the wall or below the cathode, they may not reach the gas phase. The second light signal then becomes weaker. An event involving electron recoils can consequently resemble a nuclear recoil.

To estimate how often such misclassifications occur, the researchers need to know, among other things, the affected detector regions and the frequency of multiple interactions. A process that is rare overall can be important when assessing an individual candidate.

Neutrons and the events preceding the signal

Neutrons can also cause nuclear recoils. To transfer about 250 keV to a xenon nucleus, a neutron would need an energy of at least approximately 8 megaelectronvolts.

In the model examined, a neutron flux capable of explaining the June event would imply additional recoils at lower energies. No corresponding group of events is present in the LZ data.

The collaboration also investigated what had happened before the event. A cobalt-57 calibration source on the opposite side of the detector had been removed 25 minutes earlier. The outer detector had recorded the last preceding muon 41 minutes earlier; in the time projection chamber, the last preceding muon detection had been 127 minutes earlier. Muons are penetrating particles from cosmic radiation that can also produce neutrons.

So far, these checks have not provided a convincing explanation involving known background processes. Uncertainties in their description nevertheless remain relevant.


Why the analysis is considered non-blind

LZ uses a procedure called salting. Artificial events resembling signals are mixed into the data. The analysts initially do not know which events are artificial. This is intended to prevent selection criteria from being tailored to an apparent finding.

In the new high-energy range, however, these artificial events did not adequately cover the nuclear recoil region. The collaboration therefore explicitly describes the analysis as non-blind.

The researchers adopted most of the selection criteria from their earlier study. The selection criteria and probability models were fixed before the remaining artificial events were removed. Nevertheless, this limitation of the blinding procedure remains part of the assessment of the result.


Which particle models might fit

An explanation involving dark matter must do more than allow a collision at 248 keV. It must also be consistent with the rest of the measurements.

In many simple WIMP models, low-energy recoils are more common than high-energy ones. An interpretation of the unusual event would therefore often imply additional events at lower energies. Yet the earlier LZ search found no statistically significant excess there. The 220 live days in the new analysis were already part of that earlier study.

One possibility is inelastic scattering. In the models discussed here, the dark matter particle transitions into a heavier state during the collision. This requires additional energy. With suitable model parameters, low-energy recoils are therefore strongly suppressed.

An event near 248 keV could thus be consistent with the absence of an excess at lower energies. Whether a specific explanation works, however, depends on its other predictions.


A Higgsino as a possible candidate

Katherine Freese and Dionysios Theodosopoulos proposed a Higgsino interpretation on September 1, 2026. The Higgsino is a hypothetical particle from supersymmetry, a theoretical extension of known particle physics. It is the supersymmetric partner of the Higgs fields.

In their model, its mass is approximately 1 TeV/c², roughly the mass of a thousand protons. The mass difference between two neutral states is on the order of 350 keV/c².

The interaction proceeds through a Z boson, a neutral mediator of the weak interaction. Once the model parameters are fixed, the expected collision rate cannot be adjusted arbitrarily to match the single event. This makes the explanation testable: It must also correctly describe the number and energy distribution of further events.


The test at higher energies

On September 3, 2026, Nicholas Rodd, Benjamin Safdi, Tracy Slatyer and Weishuang Linda Xu examined a related scenario involving a Higgsino with a mass of approximately 1.1 TeV/c².

Their calculation connects the present abundance of these particles with the evolution of the early universe. Higgsinos could have formed in the hot plasma at that time and annihilated one another. As the universe expanded and cooled, these processes became less frequent. For the chosen parameters, the remaining particle density could match the present density of dark matter.

A difficulty, however, arises from the LZ data themselves: The model also predicts events at still higher energies that were not observed there. The absence of such signals can constrain an interpretation.


A possible seasonal variation

In another preprint dated September 3, 2026, theoretical physicist Christopher McCabe examined how the rate of inelastic scattering might change over the course of the year.

As Earth orbits the Sun, its speed changes relative to the assumed halo of the Milky Way, an extended envelope of dark matter surrounding the galaxy. If a collision requires a high minimum energy, even a small change in speed can affect how many particles reach that threshold. In some model variants, no corresponding events would be expected during part of the year.

In some versions of the calculation, June 16 falls within a period of an increased expected event rate. A single event, however, does not establish a seasonal variation. That would require further events whose dates and energies match the prediction.


Why researchers search for dark matter in the first place

The LZ event is not the starting point of the hypothesis. For decades, astronomers have investigated observations in which visible matter alone does not explain the measured gravitational effects.

In 1933, Fritz Zwicky inferred from the velocities of galaxies in the Coma cluster that more gravitating mass had to be present than its luminosity suggested. Later measurements of individual galaxies revealed a similar problem: In their outer regions, stars and gas move faster than would be expected from the observed matter alone. These include the studies of Andromeda by Vera Rubin and Kent Ford and of M33 by Edvige Corbelli and Paolo Salucci.

Further clues come from the cosmic microwave background, radiation left over from the early universe, and colliding galaxy clusters. In the Bullet Cluster, the mass concentrations reconstructed using gravitational lensing lie near the galaxies and are spatially separated from the hot gas. In gravitational lensing, the gravity of large masses distorts the images of more distant background galaxies. The hot gas contains most of the ordinary matter in these two galaxy clusters.

Within the standard cosmological model, analysis of the Planck data indicates that ordinary matter accounts for approximately 5 percent and cold dark matter for approximately 26.5 percent of today's total energy density. Here, “cold” means that the particles moved much more slowly than light long before galaxies formed. Dark matter therefore accounts for around 84 percent of all matter in the universe. Dark energy is not included in this proportion. These values are not a direct “weighing” of the universe. They are derived from observations within the framework of the standard cosmological model, ΛCDM.

The abundance of primordial deuterium, a heavy form of hydrogen, also constrains the amount of ordinary matter. Within the corresponding calculations, it is insufficient to explain all the missing mass through faint stars, planets or cold gas.

Alternatively, researchers investigate modified laws of motion, such as Modified Newtonian Dynamics, or MOND. These approaches must likewise explain different astronomical observations together. The astronomical findings and laboratory searches address different questions: The former examine additional gravitational effects, while the latter seek the particles that could explain them. Direct detection of such particles remains outstanding.


What further measurements could clarify

LZ continues to collect data and plans a total of 1,000 live days. The crucial questions will be whether further events appear and what pattern they form.

The explanations discussed can be tested in different ways:

Energy distribution: If a model predicts additional events at lower or higher energies, those predictions must be consistent with the data.

Timing: A seasonally varying rate would have to become apparent across multiple events.

Detector behavior: A better understanding of charge loss and other background processes could change the assessment of the candidate.

The absence of signals also helps constrain models. Further measurements could help clarify the origin of the event and test the proposed explanations.

Until then, the finding remains limited: LZ has recorded an unusual event. Whether dark matter, a background process or something else lies behind it remains an open question.


Sources:

Igor Danilov (2026). Did the LZ Experiment Find Dark Matter and Why Is the Answer Far More Complicated Than a Simple “Yes” or “No”? ALLATRA Media, September 17, 2026. https://allatra.media/science/lz-experiment-dark-matter-248-kev-event

LZ Collaboration (D. S. Akerib et al.). Search for dark matter particle interactions in an extended nuclear recoil energy window with the LUX-ZEPLIN (LZ) experiment. arXiv:2609.02823 (2026). Preprint. https://doi.org/10.48550/arXiv.2609.02823

LUX-ZEPLIN. The LZ Dark Matter Experiment. Official website of the experiment. https://lz.lbl.gov/

Berkeley Lab News Center. LZ Sees Surprising Result in Search for Dark Matter. September 1, 2026. newscenter.lbl(.)gov/2026/09/01/lz-sees-surprising-result-in-search-for-dark-matter/

F. Zwicky (1933). Die Rotverschiebung von extragalaktischen Nebeln [The Redshift of Extragalactic Nebulae]. Helvetica Physica Acta 6, pages 110 to 127. English and Spanish translation by H. Andernach (2017), arXiv:1711.01693. doi(.)org/10.48550/arXiv.1711.01693

E. Corbelli, P. Salucci (2000). The extended rotation curve and the dark matter halo of M33. Monthly Notices of the Royal Astronomical Society 311, pages 441 to 447. doi(.)org/10.1046/j.1365-8711.2000.03075.x

V. C. Rubin, W. K. Ford Jr. (1970). Rotation of the Andromeda Nebula from a Spectroscopic Survey of Emission Regions. The Astrophysical Journal 159, pages 379 to 403. doi(.)org/10.1086/150317

Planck Collaboration (2020). Planck 2018 results. VI. Cosmological parameters. Astronomy & Astrophysics 641, A6. doi(.)org/10.1051/0004-6361/201833910

R. J. Cooke, M. Pettini, C. C. Steidel (2018). One Percent Determination of the Primordial Deuterium Abundance. The Astrophysical Journal 855, 102. doi(.)org/10.3847/1538-4357/aaab53

D. Clowe, M. Bradač, A. H. Gonzalez, M. Markevitch, S. W. Randall, C. Jones, D. Zaritsky (2006). A Direct Empirical Proof of the Existence of Dark Matter. The Astrophysical Journal Letters 648, pages L109 to L113. doi(.)org/10.1086/508162

D. S. Akerib et al. (LZ Collaboration) (2020). The LUX-ZEPLIN (LZ) Experiment. Nuclear Instruments and Methods in Physics Research Section A 953, 163047. doi(.)org/10.1016/j.nima.2019.163047

J. Aalbers et al. (LZ Collaboration) (2025). Dark Matter Search Results from 4.2 Tonne-Years of Exposure of the LUX-ZEPLIN (LZ) Experiment. Physical Review Letters 135, 011802. doi(.)org/10.1103/4dyc-z8zf

K. Freese, D. P. Theodosopoulos (2026). Higgsino Dark Matter Interpretation of the LUX-ZEPLIN 248 keV Nuclear-Recoil Event. arXiv:2609.01583. Preprint. doi(.)org/10.48550/arXiv.2609.01583

N. L. Rodd, B. R. Safdi, T. R. Slatyer, W. L. Xu (2026). Confronting the Higgsino Interpretation of the LZ Event with the High-Energy Sideband. arXiv:2609.04175. Preprint. doi(.)org/10.48550/arXiv.2609.04175

C. McCabe (2026). Seasonal dark matter from the LUX-ZEPLIN high-energy event. arXiv:2609.04181. Preprint. doi(.)org/10.48550/arXiv.2609.04181