Here is the contradiction most people miss about the most exciting dark matter result in years: the physicists who found it believe their detector completely, and they have almost nothing to believe in.
One event. A single particle interaction, recorded a mile beneath South Dakota on a Friday in June 2023, that a collaboration of 250 scientists spent months trying to explain away and could not.
"This is the first example in any experiment I've worked on of an outlier that appears valid in every way," says Aaron Manalaysay of Berkeley Lab. Then, immediately, the hedge: "Of course, we're still twisting our brains trying to think if there's a rare background mechanism we could've missed."
That is the whole story in one breath. Total confidence in the machine, total emptiness in the sample size.
The event is called LZ230616. It arrived inside the LUX-ZEPLIN detector, ten tons of ultrapure liquid xenon watching for the faint double-flash a dark matter particle would leave as it knocks a xenon nucleus off its perch.
The numbers are almost absurdly clean. The event landed 26.4 centimeters from the detector's cathode, well inside the quietest part of the tank, with a primary flash of 540.1 photons detected. In the specific energy region where it appeared, LZ's background model predicted just 0.0106 events, give or take 0.0008.
That is a hundredth of an expected event, and one showed up.
Statistically, that is a 3.4-sigma bump in isolation. But the collaboration had tested 616 different interaction models across a wide energy window, and once you account for all the places a fluke could have hidden, it drops to 2.6 sigma — roughly a one-in-200 chance of being a background impostor.
Physics does not call anything a discovery below five sigma. So this is not a discovery. It is a rumor with an excellent alibi.
Physics does not call anything a discovery below five sigma. So this is not a discovery. It is a rumor with an excellent alibi.
Now here is why the rumor is interesting for being wrong.
For forty years the prime suspect has been the WIMP: a heavy, sluggish particle that occasionally taps a nucleus and sends it recoiling with low energy. That is the vanilla model, and it makes a hard prediction.
Low-energy taps should vastly outnumber high-energy ones. If dark matter interacted with xenon in the simplest way, LZ should have buried itself in thousands of gentle, low-energy hits long before producing one violent 248-keV recoil.
It saw none of them. Last year's low-energy search came up empty. This year's high-energy search found exactly one event, sitting alone at the top of the plot.
As LZ spokesperson Rick Gaitskell puts it, a simple interaction would have produced "this much larger number of lower energy events" — and it didn't. So the simplest possibility is the one thing the data rules out.
That forces nature into stranger shapes. To suppress the low-energy flood and permit a lone high-energy strike, dark matter has to be structured or picky.
Maybe it has internal energy states, like an atom, and only reacts when hit hard enough to jump a level — so-called inelastic scattering. Maybe its grip on ordinary matter strengthens with momentum. Either way, the particle stops being one new thing and starts being a small, complicated regime.
And there is a seductive detail tucked in the timing. Some inelastic models predict dark matter collisions peak in summer, when Earth plows head-on into the galaxy's dark matter wind. The event landed in June.
Which is exactly where caution has to reassert itself, because one event underdetermines almost everything.
Look at the theory papers that stampeded in within days. One argues the event is a thermal Higgsino, its mass pinned near 1.1 TeV by the relic abundance, with a 371-keV internal splitting tuned to land the recoil in the right spot.
Another discards WIMPs entirely and proposes a fermionic dark matter particle of just 247 MeV being absorbed by a xenon nucleus, dumping its entire rest mass into a single kick. That paper's own authors then note the idea runs into tension with existing KamLAND data.
So a lone bump has already been read as everything from a sub-GeV absorption particle to a TeV-scale supersymmetric relic. When one data point can support interpretations separated by four orders of magnitude in mass, the data point is not telling you what dark matter is. It is telling you the field is desperate and creative.
We have watched this movie before. In 2020, XENON1T saw a tantalizing excess; its successor erased it. The graveyard of dark matter hints is well populated.
And the field keeps learning how easily a "signal" dissolves. A recent simulation study of stellar streams — those ribbons of stars torn from infalling dwarf galaxies — found that the kinks and twists astronomers had hoped were fingerprints of dark matter clumps can be produced by ordinary matter alone. Smooth streams turned out to be the rare ones. A whole class of indirect evidence quietly lost its footing.
So how do you confirm a ghost you can't reproduce on demand? Not, crucially, with more xenon.
Daniel Akerib of SLAC, an LZ member, is blunt about the recipe. You want to see the result confirmed, "learn its coupling to matter by seeing it in another isotope — possibly liquid argon," and ultimately produce the thing in a lab. Seeing something go bump, he notes, is very different from measuring its cosmic abundance.
The logic is unforgiving. Xenon can pile up more data and either grow the significance or watch it fade — and LZ has already collected at least three times the exposure it just analyzed. But xenon alone can never prove that a xenon signal is dark matter rather than some xenon-specific quirk. Only a different target, a different detector, a different failure mode, can cross-check the claim.
Which is why a machine booting up in Ontario this month belongs in the same story as a Friday in South Dakota.
SuperCDMS SNOLAB began collecting its first scientific data this summer, more than a mile underground in the Vale Creighton mine near Sudbury. Instead of a xenon tank, it uses 24 ultrapure silicon and germanium crystals, each about the size of a hockey puck.
Chilled colder than outer space, each crystal turns a particle strike into a tiny phonon — a quiver in the crystal lattice — plus a faint electrical pulse, read out by superconducting sensors that only work near absolute zero.
The shielding is its own kind of poetry. Copper, polyethylene, a radon barrier, a mile of rock — and ultrapure lead salvaged from ancient Roman shipwrecks on the floor of the Mediterranean, because modern lead is too radioactive and only two thousand years underwater burns the radioactivity out.
Roman ballast now stands guard over the lightest particles physics can imagine.
Because that is SuperCDMS's beat: not LZ's heavy quarry but the light-mass frontier, the featherweight particles whose interactions leave only the faintest trace. Its spokesperson, Priscilla Cushman, says the detectors will probe "regions where the lightest-mass dark matter particles may be lurking."
The two experiments hunt from opposite ends of the mass scale, on different materials, with different backgrounds and different ways of being fooled. That is not redundancy. That is the entire point.
If dark matter is real and structured, it should eventually leave a consistent signature that no single detector's quirks can fake. If LZ230616 is a fluke or a xenon artifact, a germanium crystal will never confirm it. The disagreement is the safeguard.
Full-sensitivity science at SuperCDMS isn't scheduled until 2027, and even then it targets a different regime than LZ's 248-keV oddity. Confirmation of this specific event will more likely come from argon or from more xenon. But the deeper shift is architectural.
For the first time, the dark matter hunt is not one heroic machine hoping for a lucky night. It is a field with enough different eyes, in different mines, made of different stuff, to interrogate a single flash and ask whether anything else in the world can see it too.
The frontier isn't a bump on a Friday. It's the growing ability to check.