Chaff

Separating the wheat from the chaff is an old enough figure of speech that most people who use it have never touched a threshing floor: toss the grain in the air and the heavy kernels fall straight back down while the light husks and stems — the chaff — blow off sideways. A physicist doing more or less the same thing to a spectrum of noisy readings calls the unwanted part background instead, but the question is identical: most of what a detector records is nothing, and the real work is deciding whether any one bump in it is grain or just chaff that happened to fall a little oddly this time.

On 1 September 2026 the LUX–ZEPLIN (LZ) collaboration — a dark-matter detector built around seven tonnes of liquid xenon, nearly a mile underground at the Sanford Underground Research Facility in South Dakota, precisely so cosmic rays can't reach it and fake a signal — told a physics conference in Tendo, Japan that they'd recorded one nuclear-recoil-like event, at about 248 keV, that their own background model can't account for. Their own estimate is that there's roughly a 0.5% chance it's ordinary background wearing a disguise. That sounds convincing until you learn the field's own bar for calling something a discovery: 5σ, odds of about one in 3.5 million against it being a fluke. What LZ actually has, in the units physicists use for exactly this judgment, is 2.6σ. Genuinely interesting. Openly, by the team's own account, not proof of anything yet — the detector keeps running through 2028 specifically to find out which it is.

Below is chaff, not grain: Math.random() standing in for a detector, with no signal hidden in it, ever, on purpose. Press the button and it draws 200 fake background readings from a plain bell curve, then reports the single highest one, in real standard deviations (σ) above the middle — the same units, the same question. Watch how often that number clears 2.6σ anyway, with nothing behind it but noise.

Honest gap: 200 isn't a real detector's live-day count or bin count — it's a round number chosen so a false alarm this size shows up often enough to actually watch happen, not so rarely you'd click for an hour first. Each dot's left-right position is its real σ value, drawn from an honest standard normal; how high it sits on the page is only so 200 of them don't smear into one line. The point isn't the count, it's the shape: any process that keeps drawing from pure noise will eventually throw a bump that looks exactly like a real one, at a rate you can calculate in advance — which is exactly why 5σ, not 2.6σ, is where physicists agree to start believing a bump is grain.

Sources: LUX-ZEPLIN collaboration, results reported 1–2 September 2026 at the TeV Particle Astrophysics conference, Tendo, Japan — not yet a peer-reviewed discovery paper, on purpose: this is a conference announcement, and the honest gap above is real. Reporting via Nature and ScienceDaily; background on the detector itself via Wikipedia's LZ experiment entry.