370 | Daniel Akerib on a Faint Hint of Dark Matter
What this covers
Sean Carroll talks with Daniel Akerib, a SLAC physicist who has searched for dark matter since the early 1990s, about a single unexplained event in the LUX-ZEPLIN xenon detector. The conversation begins with history: why particle physicists, drawn by supersymmetry and naturalness arguments, expected a WIMP, and how early searches repurposed germanium detectors built for double beta decay. It then covers the mechanics of direct detection: working deep underground, discarding events near the tank walls, using the outer tons of xenon as an active shield, and calibrating with neutron sources that scatter off xenon nuclei, which maps the nuclear recoil band.
The event itself is the centre of the middle stretch. In the 2023-2024 data, a high-energy nuclear recoil search found one event sitting on that band, far from electron-recoil backgrounds, and the team could not explain it with any known background, including accidental two-event coincidences. Akerib explains the blinding and salting used to guard against analysis bias, and why Geiger-counter randomness makes one event or a quiet stretch weak evidence. At 2.8 sigma global significance, he says, chance fluctuations of this size are routine. If it is real, the best fit is a heavy particle near 1 TeV with large error bars, and he lays out what more data could show: off-band events, on-band events, or nothing further.
The last part widens the frame. After seven or eight orders of magnitude of null WIMP results, the field diversified into dark-sector candidates spanning about 30 decades of mass. Akerib compares the situation to Ray Davis's solar neutrino shortfall. He also discusses the engineering demands of larger detectors, and why confirmation would need other xenon experiments, colliders, and decades of work.
Akerib argues that the LZ event is a low-significance, unexplained single candidate that cannot be called dark matter yet, and that only repeated events, controlled backgrounds, and independent experiments can separate a real WIMP signal from fluctuation or an unmodeled background.
- A single event at 2.8 sigma global significance is below the evidence threshold, and such fluctuations are common across the field.
- The team used blinding and salting, and the event survived extensive background checks but has no identified background explanation.
- The field has broadened from WIMPs to many dark-sector candidates, so no single detection can settle the question.
A single 2.8-sigma event is provisional, and confirming dark matter needs more events and cross-checks.
- The event's global statistical significance is 2.8 sigma, below the conventional thresholds for hint, evidence, or discovery, and because roughly one in 100 experiments produce a three-sigma fluctuation, particle physicists routinely see such results from chance.
“the global significance of this detection was 2.8 sigma, which is below the threshold for hint, evidence, discovery, et cetera... Three sigma, four sigma, five sigma.”
- Additional data could show events that are off the nuclear recoil band, indicating an unmodeled background to be understood, events on the band indicating a possible signal, or no further events, leaving the event an unexplained mystery.
“So that's one possibility is that we see more events, but they're not in the right place. Stand down, chill out, right.”
- The community conventionally requires three, four, or five sigma levels before calling a result evidence or discovery, because with many experiments running, expected fluctuations would otherwise generate false claims.
“Three sigma, four sigma, five sigma. How much should we pay attention to this?”
- The observed rate corresponds to roughly one event per four ton-years of exposure, so the team needs a bigger detector to accumulate enough events to resolve the energy spectrum, and a 2023 event sits within 220 days of data.
“If one event per 200, whatever, four ton-years of exposure is the rate.”
- A single event or single experiment should not be trusted as proof of dark matter, and everything in science remains provisional and falsifiable, so the result could eventually be dismissed if not reproduced.
“Don't trust a single experiment. Right. Right. And... Don't trust a single event. Yeah. And everything in science is provisional.”
- Many apparent signals have faded when subjected to more scrutiny, so the team should treat its sub-three-sigma result as potentially a fluctuation until more data arrives.
“many things have fluctuated away when you subject them to more scrutiny.”
- The proposed next-generation xenon experiment was prioritized through peer review by cost and breadth of program, but it was recommended under budget scenarios that have not yet been realized, so its construction depends on funding.
“And this one survived that process and was recommended to go forward, but it did so in budget scenarios that have not yet been realized.”
- Given one event in 220 days, the team computed a 10 to 90 percent confidence envelope for expected event counts as exposure grows, and it widens very quickly, so a prediction of either nothing or a detection after threefold more exposure is both reasonable.
“There's a reasonable chance that we see nothing in three times the stats. There's a reasonable chance that we see something in three times the stats.”
- A single event is insufficient to claim dark matter, because individual events fluctuate and many apparent signals have faded under scrutiny, so confirmation requires more events and eventually a spectrum that can be studied in detail, with one event turning into three, then twenty.
“Don't trust a single experiment. Right. Right. And... Don't trust a single event. Yeah. And everything in science is provisional.”
- Confirmation should come from other experiments like XENONnT and PandaX searching their own data, plus accelerator searches for related particles, because a collider finding supersymmetric cousins would show how the particle couples to ordinary matter, while direct detection alone cannot identify the model.
“Supersymmetry is something that will be discovered in an accelerator. Right? And they will work out the couplings”
- The team has analyzed only the 2023-2024 data, which is about one-third of total recorded exposure, while roughly three times that amount remains unexamined, so confirmation depends on future analysis of the existing dataset as well as new data.
“We have three times as much data recorded that we haven't looked at yet”
- A 1 TeV particle candidate, if real, could be produced at colliders such as the LHC, and the detected event would motivate bigger accelerators, making accelerator searches a key test of the signal.
“if the mass is a TeV, which is our best fit with huge error bars, if it's even something, that should be accessible in colliders, right?”
- If the event is real, it would take a rise from one event to about three and then about twenty events before the energy spectrum could be populated and studied, and at the expected rate of roughly one event per 200 units of exposure, hundreds of events would require a much larger experiment.
“one event turns into three, turns into 20, then we get to start populating that energy spectrum, which contains information.”
The case for xenon dark matter searches stands regardless of this single event.
- If the event is confirmed, Akerib would feel elation and validation, while if it disappears the motivation for the xenon program still stands, so the scientific case for the next experiment does not depend on this single event.
“If it goes away, if it's, we see four more events and none of them are on the nuclear recoil band, then the pitch that we'd been making is still there”
- Confirming a real dark matter particle requires connecting direct detection with collider discoveries, indirect detection in galaxy centers, and cosmological models of relic density, which together would take the next several decades.
“if the mass is a TeV, which is our best fit with huge error bars, if it's even something, that should be accessible in colliders, right? And so you wanna study it in detail in the lab, you wanna study it in detail its astrophysical signature”
- By roughly the 1980s or 1990s it was understood that ordinary matter (protons, neutrons, anything made in an accelerator) cannot make up the missing gravitational mass, which led particle physicists to hypothesize a new form of matter.
“ordinary stuff couldn't make up the difference, that ordinary matter, protons, neutrons, the periodic table, anything that's been made in an accelerator, can't account for this missing stuff.”
- Failure to detect WIMPs over seven or eight orders of magnitude in the search space has pushed WIMP searches toward regimes where naturalness arguments would predict a detection, but this has led the community to diversify its searches rather than abandon dark matter candidates.
“we've looked through, I don't know, seven or eight orders of magnitude on the WIMP search side. We are probably now pushed to regimes where the cognoscenti would say, 'Well, if it was natural, it would have kind of showed up by now.'”
Apparent signals must be scrutinized against statistical flukes and alternative explanations.
- Random processes like a Geiger counter produce clusters and streaks, so people systematically underestimate runs of events, and a single event or a quiet stretch carries little information about the underlying rate.
“When you ask people to make up a random distribution of coin flips, they invariably don't make enough in a row of heads or of tails, right?”
- After a single anomalous event, theorists quickly publish papers matching the event to their favored models, so the event could be explained by up to a couple hundred dark matter possibilities.
“If you would like to read about a couple of 100 of the possibilities, you could go online and look at some of the recent papers. I think it's fair to say it could still be just about anything.”
- A rare-event search must consider whether two overlapping events, one at the detector edge that lost light or charge, could accidentally mimic a nuclear recoil, so every such conspiracy scenario must be checked.
“are there conspiracies where two events could arrive in the detector overlapping, and maybe one of them was on the edge of the detector and lost some charge or lost some light”
- In the 2023-2024 data, the high-energy nuclear recoil search found one event lying exactly on the calibrated nuclear recoil band and far from electron-recoil backgrounds, which the team could not explain with any known background after considering accidental two-event coincidences.
“So in the data set that we'd accumulated in 2023 and 2024, the team that was searching for these high-energy nuclear recoils found an event. And it lied exactly on this nuclear recoil band”
Direct detection separates WIMP recoils from electron backgrounds using spectra, charge-light ratios, and calibration.
- Theorists used non-relativistic effective field theory to enumerate the most general ways WIMPs could couple to xenon nuclei, including velocity-dependent and spin-dependent couplings, which produce recoil spectra extending to hundreds of keV rather than only the 10 keV scale.
“if we can include velocity dependence and spin and what have you, that you might get events that go out to significantly larger energies.”
- Electron recoils from gamma rays produce a higher ratio of charge to light, while nuclear recoils from WIMP-like scattering produce a lower ratio, so this ratio is the characteristic signature used to separate signal from the dominant electron-recoil background.
“Electron recoils, gamma ray comes in and scatters, it produces a certain ratio of charge to light. Nuclear recoils produce a lower ratio of charge to light.”
- If WIMPs scatter like billiard balls, the energy spectrum of recoils falls exponentially, so the logarithm of event counts per energy bin is a straight line decreasing with energy.
“even simple calculations that I can follow, show you that the energy spectrum of WIMPs would look like a falling exponential.”
- Direct detection searches for WIMPs scattering directly off a nucleus in a detector, while complementary methods look for WIMPs collecting in stellar or galactic centers and annihilating into gamma rays, and for supersymmetric particles at colliders.
“So there are what we do, we call direct detection. This is a WIMP directly scattering from a nucleus in our detector and registering a signature.”
- WIMP experiments remain theory-agnostic in principle, but they report results by excluding parameter space of specific models such as supersymmetric ones, which allows comparison with collider searches and indirect detection.
“when we run an experiment and we exclude some parameter space, we can say, some particular parameter space of supersymmetric models appears not to explain the dark matter.”
- Because WIMPs cannot be beamed, the team calibrates detector response by bringing neutron sources near the detector, since neutrons are neutral and do not produce electron recoils but do billiard-ball scatter off xenon nuclei, mapping the region where nuclear recoils should appear.
“since neutrons are neutral particles, they won't create these electron recoils, which are a background, but they will scatter, billiard-ball scatter, from Xenon nuclei.”
Sensitive underground xenon detectors with strong background suppression make rare-event searches possible.
- LUX and ZEPLIN merged into LZ, while the XENON and PandaX collaborations compete, and each has at times held the most sensitive exposure, so a new data run can offer discovery potential if WIMPs are near.
“And those three collaborations have sort of been, over time, been leapfrogging each other.”
- Liquid xenon detectors such as XENON10 and ZEPLIN-III became more sensitive than germanium cryogenic detectors around 2007 because they could also be built with very low natural radioactivity and performed well at very low energies.
“And basically in 2007, XENON10 had a better... More sensitive search for dark matter than these germanium experiments.”
- Rare-event experiments are placed deep underground because cosmic rays create overwhelming high-energy background at the surface, just as astronomy is done at night because scattered sunlight swamps faint stars during the day.
“It would be like trying to do astronomy during the day. Yes, the stars are on during the day, but there's an enormous amount of scattered sunlight. So let's do astronomy at night.”
- Early dark matter searches were repurposed germanium detectors built for neutrinoless double beta decay, reconfigured with ultra-low thresholds from MeV scale down to roughly 10 keV because WIMP interactions were expected to be rare, like double beta decay.
“so a handful of double beta decay experiments became the first dark matter experiments.”
- Annual world xenon production is about 60 to 100 tons, the experiment's seven-ton inventory is about a tenth of that, xenon is distilled from liquefied air at about one in 10 million atmospheric atoms, and it is not consumed, so the team tries never to release it.
“So the annual production of Xenon worldwide is somewhere between 60 and 100 tons. So we're using a tenth of world production.”
- A detector ten times larger run for five times as long must reduce backgrounds by a factor of about 150, so beating down backgrounds is the central engineering challenge of next-generation xenon experiments.
“if you're gonna build a 10 times bigger thing and you're gonna run it for five times as long, you have to push your backgrounds down by a factor of 150.”
- The LZ detector holds about seven tons of liquid xenon cooled to roughly minus 100 Celsius; a particle interaction produces a prompt scintillation light flash and ionized electrons that are drifted to the gas phase to produce a second light signal, and both are recorded by photomultiplier tubes to reconstruct each event.
“when particles interact with it, it produces small flashes of light, and some of the Xenon gets ionized. So electrons are freed up.”
- Events near the edges of the xenon tank are discarded because radioactivity from the walls is likely to cause them, so the search focuses on the inner four or five tons, with the outer one or two tons serving as an active shield.
“let's only focus our dark matter search on the inner four or five tons, and we'll use that outer ton or two as a shield, as an active shield.”
- The detector sits about a mile underground in a former gold mine surrounded by 8 meters of purified water, because cosmic rays create a high background at the surface, and the rock overhead shields the rare-event search much as night darkness helps astronomy.
“It would be like trying to do astronomy during the day. Yes, the stars are on during the day, but there's an enormous amount of scattered sunlight. So let's do astronomy at night.”
The dark matter identity, mass, and interaction remain open pending additional events and collider evidence.
- If the event is real, the best-fit mass is about one TeV with large error bars, and the four ton-year exposure needed to produce a single event implies a heavy particle with weak interactions, though the coupling and mass remain poorly constrained.
“if the mass is a TeV, which is our best fit with huge error bars, if it's even something, that should be accessible in colliders”
- Confirming a dark matter particle could follow the model of Big Bang nucleosynthesis, where lab measurements, observations, and cosmological models together predict light-element ratios, so the particle's relic density could in principle be calculated from its known physics.
“We can tie that whole story together and literally calculate the relic density based on the physics of this species.”
- One event cannot determine the particle's identity, and only additional events and collider discoveries of cousin particles could say whether the candidate is supersymmetric, so the dark matter identity remains open.
“Supersymmetry is something that will be discovered in an accelerator. Right? And they will work out the couplings”
Galactic motion anomalies and past neutrino discovery favor new matter over changing gravity.
- Ray Davis's chlorine experiment detected only about one-third of the solar neutrinos predicted, a discrepancy later resolved when the SNO experiment found the missing neutrinos had changed flavor, showing neutrinos have flavor and oscillate on their journey from the sun.
“He only detected one-third the amount that was predicted from the standard solar model.”
- Akerib, as a particle physicist, argues the community should not discard the laws of gravity and instead assumes a new form of matter, following a historical pattern in which conservation-law anomalies led to new particle discoveries such as the neutrino.
“Let's not be so cavalier to throw away the laws of gravity. Let's see if we can find the missing stuff.”
- Stars and gas in the Milky Way move too fast given the visible mass present, so gravitational effects are seen that ordinary matter cannot account for, a mystery dating back about 100 years.
“Apply that same physics to the galaxy as a whole, the Milky Way, and things seem to be moving much too fast given the amount of stuff that is present in stars.”
After null WIMP searches, the field broadened from supersymmetry toward axions and a wider dark sector.
- In the 1990s supersymmetry seemed a natural explanation whose dark matter would sit at a natural mass and scattering scale, and after searching seven or eight orders of magnitude without success, the field began asking whether nature chose something else.
“And it was thought to be a very natural scale at which to produce dark matter. And we've looked through, I don't know, seven or eight orders of magnitude on the WIMP search side.”
- The strong CP problem, a fundamental symmetry that is apparently not violated though it could be, was resolved by Peccei-Quinn symmetry, which predicts the axion, a particle that would appear in detectors as a wave rather than billiard-ball scattering.
“And so the theory was come up with to solve the strong CP problem. And this was Helen Quinn and Roberto Peccei, and they proposed a new symmetry, and that new symmetry led to the prediction of another particle called the axion”
- About 10 to 15 years ago, after WIMP and axion searches had not found anything, the field broadened its search to any particle not forbidden by known physics that could produce dark matter, giving rise to the dark sector concept with candidates such as dark photons, spanning about 30 decades of mass.
“let's expand the portfolio and let's change the question a little bit and say, is there anything that isn't disallowed, did I get my double negatives right? Isn't disallowed by particle physics that could also produce dark matter?”
- Axion searches differ fundamentally from WIMP searches: WIMPs are sought as billiard-ball scatterings off nuclei, while axions would appear as wave-like signals, so experiments tuned for axions are like tuning a radio station.
“They would be wave-like. I'm sure we'll go into this more. We're looking for billiard ball scatters. They're looking to tune in the axion radio station.”
- Theories such as supersymmetry were proposed because the Standard Model's particle spectrum and mass scales look arbitrary and unnatural, and these theories predict new particles, and astrophysicists' search for missing mass seemed to point the same direction as particle physicists' search for a deeper theory.
“And the connection that was made, I don't know, 30-plus years ago was that, "Oh, astrophysicists and cosmologists are looking for this missing mass, and particle physicists are looking for these underlying explanations to the standard model, and they seem to be pointing in the same direction. We should go look."”
Blinding, salting, and null-result publication guard against analysis bias.
- The collaboration uses blinding to hide the signal region during analysis, and salting, in which a working group inserts fake signal-like events that analysts must find or reject before unblinding, to test whether the analysis is biased.
“So we can blind ourselves to the signal region, make all of the decisions that we're gonna do to determine whether the big reveal doesn't suffer from bias.”
- The team commits to publishing null results as well as positive ones, so the absence of a signal in future data would still be reported, which also prompts competing experiments like XENONnT and PandaX to check their own data.
“every new null result we publish. So certainly, like, while we'd identified this as a data set of interest, we saw something, we're certainly gonna publish that.”
- The salting procedure used for this analysis was not robust, and later refined nuclear-recoil calibration made the event look like it might belong to an older calibration, so the team is partly re-salting new data after improved calibration.
“Now, we're not super happy with the way that the salting was done for this result, so we weren't putting a lot of emphasis on it for this analysis.”
- Every null result is published, and if more data show nothing on the nuclear recoil band, the motivation for dark matter searches and the case for building the next xenon experiment remain unchanged.
“every new null result we publish. So certainly, like, while we'd identified this as a data set of interest, we saw something, we're certainly gonna publish that.”
Collaboration size, costs, and review pace shape how science gets done.
- Physicists favor cheap solutions, such as using an existing mine cavern rather than digging new space, because there are more good ideas than funding, which creates natural competition among proposals.
“There's never... There's just more good ideas to pursue in science than there is money to pursue them.”
- LZ and XENONnT merged into a single collaboration called XLZD, combining the best techniques from both teams and sharing costs, with a proposed detector that is designed to do both dark matter and neutrinoless double beta decay searches in xenon-136.
“So we've merged our teams to pitch and build this new experiment and draw from the broader pool of expertise.”
- The collaboration is debating whether to analyze the new data slowly and deliberately or quickly, because the analysis team is exhausted and the first paper draft drew 300 comments, showing that careful review takes significant time.
“the first paper draft had 300 comments. So this is a very deliberate process.”
- As collaborations grow, meeting time seems to scale with collaborator count, so researchers spend less time in the lab and more in meetings, which made Akerib prefer a small dark matter group of about 30 people over a 200-person experiment.
“it seems to be scale with... the number of collaborators. ... I seem to be spending less and less time in the lab and more and more time in meetings.”