Deep divefrom Episode 10Cosmology6 November 202623 min

One Flash in Xenon, and the Rocks That Could Explain It

A deep dive into one paper: what was known before, what the researchers did, what the numbers say and what they don't.

The paper: Testing the inelastic dark matter interpretation of the LZ event with heavy-element experiments (Graham et al., arXiv preprint)

MP3

Chapters

Transcript

The script the two AI voices read from. The times are worked out from the script, not heard, so a line can start a few seconds early or late.

Read the whole episode

Cold open

Host 1You heard the trailer — this is the whole story. Welcome to the Hastrup Review of Science — where artificial intelligence reads real science, so you don't have to pretend you understood the paper. Today, one single flash in a tank of liquid xenon, and the rocks and the lead that might tell us what made it.

Host 2The paper is Testing the inelastic dark matter interpretation of the LZ event with heavy-element experiments, by Peter Graham, Juan Luengas, Harikrishnan Ramani and Samuel Wong, from Stanford, Delaware and Washington. It is a theory paper, a preprint, so no new data of its own. It asks what one strange detector event could mean, and how we would check.

Why we hunt for dark matter

Host 1Before the event itself, can you remind me why anyone builds a tank of xenon to look for dark matter in the first place?

Host 2Because most of the matter in the universe seems to be something we cannot see. In the wider picture of the field, galaxies spin too fast and light bends too much for the visible stars and gas to explain, and that has been argued since the nineteen thirties and firmed up through the seventies. Whatever it is, it does not shine, so the idea is to wait for it to bump into an ordinary atom.

Host 1And a bump means what, physically, inside the detector?

Host 2A dark matter particle hits a xenon nucleus and kicks it, like a cue ball hitting a billiard ball. The recoiling nucleus makes a tiny flash of light and a few freed electrons, and the detector measures how much energy the kick carried. That energy is called the recoil energy.

Host 1So who is LZ, and how long have experiments like it been doing this?

Host 2LZ is a large liquid xenon detector deep underground in South Dakota, and it has two big sister experiments, XENON and PandaX, which the paper also names. As the wider field tells it, these detectors have grown for about two decades from kilograms to many tonnes, and for most of that time the headline result was silence. That silence ruled out a lot of the simplest dark matter ideas.

What was known before

Host 1What kind of dark matter were they mostly expecting to see, back when this started?

Host 2The favourite was the weakly interacting massive particle, usually shortened to WIMP. It is a heavy particle that interacts through something like the weak nuclear force. The simplest version scatters elastically, meaning it bounces off the nucleus and stays exactly the same particle.

Host 1If it bounces off and stays the same, what does that predict about the energies the detector should see?

Host 2Mostly small kicks, with many more low energy events than high ones. The paper notes that ordinary elastic scattering would make a spectrum piled up at low energies. So if you saw only one event, at high energy, with nothing below it, that would look odd for a plain WIMP.

Host 1And the alternative they bring in is called inelastic dark matter, so what does inelastic mean in this setting?

Host 2It means the dark matter cannot just bounce; it has to change into a slightly heavier sister state when it scatters. The two states differ by a small amount of mass, which the paper calls the mass splitting. Paying that extra mass costs energy, so slow particles simply cannot scatter at all.

Host 1So it is a bit like a turnstile that needs a coin, and only particles with enough speed have the coin?

Host 2That is a fair picture. There is a kinematic threshold, and below it nothing happens. A side effect is that low energy recoils get suppressed, which is exactly the shape you would want if all you see is one high energy event.

Host 1You mentioned a particular candidate in the abstract, the Higgsino, so what is that in plain words?

Host 2It is a particle predicted by supersymmetry, a long-standing extension of particle physics where every known particle has a partner. In a simple limit, the two neutral Higgsino states form exactly this kind of inelastic pair. The paper stresses that it is very predictive: the strength of its interaction is fixed, and its mass is fixed too.

Host 1Fixed how, exactly, if nobody has ever seen one?

Host 2By the early universe. If Higgsinos were made hot in the Big Bang and then froze out as the universe cooled, getting today's amount of dark matter requires a mass of about one point one tera-electron-volts, roughly twelve hundred times the mass of a proton. Then the only real free knob left is the mass splitting.

The open question

Host 1So what exactly did LZ see that set all of this off?

Host 2One single nuclear recoil at two hundred and forty-eight kilo-electron-volts of energy. The paper says this is much higher than the usual window where people look for WIMPs, and there were no matching low energy events. That high, lonely event is what makes inelastic dark matter tempting.

Host 1Is one event a discovery, or could it just be something ordinary that happened to land there?

Host 2One event is not a discovery, and the paper never claims it is. It treats the event as a tentative signal and asks what it would imply if real. The authors say even confirmation by other xenon experiments would leave too much ambiguity to pin down what dark matter is.

Host 1You said there were strong constraints on the Higgsino story, so what is standing in its way?

Host 2Two things. First, the Sun: Higgsinos would be captured in the Sun and annihilate, making neutrinos, and the absence of those rules out mass splittings below five hundred and sixty-six kilo-electron-volts for the thermal Higgsino. Second, LZ saw no further events above two hundred and forty-eight, which cuts away more of the space.

Host 1And the simple fit to the LZ event lands where, compared to that five hundred and sixty-six line?

Host 2Below it, which is the problem. With the standard assumptions and LZ's own energy window, the fit wants splittings around three hundred and forty to three hundred and sixty in the standard halo, or about four ninety-five to five fifteen with the Large Magellanic Cloud model. Both are inside the region the Sun already excludes.

Host 1So the real puzzle is whether there is any honest way for a thermal Higgsino to explain this one event without breaking the solar limit?

Host 2Yes, that is the first half of the paper. The second half asks, whatever the model, which other experiments could actually test it. And the hinge for both is the speed of the fastest dark matter near Earth.

How the speed tail decides everything

Host 1Why would the very fastest particles matter so much more than the typical ones?

Host 2Because of the turnstile. A larger mass splitting needs a faster particle to pay the toll, so for inelastic dark matter only the high speed tail of the distribution can scatter at all. The paper says this tail is exactly where the usual model is least trustworthy.

Host 1What is the usual model, and what does it assume about speeds?

Host 2It is called the Standard Halo Model. It treats the dark matter around our galaxy as a gas with a bell-shaped spread of speeds, typical speed two hundred and twenty kilometres per second, cut off at the galaxy's escape speed of five hundred and forty-four. Seen from the moving Earth, the fastest particle in that model hits us at seven hundred and ninety-three kilometres per second.

Host 1And the Large Magellanic Cloud, which I know as a smudge in the southern sky, how does it change that picture?

Host 2It is a satellite galaxy that made its closest pass to the Milky Way roughly fifty million years ago. Recent galaxy simulations suggest that pass stirred up a population of extra fast dark matter, some of it not even bound to our galaxy. And it comes from the right direction to add to Earth's own motion.

Host 1Is the timing of the LZ event connected to that direction somehow?

Host 2Suggestively, yes. Inelastic scattering is easiest when Earth is moving fastest against the dark matter, which is around June, and the event was seen in June near that annual peak. The paper calls this suggestive, not proof.

Host 1How do they know what this Magellanic tail looks like, if nobody can measure dark matter speeds directly?

Host 2From a big simulation called Auriga, which models a galaxy like ours with its satellite. But the simulation only reports the flux down to a certain floor, and at that floor the top speed is nine hundred and fifty-four kilometres per second. The authors suspect that cutoff might partly be the simulation running out of particles, not nature stopping.

What the paper did

Host 1So what did they actually do with that uncertainty, since they are not running a new simulation?

Host 2They built a few hand-made versions of the tail and asked which ones fit. In one, they keep the simulation's shape but turn its fast flux down, mixing it with the standard model; at one percent strength they call it the y equals one percent model. In another, they stretch the top speed to eleven hundred kilometres per second at ten percent strength.

Host 1Is one percent of the simulated tail a tiny tweak, or a big one?

Host 2Big in a sense. The paper points out that in the usual language, the Magellanic Cloud's mass fraction becomes about zero point zero zero two six percent, which is one hundred times smaller than the benchmarks other papers use. And for the eleven hundred model, the extra flux sits ten times below the smallest flux the simulation even reports.

Host 1You said a fourth shape as well, involving black holes, so where does that one come from?

Host 2Pairs of black holes in the galaxy can fling a tiny fraction of dark matter to very high speeds. They include it as an example of any rare process that might boost a few particles beyond escape speed. It is an illustration, not a favoured scenario.

Host 1Once they have these speed shapes, how do they turn one event into a region of allowed answers?

Host 2They compute, for each cross section and mass splitting, how many events LZ should see and at what energies, using LZ's exposure of two point eight four tonne-years. Then a likelihood method draws the ninety percent confidence region around the best fit to a single event at two forty-eight. They also smear the energy by the detector's measurement uncertainty, so the true recoil could sit a bit above what was recorded.

Host 1You said earlier that LZ saw nothing above that event, so how far up does that silence actually count?

Host 2That is a real choice they make. LZ officially quotes its efficiency up to roughly two hundred and fifty kilo-electron-volts, but its calibration was checked against simulation up to six hundred and fifty. The authors mostly assume the detector works all the way to six hundred and fifty, which turns that empty region into evidence against models that would have put events there.

Host 1Does that make the fit easier or harder for the Higgsino?

Host 2Harder in the simplest case. With the full window, the standard and Magellanic models predict too many high energy events, so the allowed region slides to smaller cross sections, away from the Higgsino's fixed value. Splittings below about two hundred are also disfavoured, because they would predict events below two forty-eight that were not seen.

Host 1There is one honest warning in their statistics, I think, about what the region does and does not tell you?

Host 2Yes, the method only measures how likely a point is relative to the best fit, not how good the fit is in absolute terms. The authors say they checked that the best fits do describe the event reasonably well. But a region on the plot is not a stamp of approval.

What they found

Host 1So, the numbers: does any of their speed tails let the thermal Higgsino explain the event and survive the Sun?

Host 2One does, partly. With the top speed stretched to eleven hundred kilometres per second at ten percent strength, the allowed region crosses the Higgsino line between about five hundred and six hundred kilo-electron-volts of splitting. The slice above five hundred and sixty-six escapes the solar limit.

Host 1And the one percent version, the one with the flux turned down?

Host 2Turning the fast flux down a hundredfold pushes the region up onto the Higgsino line, overlapping between four hundred and forty and five hundred and ten. That is a better fit, but still inside the solar exclusion. So less fast dark matter, oddly, helps, because the full simulated tail predicts too many events.

Host 1And the black hole version, does that rescue anything?

Host 2It opens a region at very large cross sections, roughly ten to the minus thirty-six to ten to the minus thirty-two square centimetres, for splittings from four hundred to fourteen hundred. But to reach the thermal Higgsino past the solar limit, that boosted flux would have to be a thousand times larger. The authors say it is unclear whether that is possible.

Host 1What if the Higgsino is not the thermal one, but heavier?

Host 2They test ten tera-electron-volts, which needs a non-standard history of the early universe. There the solar limit is weaker, around three hundred and sixty-five, and the fit just touches the Higgsino line in the standard halo at three sixty-five to three eighty. With the modified tails the overlap is large, for example four hundred and fifty to five ninety-five with the eleven hundred model.

Host 1So if I have this right, the best-motivated Higgsino only works if the fastest dark matter is faster than any simulation has yet shown, but in amounts too small for simulations to see?

Host 2That is right, and the authors say so plainly. They call that extended tail speculative, and their main conclusion from this half is a request for higher resolution simulations. It is a statement about what would be needed, not evidence that it exists.

Host 1Before we get to the rocks and the lead, is there a word from someone who pays for this?

A word from our sponsor

Host 2This episode is brought to you by Witness Rock. A plain pebble, quarried five kilometres underground, that quietly records every footstep in your hallway as a tiny damage track.

Host 1I have had mine on the hall table for six weeks. It has seen everything, and it is not saying anything yet.

Host 2Readout takes about a billion years, by X-ray, in a sample the size of a sugar cube. For the best signal to noise, buy your Witness Rock fifty million years ago.

Host 1I did buy mine fifty million years ago. Delivery was slow, but the continents were in a different place then.

Host 2Witness Rock. It does not forgive, it does not forget, and it has no opinion about your shoes.

Why heavier elements help

Host 1Back to the paper, then, so why would xenon alone not be enough, even if XENON and PandaX saw more events?

Host 2Two reasons. A handful of events could also come from a different model, elastic scattering whose strength depends on the momentum of the kick. And even within inelastic dark matter, a xenon spectrum only fixes combinations of the mass, the splitting, the cross section and the speed distribution, so they stay tangled together.

Host 1Why does a heavier nucleus, like lead, untangle that?

Host 2Simple kinematics: a heavier target gives more collision energy to work with, and the largest splitting it can reach grows in step with its mass. Xenon has mass number one hundred and thirty-one, lead two hundred and seven. A splitting that sits right at xenon's limit sits comfortably inside lead's, so lead sees far more of the speed distribution and much higher rates.

Reading ancient rocks

Host 1The first experiment is a paleodetector, so what is that in everyday language?

Host 2It is a rock used as a detector that has been switched on for millions of years. When a nucleus inside a mineral recoils, it leaves a damage track a few nanometres to a few hundred nanometres long, and certain minerals keep those tracks for geological time. You dig up the rock and read the tracks.

Host 1How do you even read a scratch that is a few billionths of a metre long inside a stone?

Host 2The paper assumes small angle X-ray scattering tomography, which can map tracks in three dimensions with a resolution of about fifteen nanometres across sixty cubic centimetres of sample. It also mentions a helium ion beam microscope reaching about one nanometre, but on only up to six cubic millimetres. The rock has to come from about five kilometres down, away from cosmic rays, and be very low in uranium.

Host 1Why did older paleodetector proposals not already cover this?

Host 2Because they used minerals whose heaviest element is iron, mass fifty-six, which is lighter than xenon. That makes them blind to the region the LZ event points to. The authors earlier proposed lead-bearing minerals instead, for example one called laurionite, a lead chloride hydroxide that forms from deep geothermal brines and whose hydrogen helps damp neutron backgrounds.

Host 1So how much of the LZ region could a laurionite sample actually cover?

Host 2In the standard halo, with a billion years of exposure and the sixty cubic centimetre readout, it covers the vast majority of the region. Even a sample a hundred thousand times smaller, zero point six cubic millimetres, covers part of it. In the modified tails, it covers every place where the region overlaps the Higgsino line, with Higgsino reach up to about seven eighty and six ninety-five kilo-electron-volts.

Host 1Wait, a billion years for the standard case, but what exposure do they use for the Magellanic cases?

Host 2Fifty million years, because the fast tail only arrived with that pass. They assume the halo was plain standard before then, so an older rock just collects more background without more signal. That gives a test no live experiment can do.

Host 1What test is that, in plain terms, using rocks of different ages?

Host 2Line up minerals both much younger and much older than about fifty million years. If the signal comes from the Magellanic pass, signal grows with age only up to that point and then stops, while background keeps growing, so signal to noise peaks near fifty to one hundred million years. Seeing that peak would point to the Magellanic tail and favour the thermal Higgsino.

Host 1And you said two different minerals could pin down the mass and the splitting, so how does that work?

Host 2From the track lengths you rebuild the energy spectrum, and its peak sits at a special recoil energy that depends on the dark matter mass, the splitting and the target's mass. One element gives one equation with two unknowns. A second element gives the second equation, although this works best when the dark matter is not too much heavier than the nucleus.

Making dark matter glow

Host 1The second experiment uses light, so how does a dark matter particle make a flash far from any detector?

Host 2If the heavier sister state can decay back by emitting a photon, then a particle that up-scatters on lead or uranium in the Earth's rock travels on and later releases a photon whose energy equals the mass splitting exactly. A large underground neutrino detector, such as JUNO in China, could catch those photons as a sharp line. For the Higgsino this decay length is calculable.

Host 1And the heavy-element supplement means literally piling lead around the detector?

Host 2Yes, a shell of lead or depleted uranium around it, in volumes of five or twenty metres on a side. The paper notes there are about a million tonnes of depleted uranium in the world with no large use, so space and digging, not material, are the limits. And you can check a signal by taking the supplement away.

Host 1How well would that do against the LZ region, over what time span?

Host 2They assume about ten years of JUNO data. In the standard halo, JUNO alone already touches part of the region, and large supplements reach its upper half. In the eleven hundred model, the twenty metre supplements cover the window between five hundred and sixty-six and six hundred, the very slice where a thermal Higgsino would survive the Sun.

Host 1What would a positive signal tell us that xenon never could?

Host 2It would show the dark matter is inelastic and that it couples to light, which is not guaranteed for inelastic models. The photon's energy gives the splitting directly. And by varying the supplement you measure the decay length, which would say whether it is the Higgsino.

What it means

Host 1So stepping back, what changes, and for whom, if this paper is right?

Host 2Mostly it changes the to-do list for physicists. Simulation groups get a concrete request to resolve the fastest, rarest dark matter better, and experimental groups get two specific ways to test the event that do not depend on more xenon. For everyone else, nothing changes today; this is about how to check a single, unconfirmed event.

Host 1What should a listener definitely not conclude from all this?

Host 2That dark matter has been found, or that it is the Higgsino. The paper shows a thermal Higgsino can fit only with an assumed speed tail beyond current simulations, and the two experiments are projections, not results. The authors add that these methods are worth doing even if the LZ event turns out not to be dark matter.

Honest caveats

Host 1Where are the softest spots in the argument, if you had to name them?

Host 2First, everything rests on one event. Second, the speed tails are hand-built examples, chosen partly to fit the event, and the authors label the eleven hundred kilometre per second tail speculative. Third, they use LZ's efficiency up to six hundred and fifty rather than the officially quoted window, and that choice moves the contours.

Host 1And the experiments themselves, are those ready to switch on?

Host 2No. Lead-bearing paleodetector minerals of the needed purity still have to be found, and the readout volumes are assumptions. The lead and uranium shells around JUNO are proposals by the same authors, and they note issues like space and excavation cost.

Host 1Is it a concern that the authors are promoting experiments they themselves invented?

Host 2It is fair to note, and they say so openly: both ideas were proposed earlier by some of them. That does not make the physics wrong, but independent groups should check the projected reach. Their simplified Poisson limit and simple Earth model are also approximations.

What to watch next

Host 1So what should we watch for over the next few years to see how this plays out?

Host 2First, whether LZ, XENON or PandaX see more high energy events, or none. Second, higher resolution simulations of the Magellanic Cloud's effect on the fastest dark matter. Third, any search for lead-bearing minerals pure enough for paleodetection, and whether JUNO's data gets searched for a photon line below a million electron-volts.

Host 1And if more xenon events do arrive, what would make you take the inelastic story more seriously?

Host 2An edge: a clear minimum recoil energy with nothing below it, which the paper says only a large event sample in one target can resolve. Then a matching signal in a heavier element would be the real cross-check. One flash is a question; it is not yet an answer.

Sign-off

Host 1That was the deep dive on what one high energy flash in LZ could mean. Every weekend this same feed carries the roundup, the week's new papers in one short episode.

Host 2Thanks for listening to the Hastrup Review of Science. Until next time.