Deep divefrom Episode 6Cosmology15 October 202621 min

How Fast Can Dark Matter Hit a Detector in June?

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: The high-velocity dark matter halo of the Milky Way in light of the LZ 248 keV event (O'Hare, 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 event in a dark matter detector, and the question of how fast anything in our galaxy can actually be moving.

Host 2The paper is called The high-velocity dark matter halo of the Milky Way in light of the LZ two hundred forty-eight kilo-electron-volt event. It has a single author, Ciaran O'Hare at the University of Sydney. It was posted to arXiv in September twenty twenty-six.

Host 1So a preprint, not yet peer reviewed.

Host 2Correct. Every number today comes from that preprint, and where I add textbook background, I will say so.

Host 1The claim, in two sentences.

Host 2If that one event really was dark matter, some popular models need it to come from the very fastest particles in the galaxy, a part of the distribution nobody has measured. Using ninety-eight simulated Milky Way look-alikes, the paper finds the fastest dark matter reaching Earth on the day of the event moved at about seven hundred ninety-three kilometers per second, with a plausible range of roughly seven hundred to eight hundred seventy-five.

Why this field exists

Host 1Start with background. Why are people looking for dark matter in a tank underground at all?

Host 2Background, not from this paper: since the nineteen thirties, and especially since Vera Rubin's galaxy rotation curves in the nineteen seventies, astronomers have seen more gravity than visible matter can explain. The leading explanation is a halo of invisible particles surrounding every galaxy, including ours. If that is right, those particles are streaming through the Earth constantly.

Host 1And a detector tries to catch one bumping into an atom.

Host 2Yes. A particle occasionally hits an atomic nucleus and the nucleus recoils, like a struck billiard ball. Experiments such as LUX-ZEPLIN, a tank of several tonnes of liquid xenon deep underground in South Dakota, watch for the faint flash of light and the electrons that recoil releases.

Host 1What decides how hard the nucleus gets hit?

Host 2Two things: the physics of the particle, and how fast it is moving when it arrives. The paper's point is that the second ingredient is an astrophysical input that experiments mostly assume rather than measure.

What was known before

Host 1So what is the standard assumption about those speeds?

Host 2It is called the Standard Halo Model. The dark matter near the Sun is treated as a cloud whose velocities follow a simple bell curve, the same in every direction, cut off at the galaxy's escape speed. The paper uses the conventional values: a typical speed of two hundred thirty-eight kilometers per second and an escape speed of five hundred forty-four.

Host 1Why cut it off at the escape speed?

Host 2Anything faster than escape speed should not stay bound to the galaxy, so in a settled halo you expect essentially none of it. The escape speed is set by how deep the gravitational well is at the Sun's position.

Host 1Is that number well measured?

Host 2Not very. It is inferred from the fastest nearby stars, and the paper notes that even with the Gaia satellite, estimates range from about five hundred to five hundred eighty kilometers per second, with large error bars. Those results depend on assumptions about the shape of the stellar velocity tail, on substructure, and on the galaxy not being fully settled.

Host 1And stars are not dark matter.

Host 2That is the second problem. Stars and dark matter have different overall distributions, so a stellar escape speed does not translate cleanly into the dark matter tail. On top of that, earlier simulations suggest the Large Magellanic Cloud, a satellite galaxy falling into ours, may have dragged in dark matter faster than the local escape speed.

Host 1How much has the field worried about this until now?

Host 2Moderately. For the usual searches the fast tail is already exponentially rare, so changing it shifts the total event rate only modestly. People have explored many departures from the standard model, including flattened halos, uneven velocities, a co-rotating dark disk and streams from shredded satellites, but the tail was usually a secondary concern.

The open question

Host 1Then what changed?

Host 2Two kinds of models make the tail decisive. One is very light dark matter, where only the fastest particles carry enough energy to register. The other is called inelastic dark matter, and that is where this paper lives.

Host 1Explain inelastic.

Host 2In these models the dark matter particle has a slightly heavier sibling state. To scatter off a nucleus, the incoming particle must be promoted to that heavier state, which costs a fixed amount of energy called the mass splitting. If the particle is too slow to pay that cost, nothing happens at all.

Host 1So there is a speed threshold.

Host 2Yes, and a well-motivated example is the Higgsino, a particle from supersymmetry. The paper notes that at a mass of about one point one tera-electron-volts it would be produced in the early universe in just the right amount to be all of the dark matter, and it has two neutral states separated by less than a mega-electron-volt.

Host 1Where does the LUX-ZEPLIN event come in?

Host 2LUX-ZEPLIN analysed an extended window of high recoil energies, up to two hundred seventy kilo-electron-volts, with an exposure of two point eight four tonne-years. They saw one event consistent with a nuclear recoil at two hundred forty-eight kilo-electron-volts, plus or minus twenty-three statistical and twenty-three systematic, in a region where essentially no background is expected.

Host 1One event.

Host 2One. The paper is explicit that a single event cannot establish a dark matter signal. What makes it interesting is the absence of the usual pattern: ordinary elastic scattering would produce many more low-energy recoils falling off smoothly, and those were not seen.

Host 1And inelastic scattering explains that absence.

Host 2Naturally, because low-energy recoils are kinematically forbidden. But for many parameter choices, a recoil that energetic requires particles near the very top of the speed distribution. So the open question becomes concrete: how fast is the fastest dark matter hitting Earth, and how sure are we?

What the paper did

Host 1What did this paper actually do?

Host 2Since the tail cannot be observed, it turns to simulations of galaxies that grew inside a full model of the universe. The main sample is ninety-eight Milky Way look-alikes from IllustrisTNG, a large cosmological simulation that follows dark matter, gas, stars and black holes together in a cube about fifty-two megaparsecs on a side.

Host 1How were the ninety-eight chosen?

Host 2A previous study selected them: galaxies with between forty and seventy-three billion solar masses of stars, and isolated enough to resemble our environment. Some have histories resembling ours, including a merger like the Gaia-Sausage-Enceladus collision and an ongoing interaction with a Magellanic Cloud-like satellite.

Host 1Why add a second simulation?

Host 2As a cross-check. The paper uses three publicly available higher-resolution galaxies from the FIRE two Latte suite, which zooms in on single galaxies. They have similar halo masses but different assembly histories, and different handling of the gas and star physics.

Host 1Ninety-eight is a lot of galaxies, but none of them is the Milky Way.

Host 2That is the first methodological problem the paper tackles. The simulated disks are generally less centrally concentrated than ours, so placing a fake Sun at eight point three kiloparsecs from the centre would not put it in an equivalent spot.

Host 1How do you fix that?

Host 2With an energy-conserving rescaling from an earlier study, which compresses each galaxy in position and velocity until the circular orbit speed at the Sun's radius matches the Milky Way's measured value, two hundred thirty-eight kilometers per second. That value is known to about five kilometers per second, thanks to precise measurements of the black hole at the galactic centre and of the Sun's own motion.

Host 1Then they pick out the dark matter near the fake Sun.

Host 2Right. Each galaxy is rotated so its disk lies flat, and every dark matter particle inside a ring-shaped tube around the galaxy is counted as local. The ring sits at eight point three kiloparsecs with a cross-section radius of one kiloparsec, and the paper says the conclusions do not change when that radius moves by half a kiloparsec either way.

Host 1Why a whole ring rather than one spot?

Host 2There is no principled way to choose where on the circle the Sun sits, so averaging around it removes an arbitrary choice and gains particles. Velocities are then described in three directions: radially outward, along the direction of rotation, and vertically out of the disk.

Host 1And then they turn galactic velocities into what a detector would see.

Host 2Yes. The galaxy-frame velocities are boosted by the Sun's motion, about two hundred fifty kilometers per second along the direction of rotation, plus Earth's orbital speed of about thirty kilometers per second, which adds in one season and subtracts in another. June sixteenth, the date the LUX-ZEPLIN event was recorded, falls near the time of year when the combined speed is greatest.

Host 1What quantity do they compute from the particles?

Host 2For standard interactions, everything the velocities contribute to the event rate is captured in one function, the average inverse speed of all particles faster than some minimum. The paper computes it straight from the simulation particles, with no binning and no fitted curve, which matters when you care about a sparse tail.

Results

Host 1Results. First, does the simulated halo look like the textbook model?

Host 2In bulk, mostly yes. The radial velocities are slightly hotter than the standard model and the vertical ones slightly colder, and all three directions have heavier tails than a bell curve, most visibly in the vertical direction.

Host 1But the halo spins.

Host 2That is the notable departure. Consistent with earlier work, the dark matter halos rotate in the same sense as the disk. Across the ninety-eight halos, the middle sixty-eight percent have a mean rotational velocity between five and fifty-eight kilometers per second.

Host 1Why would a cloud of dark matter rotate with the stars?

Host 2The paper says the reasons have not been fully investigated. Halos in the standard cosmological model are expected to pick up spin from torques during assembly, and if a disk forms early, incoming satellites may merge preferentially along its plane, adding angular momentum in a common direction.

Host 1Here is what I would expect. If the halo moves with us, the head-wind of dark matter is weaker, so the fastest hits should be slower.

Host 2That is the intuitive expectation, and the paper's main conceptual finding is that it largely fails. The rotational velocity distribution is not slid sideways as a whole; it is skewed. The bulk leans toward the rotating direction while a long tail stretches out in the opposite direction.

Host 1And the opposite direction is the one that matters.

Host 2Exactly. The fastest particles in a detector are the ones flying against the Sun's motion, and that tail stays roughly where it was. Across all halos and all three directions, the maximum speeds sit in roughly the same range.

Host 1Do the high-resolution galaxies agree?

Host 2Broadly. The three FIRE galaxies land in the same region of the summary statistics, and one, called m twelve f, has a strong negative skew and a mean rotation of about seventy kilometers per second, close to the most extreme of the Illustris halos.

Host 1Now the headline number.

Host 2On June sixteenth, the maximum dark matter speed in the laboratory frame has a median of seven hundred ninety-two point seven kilometers per second across the sample. Ninety-five percent of halos fall between six hundred ninety-seven point one and eight hundred seventy-five point six.

Host 1How does that compare with the standard model?

Host 2The median is very close to the standard model value. The paper is candid that this is partly by construction, because the rescaling forces every halo to share the standard model's rotation speed.

Host 1What about the shape of that inverse speed function?

Host 2It hugs the standard model at moderate speeds. Below about two hundred kilometers per second it is slightly higher, and above that slightly lower, because the co-rotation shifts the bulk toward lower lab speeds. Above about seven hundred, near the resolution limit of the simulations, the differences between halos dominate.

A word from our sponsor

Host 1Before the season effect, a quick word from our sponsor.

Host 1Is your water too warm, and too liquid, for the life you want to live? The Hastrup Review of Science is brought to you by ice cube trays.

Host 2I like the silicone ones. You bend them gently and everything just pops right out.

Host 1Fourteen compartments, each one holding its shape all night. They are very reasonably priced and in stock right now.

Host 2I use mine every evening. Firm, reliable, and they never leave me without something cold and hard when I need it.

Host 1Ice cube trays. Our referral link is in the show notes.

The annual rhythm

Host 1Does the speed limit change through the year?

Host 2It does, and this is where the tail becomes testable. Because Earth's orbital motion adds to the Sun's in summer and subtracts in winter, a signal that only comes from very fast particles switches on and off with the seasons. Textbook background: this annual modulation idea dates to the nineteen eighties.

Host 1How strong does it get?

Host 2The paper integrates only the contribution from particles faster than seven hundred kilometers per second. Most halos still peak around June, with a spread of about a month, and on average they match the standard model almost exactly.

Host 1And the outliers?

Host 2A subset switch on much more sharply in season, because their tails are slightly suppressed. In the most extreme cases, about five percent of halos, the rate almost vanishes for three-quarters of the year, even at that modest threshold.

The LUX-ZEPLIN event

Host 1Now apply this to the event itself.

Host 2The paper builds the expected recoil spectrum for an inelastic model on xenon, using the simulated velocity curves. It chooses a dark matter mass of one tera-electron-volt and a mass splitting of three hundred seventy-five kilo-electron-volts, a combination able to produce the event and chosen to align with the thermal Higgsino.

Host 1What goes into that calculation?

Host 2Equal couplings to protons and neutrons, nuclear structure factors for all the stable xenon isotopes from shell-model calculations, and a blurring for detector resolution matched to the event's quoted uncertainty. The total rate is left arbitrary; only the shape of the spectrum is compared.

Host 1And the shape depends on the halo.

Host 2Substantially. Averaged over the simulated halos, the predicted spectrum is boosted below the observed energy and suppressed at higher energies.

Host 1Why do the higher energies matter?

Host 2Because this Higgsino-like model has a known problem: it predicts events in a higher-energy sideband from about three hundred fifty to six hundred eighty kilo-electron-volts where LUX-ZEPLIN saw nothing. The simulations' shorter tail somewhat eases that tension, but the paper notes the sideband is not calibrated for nuclear recoils, so an empty sideband is itself hard to interpret.

Host 1There is also a constraint from the Sun.

Host 2Yes. Higgsinos would be captured by the Sun and annihilate into neutrinos that the IceCube detector at the South Pole would see. That sets a lower limit on the mass splitting.

Host 1And the speed limit sets an upper limit.

Host 2Right. If the splitting is too large, no particle in the halo is fast enough to scatter on xenon, so for each halo there is a maximum allowed splitting. The paper computes that using the heaviest common xenon isotope, xenon one hundred thirty-six, which has the most kinematic reach.

Host 1So the model has to fit between the two.

Host 2For the pure thermal Higgsino near one tera-electron-volt, with one choice of the IceCube bound, only the most extreme two and a half percent or so of halos supply particles fast enough. Even then, the paper says, a good fit would need the measured recoil energy to have fluctuated well below its true value.

Host 1So the simulations make it harder for the Higgsino, not easier.

Host 2Slightly harder. On average the simulated halos give a lower top speed on June sixteenth than the standard model, so this speed constraint is a little more severe than the textbook version.

Why it matters

Host 1Step back. What is the principle here?

Host 2Near a kinematic threshold, the astrophysics stops being a nuisance and becomes a leading uncertainty. A modest change in the fastest few particles can change a predicted rate by orders of magnitude, and here the paper puts an honest range on that input instead of one assumed number.

Host 1Who should change what they do?

Host 2Theorists fitting this event with inelastic models can use the spread of top speeds, roughly seven hundred to eight hundred seventy-five kilometers per second, instead of a single standard-model cutoff. Experimenters get a reason to look at when events happen in the year, because a fast-tail signal should cluster around June.

Host 1And the rotating-halo point?

Host 2It closes off a tempting argument. Co-rotation is real in these simulations, but it does not shave meaningfully off the maximum speed, so it cannot be used to rescue or rule out a model by lowering the speed limit.

Host 1What does it not mean?

Host 2It does not mean the event was dark matter, and it does not mean dark matter is a Higgsino. It shows the leading inelastic explanation sits close to the edge of what the galaxy can kinematically supply, and that the standard model's speed limit is a reasonable middle estimate.

Caveats

Host 1Caveats.

Host 2The biggest is that everything rests on one event. One recoil cannot distinguish dark matter from an unmodelled background, and the paper never claims otherwise.

Host 1And the simulations?

Host 2They run out of particles just where the question lives. Above about seven hundred kilometers per second the tail is sparsely sampled, so the ninety-five percent range partly reflects resolution, not just real galaxy-to-galaxy variation.

Host 1The rescaling also bothers me.

Host 2Fairly so. It forces every halo to match the Milky Way's rotation speed, which is why the median lands near the standard model, and it assumes a simple energy-conserving compression works for galaxies with different structures. The look-alikes are still not our galaxy.

Host 1Anything left out entirely?

Host 2Dark matter boosted to very high speeds by cosmic rays is explicitly set aside, as are particles from outside the galaxy. The specific Magellanic Cloud effect is discussed but not isolated, and the Illustris halos include only some galaxies with such a satellite.

Host 1And the model choice?

Host 2The paper only works through the endothermic case, where energy is absorbed. It notes the opposite case, where energy is released, can also explain the event, and that would not lean on the fast tail in the same way.

Host 1Anything about the work itself?

Host 2It is a single-author preprint, and the author discloses using large language models to help write code, while taking responsibility for all scientific judgments. That is a transparency note, not a red flag, but independent reproduction of the numbers would be welcome.

What to watch next

Host 1What should we watch for next?

Host 2More exposure. LUX-ZEPLIN is still running, other xenon experiments are collecting data, and larger next-generation detectors are planned. If more unexplained high-energy recoils appear, the paper lists what would test the fast-tail story.

Host 1Which is?

Host 2The shape of the energy spectrum, the timing of events through the year, and eventually the direction recoils come from, which should be strongly lopsided in this regime. Directional measurements, though, need more advanced detectors than exist at this scale.

Host 1And if no more events appear?

Host 2Then the event stays an outlier, and the lasting result of this paper is the method: a simulation-based range for the galaxy's dark matter speed limit, ready for the next anomaly.

Sign-off

Host 1That is the deep dive: The high-velocity dark matter halo of the Milky Way in light of the LZ two hundred forty-eight kilo-electron-volt event, by Ciaran O'Hare. The link is in the show notes.

Host 2Short version: ninety-eight simulated galaxies put the fastest dark matter reaching Earth on June sixteenth at about seven hundred ninety-three kilometers per second, close to the textbook value, and halo rotation does not lower it much. The leading Higgsino explanation of the event sits near that edge, and one event is still one event.

Host 1Every Friday this same feed carries the roundup episode: five new papers as trailers, and you pick which become deep dives. This has been the Hastrup Review of Science.