Deep divefrom Episode 11Cosmology9 November 202619 min

Same Thirteen Events, a Looser Limit on Dark Black Holes

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: Same Events, Weaker Bound: Model Dependence in Magellanic Microlensing Limits on Primordial Black Holes (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.

Host 2Today is one paper, a preprint dated the twenty-third of October by David Morales-Zapien, Lucas Brown and Stefano Profumo, at California State University Long Beach and the University of California, Santa Cruz. It asks how much the best limits on black holes from the early universe as dark matter depend on the model of our galaxy used to calculate them.

Host 1Before we start, has anyone outside the team checked this work yet, or is it still a draft?

Host 2It is a preprint, so it has not been through journal review. It is also careful to say what it is not: the authors call their main result a measure of sensitivity, not a new calibrated limit.

Black holes from the first second

Host 1Let's begin with the basics then: what is a primordial black hole, and why would anyone think it could be dark matter?

Host 2A primordial black hole is one that would have formed in the very early universe, straight from dense patches of the hot young cosmos, rather than from a dying star. The idea goes back to Zel'dovich and Novikov in nineteen sixty-seven and to Stephen Hawking in nineteen seventy-one, and both are cited in this paper.

Host 1And the link to dark matter, how does that come in?

Host 2In the wider picture, dark matter is the unseen mass that holds galaxies together, roughly five times more than all ordinary matter, and nobody knows what it is. Black holes made early enough would be dark, heavy and cold, so in principle they could be some or all of it, and the paper says they remain a viable candidate over a wide range of masses.

Host 1Why is this idea getting attention again now, rather than being a nineteen-seventies curiosity?

Host 2The paper names two reasons. Gravitational-wave detectors have been hearing black holes merge, and a survey toward the centre of our galaxy found an excess of very short microlensing events that could come from objects of roughly planetary mass.

How you weigh the invisible

Host 1You just said microlensing. What is that, in plain words?

Host 2Microlensing is when a dark object drifts almost exactly in front of a distant star and its gravity bends the starlight, so the star brightens and then fades back over days or weeks. You never see the lens itself, only the brief brightening of the star behind it.

Host 1So the black hole is a kind of invisible magnifying glass that passes in front of a star?

Host 2Exactly that, and the useful part is the duration. The time it takes, which astronomers call the Einstein timescale, grows with the mass of the lens, so heavier lenses make longer events, though speed and distance blur that relationship.

Host 1Which stars do astronomers watch to catch these events?

Host 2For the dark matter question, the Magellanic Clouds, two small galaxies orbiting ours about a hundred and sixty thousand light-years away. Light from their stars crosses the dark matter halo of the Milky Way on its way to us, so if that halo were made of black holes, some of those stars would brighten.

Host 1How long have people been running this kind of search?

Host 2The method was proposed by Bohdan Paczynski in nineteen eighty-six. In the wider history of the field, the American-Australian MACHO project and the French EROS project watched the Clouds through the nineteen-nineties and two-thousands, and the paper builds on EROS-two and, above all, on the Polish survey called OGLE.

The limit everyone uses

Host 1What had OGLE concluded before this paper came along?

Host 2OGLE's twenty-year survey of the Large Magellanic Cloud, about eighty million light curves from two thousand and one to twenty twenty, picked out thirteen events with objective criteria. From that, OGLE limits these black holes to at most a few percent of the dark matter, over masses from a hundred-millionth of the Sun up to a thousand Suns.

Host 1Thirteen sounds tiny for eighty million stars. Is that a lot of events or very few?

Host 2Very few, and that is the point. If the halo were full of black holes in that mass range, you would expect far more, so a small count rules them out, and the paper notes that these curves are routinely used as fixed inputs by other cosmology and particle-physics analyses.

Host 1You said a few percent. Is that from assuming none of the thirteen events came from a black hole?

Host 2That is one version. OGLE publishes a strict limit, which assumes no detected event is a black hole, and a relaxed limit, where both ordinary stars and black holes can explain the detections; the two differ by at most about a factor of four, near a quarter of the Sun's mass.

The open question

Host 1So where is the argument? If OGLE already gave two versions, what is left to dispute?

Host 2The other half of the calculation: how many events you would expect if black holes were the dark matter. That prediction needs a model of the Milky Way's halo, of the Large Magellanic Cloud's own halo, and of all the ordinary stars that can act as lenses too.

Host 1And people disagree about those models?

Host 2Yes, publicly. Hawkins and García-Bellido argued last year that the halo model and detection efficiencies were wrong and that black holes could be most of the dark matter, and the OGLE team replied that plausible changes to the Milky Way halo move the relevant quantity by only a few percent.

Host 1So this paper is a kind of referee, trying to see who is right?

Host 2More of a stress test than a referee. They keep OGLE's events and OGLE's detection efficiencies exactly as published, change only the galaxy model, each change motivated by recent measurements, and see how far the limit moves.

Same events, new galaxy

Host 1Did they get OGLE's actual data, or did they rebuild it from the published papers?

Host 2The real thing. The paper thanks Przemek Mróz of OGLE for supplying detection efficiencies, star counts per field, and the timescale measurements of the thirteen events, as two hundred thousand samples each describing the uncertainty in each event's duration.

Host 1Was that only the twenty-year survey, or something else too?

Host 2Also OGLE's high-cadence campaign from October twenty twenty-two to May twenty twenty-four, which imaged five fields in the Large Cloud and four in the Small Cloud every sixteen to twenty minutes, about thirty-five million stars. That one is built to catch very short events from planet-mass objects, and it found one event, lasting about ninety days, which is far too long for those.

Host 1Start with the Milky Way then. What did they change about our own galaxy's halo?

Host 2OGLE used a round halo. This paper uses a flattened one, squashed toward the plane of the disk, because recent analyses with the Gaia satellite favour a ratio of short to long axis of about zero point seven five to zero point seven eight.

Host 1Why would squashing the halo matter for a star in another galaxy?

Host 2Because the sightline to the Large Cloud leaves our galaxy well below the disk. At the same density near the Sun, a flatter halo puts less dark matter along that line, and in the benchmark, with an axis ratio of zero point seven two, the flattening alone cuts the dark matter column by about twelve percent and the expected black hole events from our halo by nineteen percent.

Host 1What did they assume for the density of dark matter near the Sun?

Host 2For the benchmark, zero point three in the units physicists use, nine percent below the zero point three three in OGLE's model and consistent with local estimates. They also checked that this galaxy still spins at the speeds measured from its stars, roughly two hundred and fifteen kilometres per second at the Sun.

Host 1You keep saying benchmark. Is there a second, more extreme version?

Host 2Yes, a stress test, with a local density of zero point two and an axis ratio of zero point six, deliberately beyond what the data favour. Its rotation speed far out, about a hundred and thirty-three kilometres per second, is well below the roughly one hundred and eighty implied by distant halo stars, so the authors themselves call it deliberately extreme.

Host 1There was also the idea that the Large Cloud is so heavy it disturbs our halo. Does that help the black holes?

Host 2It is the opposite of what critics assumed. Using a live computer simulation of the Milky Way and the Large Cloud together, they find the halo is denser on the Cloud's side, which raises the expected black hole signal by about six percent.

Host 1So the disturbance makes the limit stricter, not looser?

Host 2Slightly stricter, yes. Only about seven percent of the lensing comes from beyond about a hundred thousand light-years from the galactic centre, where the visible wake lies, and to stay on the safe side they leave this six percent boost out.

Host 1What about the dark matter halo of the Large Cloud itself?

Host 2OGLE used a halo that gets densest at the centre. The benchmark uses a cored profile fitted to the Cloud's own rotation, flatter in the middle, which lowers the dark matter right where the source stars sit, while still matching the Cloud's measured mass once stars and gas are added.

Host 1And the extreme version did something stronger there?

Host 2It used a smaller cored halo cut off at fifteen kiloparsecs, which holds only about ten billion Suns within the radius where the mass is measured. That would need about seventeen billion Suns of ordinary matter to match measurements, far more than the Cloud's stars, so again it is an extreme case, not a consistent model.

Host 1Then there are the ordinary stars that also cause events. What changed there?

Host 2Three things. Heavier Milky Way disks from a recent Gaia-era model, a lower stellar mass for the Large Cloud, two point one instead of two point seven billion Suns, and a different initial mass function, which is the recipe for how many stars of each size a population contains.

Host 1Why would the recipe of star sizes change the result?

Host 2Because it shifts how long stellar events last, and the black hole limit depends on which events could have come from stars instead. They chose shallower slopes, each at the low edge of its published uncertainty range, and they say plainly that picking all four edges at once is far less probable than any one of them.

Host 1Once the model is built, how do they actually get a limit out of it?

Host 2They predict the event rate and the distribution of event durations, field by field, and compare it with the thirteen measured durations in a statistical likelihood, the same structure as OGLE's relaxed analysis. Each detected event is shared between stars and black holes in proportion to how likely each is to produce that duration, and the limit is the abundance that ninety-five percent of the probability lies below.

Host 2We will get to what those changes do to the limit after this.

A word from our sponsor

Host 1This episode is brought to you by Kindweigh bathroom scales, who are voluntarily recalling their spring model. The scale held your body fixed and changed its assumptions about the floor.

Host 2I had one. Same me every morning, weaker reading, by up to a factor of six.

Host 1Under an extreme stress test, it assumed a flattened house, and several owners were recorded as lighter than their own shadow.

Host 2Mine started clustering my weight in the hallway.

Host 1Replacement scales quote every reading together with its modelling assumptions. Kindweigh has already weighed the moon, and found it consistent with one cat.

What the numbers say

Host 1Back to the galaxy. Before the limit itself, did the new model change how many ordinary star events it predicts?

Host 2A lot. Self-lensing by stars inside the Large Cloud drops from five point seven predicted events to two point nine, mainly because of the lower stellar mass and the new star recipe, while Milky Way disk events rise from seven to about eight point eight because the disks are heavier.

Host 1And the headline: how much weaker is the black hole limit?

Host 2Against OGLE's relaxed limit, which is the fair like-for-like comparison, the benchmark weakens it by at most a factor of about one point six, near a third of the Sun's mass. The extreme stress test weakens it by up to about three point four, near one twentieth of the Sun's mass.

Host 1The title said up to six, though. Where does six come from?

Host 2From comparing with the strict limit instead, where the factors reach six point two for the benchmark and thirteen point two for the stress test. But that number also includes the difference between strict and relaxed, which OGLE already published, so the part that really comes from the galaxy model is the one point six.

Host 1So if I have this right, a few percent might be more like several percent?

Host 2Roughly, at the most sensitive masses. At a typical star mass, around zero point four five of the Sun, their benchmark limit is about two point six percent of the dark matter, and it is mass dependent, so no simple rescaling of event counts reproduces it.

Host 1Which single change mattered most?

Host 2Swapping the star recipe back to the standard one tightens the benchmark limit by a factor of about one point five, the largest single effect. Restoring all halo ingredients or all stellar ingredients each tightens it by about one point three, and the heavier Milky Way disks actually push the other way.

Host 1Did anything surprising show up in the fit itself, apart from the limit moving?

Host 2Both OGLE's fit and theirs prefer a small nonzero black-hole abundance, with a best-fit mass around ten to the minus one point five solar masses and a fraction of order three times ten to the minus three, driven mainly by the shortest event of seven days; in the new models that preference is stronger, with a Δχ² statistic of about nine to ten compared with OGLE's three point seven.

Host 1Does that mean they found black holes?

Host 2No, and they do not say so. It is still the same thirteen events, and a stronger preference inside a model whose parameters were chosen to be lenient is not a detection.

Host 1What happened at the small planet-like masses, from the fast survey?

Host 2At two millionths of a solar mass, roughly two thirds of an Earth, OGLE's published limit is about an eighth of one percent of the dark matter. The benchmark weakens it by a factor of one point two and the stress test by two point four, because with no short events there, the limit simply follows how much dark matter lies along the sightlines.

Host 1And the clustered black holes in the summary, what does clustering mean here?

Host 2If primordial black holes formed in tight clumps instead of spread evenly, a survey could look through gaps between clumps and see nothing, even if they are plentiful. The paper models clusters thirty parsecs across and caps their size with the Lyman alpha forest, the pattern of absorption in distant quasar light that would show graininess from heavy clumps.

Host 1How much does clumping loosen things?

Host 2At planetary masses, the limit becomes about three point four times weaker than published for the benchmark, and six times for the stress test. For the twenty-year survey of heavier masses, clumping changes the limit by at most sixteen percent.

What changes and for whom

Host 1Why does any of this matter to people who are not black hole hunters?

Host 2Because these curves get copied. Other analyses treat OGLE's limit as a fixed fact, and this paper shows the same data supports a limit that is a factor of one point two to one point six looser under reasonable, recent galaxy models.

Host 1Does this rescue the idea that black holes are all the dark matter?

Host 2No. The authors say their likelihood, which keeps the measured durations, does not support loosening of the size the critics argued for, and even the stress test leaves black holes as a small fraction across most of this mass range.

Host 1So what are they actually asking others to do differently?

Host 2To quote microlensing limits together with their halo and star assumptions and say whether a curve is strict or relaxed. Where robustness matters, they offer their benchmark curves as a conservative alternative.

What it cannot tell us

Host 1What should we be careful not to read into this?

Host 2First, every halo parameter is fixed, not averaged over its uncertainty, and the values were chosen to suppress the signal, so the benchmark measures sensitivity, not a new lower envelope. Second, they assume all black holes have one mass; a spread of masses changes things in ways that need not weaken the limit.

Host 1Were there gaps in the data they had?

Host 2Two. They did not reanalyse OGLE's long-duration survey of the Small Cloud, and they had star brightness distributions for only one Large Cloud and one Small Cloud field, which they applied to all fields.

Host 1And the clustering results, are those firm?

Host 2They are the softest part. The microlensing and Lyman alpha limits are imposed separately rather than in one joint fit, they assume clusters formed in the early universe and are scattered independently, and using a budget from stellar streams would cut both to about a factor of two point four.

Host 1What about the short events toward the galaxy's centre that started this renewed interest?

Host 2The paper explicitly does not claim to reconcile with them. It mentions a separate study finding a statistical preference for an unexplained lens population toward the bulge, and treats it as an unconfirmed anomaly that could come from free-floating planets, brown dwarfs or binary lenses.

What to watch next

Host 1So what should listeners watch for next?

Host 2The authors want one joint analysis of the bulge and Magellanic Cloud samples, with a shared halo model and a shared black hole mass distribution. That would test the bulge hint and the empty Magellanic result in the same framework.

Host 1Will new telescopes help settle it?

Host 2In the wider picture, yes: the Vera Rubin Observatory's survey and NASA's Roman space telescope should find many more microlensing events. Better maps of the halo's shape from Gaia and stellar streams would narrow the very assumptions this paper shows matter.

Host 1If you had to put the whole thing in one sentence, what would it be?

Host 2Same thirteen events, same efficiencies, and a galaxy model built from recent measurements loosens the leading limit by up to about one point six, so black holes are still not most of the dark matter, but the exact few percent depends on the halo you assume.

Sign-off

Host 1That's the deep dive. Every weekend this same feed carries the roundup, the week's new papers in one short episode.