Deep divefrom Episode 9Cosmology5 November 202622 min

The Heaviest Black Hole Crash Had an Oval Orbit

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: Evidence for orbital eccentricity supports a hierarchical origin for GW231123 (Wang et al., arXiv preprint)

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, the heaviest pair of black holes ever caught colliding, and the question of whether their orbit was slightly oval.

Host 2The paper is Evidence for orbital eccentricity supports a hierarchical origin for G W two three one one two three, by Baoxiang Wang, Tao Yang and four colleagues from Wuhan University, Beijing Normal University, the Albert Einstein Institute in Potsdam and Ningbo University. It is a preprint on arXiv, so it has not yet been through peer review. Their claim is that this collision makes more sense if the orbit was not a perfect circle, and that this changes the story of where the black holes came from.

Why we listen for black holes

Host 1Before we get to the oval orbit, can you remind me what it actually means to catch two black holes colliding?

Host 2It means detecting gravitational waves, tiny ripples in space itself that spread out when very heavy objects whirl around each other and merge. To give the wider picture, Einstein predicted them in nineteen sixteen, and the two LIGO detectors in the United States first heard one in twenty fifteen, which earned a Nobel Prize two years later. Since then the field has catalogued a few hundred of these mergers.

Host 1How does a detector on Earth notice a ripple that started in some distant galaxy?

Host 2Each LIGO detector is an L-shaped instrument with arms four kilometres long, and laser light measures whether one arm stretches while the other squeezes. A passing wave changes the arm length by far less than the width of a proton, which is why the signals are buried in noise. That background is general to the field, not something this paper adds.

Host 1And the name G W two three one one two three, is that just a date?

Host 2Yes, it is gravitational wave, then the date: twenty-three November twenty twenty-three. The Hanford and Livingston detectors both caught it, and the LIGO, Virgo and KAGRA collaboration reported it as the most massive black hole merger seen so far, with a total mass of one hundred and ninety to two hundred and sixty-five times the mass of the Sun.

What looked strange about this event

Host 1Is a black hole a hundred times heavier than the Sun unusual, or are they all roughly that size?

Host 2It is unusual, and that is the point. In the standard analysis the bigger black hole came out at about one hundred and thirty-seven solar masses and the smaller at about one hundred and one. Both land in or at the edge of what astronomers call the pair-instability mass gap.

Host 1What is the pair-instability mass gap, in words someone without a physics degree can follow?

Host 2It is a range of masses where ordinary dying stars should not leave black holes behind. As the wider picture in astrophysics has it, the core of a very massive star gets so hot that light turns into pairs of electrons and positrons, pressure drops, the core collapses and a runaway explosion can blow the whole star apart, leaving nothing. The paper uses sixty to one hundred and thirty solar masses as its reference range for that gap.

Host 1So if stars cannot make black holes in that range, how does one end up there anyway?

Host 2The leading idea is a hierarchical merger: two smaller black holes merge, and the product later merges again. That needs a crowded place, such as a dense star cluster or the gas disc around a giant black hole at a galaxy's centre, where the merged black hole is not kicked out and can find a new partner.

Host 1Does a black hole made by an earlier collision carry some kind of fingerprint that tells you that is where it came from?

Host 2It carries a typical spin. When two black holes of similar mass with little spin of their own merge, the result spins at about zero point seven on a scale where one is the fastest a black hole can turn. The paper notes that this depends on the masses and spins going in, so it is a typical value, not a fixed one.

Host 1And what spins did the original analysis find for these two?

Host 2Very high ones: about zero point nine for the larger black hole and zero point eight for the smaller, near the maximum. That is higher than the zero point seven fingerprint, which made the event harder to fit into a simple story.

The question the paper asks

Host 1So the open question is whether those extreme spins are real, or whether something in the analysis pushed them up?

Host 2That is exactly how the authors frame it. The event lasted a remarkably short time in the detectors, with only a few wave cycles before the merger, so the results depend heavily on the mathematical template used to model the signal near the end.

Host 1What do you mean by a template here, and why would a short signal make it matter more?

Host 2Analysts compare the data against predicted waveforms, the exact wiggle pattern a merger with given masses, spins and orbit would produce, and keep whichever settings fit best. With many cycles, different effects leave distinct marks; with only a few, different physics can produce similar-looking wiggles.

Host 1Which two effects are we talking about getting mixed up?

Host 2Eccentricity and precession. Eccentricity means the orbit is an oval rather than a circle; precession means the spins are tilted so the orbit wobbles like a slowing top. Earlier studies have shown both distort the waveform's phase and amplitude, and over a handful of cycles they are hard to tell apart.

Host 1So if your template does not allow an oval orbit, the analysis might blame the oddness on spin instead?

Host 2That is the concern, and the paper cites prior work showing that leaving out non-circular motion can bias inferred spins. The standard analyses of this event assumed a quasi-circular orbit, meaning circular except for the slow shrinking as energy is radiated away.

Host 1Why would anyone expect the orbit to be oval in the first place?

Host 2Because of where the pair formed. Gravitational radiation rounds off an orbit over time, so a pair of stars that grew up together usually arrives circular. In crowded environments, captures and close encounters with third objects can produce pairs that are still oval when they enter the detectors' range, so eccentricity points to the same crowded places that make hierarchical mergers.

How they did it

Host 1How did they actually go about testing this, step by step?

Host 2They reanalysed the public data from both LIGO detectors with a waveform model called TEOB-ResumS Dalí, which can handle an oval orbit, tilted spins and the fainter higher harmonics of the signal at the same time. They ran it twice, once allowing eccentricity and once with eccentricity fixed at zero, keeping everything else identical. Then they compared how well each version explained the data.

Host 1You mentioned frequencies in the trailer, eleven to four hundred and forty-eight hertz. Why does the lower number matter so much?

Host 2Because eccentricity is largest early, at low frequencies, and shrinks roughly as an inverse power of frequency as the pair spirals in. Previous studies started at twenty hertz; this one used Hanford data down to eleven hertz, which keeps an earlier, more oval-looking stretch of the signal.

Host 1Was there a reason earlier studies stopped at twenty hertz, like a problem with the data down there?

Host 2Partly. The Livingston detector had a glitch, a burst of instrument noise, between ten and twenty hertz, so the authors kept Livingston at twenty hertz to be cautious, even though the collaboration found the glitch had no measurable effect. For Hanford they used the officially released version with a glitch subtracted, which passed the collaboration's checks against ordinary Gaussian noise.

Host 1So the low-frequency evidence comes from one detector only?

Host 2In the main analysis, yes, only Hanford contributes below twenty hertz. As a robustness check they also ran Livingston down to eleven hertz and got nearly the same eccentricity, but the headline numbers use the cautious setup.

Host 1How much data are we talking about in total, is it hours of recording?

Host 2Eight seconds per detector: six before the trigger and two after, sampled one thousand and twenty-four times a second. That shortness is part of why the modelling choices weigh so heavily.

Host 1And when they say they accounted for calibration uncertainty, what does that guard against?

Host 2Detectors do not report the wave's size and timing perfectly; there are small frequency-dependent errors in how the instrument's response is known. The authors let those errors float, using ten adjustable points per detector with ranges taken from the collaboration's released calibration data, so the fit cannot exploit a calibration flaw to fake an effect.

Host 1How do they decide which version, oval or circular, wins?

Host 2With a Bayes factor, which is the ratio of how probable the data are under each model, averaged over all the settings each model allows. A model with an extra knob gets penalised automatically if that knob does not earn its keep. They computed it with a standard sampling program called dynesty inside the Bilby software, using two thousand live points.

Host 1What range of eccentricity did they let the model consider before seeing the data?

Host 2Anything from zero to zero point five, all equally likely, with eccentricity defined at a reference frequency of five point five hertz. That reference sits at half the lowest analysed frequency so the higher harmonics are fully included.

What they found

Host 1So what number came out for how oval the orbit was?

Host 2An eccentricity of zero point three four at five point five hertz, with a ninety percent credible range from about zero point one one to zero point four six. Zero would be a perfect circle, so the range does not reach a circle, though its lower end comes close to it.

Host 1Is zero point three four a lot, or would I barely notice it if I could look at the orbit?

Host 2It is a clearly oval orbit, not a slight flattening, at the moment the signal was that low in frequency. By the time the pair reached higher frequencies it would have rounded off considerably, which is why the twenty hertz analyses see much less.

Host 1And how strongly does the data prefer the oval model over the circle?

Host 2The Bayes factor is ten point two in favour of eccentricity. That means the data are about ten times more probable under the eccentric model; by the usual conventions that counts as meaningful but moderate, not decisive.

Host 1What happens if you redo it the old way, starting at twenty hertz?

Host 2Then eccentricity at ten hertz comes out at zero point one two, with a range from about zero point zero one to zero point three nine, which is consistent with a circle. That matches earlier work, including an independent analysis with a different eccentric model that found zero point one zero.

Host 1So one way of looking says oval and the older way says possibly round, how do they know which one to believe?

Host 2They built a fake event. They took the best-fitting eccentric signal from their own analysis, with eccentricity zero point three zero, and ran it through both setups as if it were real, with no added noise. This is called an injection, and it shows what a truly oval signal of this kind would look like to each analysis.

Host 1And did the fake one behave like the real one?

Host 2Closely. Over eleven to four hundred and forty-eight hertz they recovered zero point three seven with a log Bayes factor of two point six one, against two point three two for the real data. Over twenty hertz and up, the same oval signal came back at only zero point one four, with plenty of room for a circle.

Host 1So if I have this right, a signal that genuinely is oval can look round if you only listen from twenty hertz upwards?

Host 2Yes, for a signal like this one. The authors also point out that earlier studies found genuinely eccentric short signals often do not produce a strong Bayes factor, so a moderate number is roughly what one would expect even if the orbit really is oval.

Host 1Now for the spins, which was the original puzzle. What happens to them once the oval orbit is allowed?

Host 2The smaller black hole's spin drops. With a circular orbit its median is zero point nine zero, range about zero point eight two to zero point nine eight; with eccentricity allowed the median is zero point seven four and the range widens from about zero point five one to zero point nine seven.

Host 1That lower range is enormous. Does that mean they now know the spin is about zero point seven, or that they know less?

Host 2Mostly that they know less, and the authors say so: the distribution extends towards much smaller spins, while its median moves near zero point seven. The firm conclusion is that the strong preference for a near-maximal spin disappears, not that the spin is pinned at zero point seven.

Host 1Can they show that ignoring the oval orbit is what inflated the spin, rather than just two models disagreeing?

Host 2That is the second use of the fake event. When they analysed the oval injection with the circular template, its secondary spin shifted upward by about zero point one seven and looked much like the real event's circular result. So leaving eccentricity out can by itself produce the high spin seen before.

Host 1And the bigger black hole, does its spin come down too?

Host 2No, it stays high under both models, around zero point nine three. In the injection, switching models moved it by only about zero point zero two, compared with zero point one seven for the smaller one.

Host 1Why would one black hole respond and the other not?

Host 2The authors offer an interpretation rather than a proof: the primary's spin is already pressed against the upper limit, so it has little room to absorb extra structure in the signal, while the secondary's spin can move. They are careful to call it a natural interpretation.

A word from our sponsor

Host 1Before we get to what all this means, a word from our sponsor?

Host 2The Hastrup Review of Science is brought to you by epilators. An epilator removes hair at the root, so the result persists well beyond the reference frequency of an ordinary razor.

Host 1Does it work on someone who has only a few cycles of hair?

Host 2Results are consistent with no hair, within the ninety percent credible interval. Epilators are sold on Amazon at ordinary prices, and the link is in the show notes.

Testing the result

Host 1Back to the black holes. Did they check whether something else could be distorting the signal, not just an oval orbit?

Host 2They checked gravitational lensing, which several groups have proposed for this event. As the field generally describes it, a star or other compact mass between us and the source can bend and slightly reshape a gravitational wave, and here they modelled a single point-like lens and fitted it alongside everything else.

Host 1And did adding a possible lens make the oval orbit go away?

Host 2No. With lensing allowed the eccentricity was zero point three two and the log Bayes factor two point two seven, both essentially unchanged, while the lens's mass and position stayed spread across their whole allowed range, meaning the data said little about them. The authors suggest some earlier support for lensing might partly reflect missing source physics such as eccentricity, but they offer that as a possibility.

Host 1Were there other checks on the method itself, things like how they estimated the noise?

Host 2Yes. They estimated detector noise from data away from the event, then repeated the analysis with noise estimates built the collaboration's way extended down to eleven hertz, and the eccentricity stayed consistent. The fake event also matched the real one across mass, mass ratio and distance, not only eccentricity and spin.

Host 1Why does it matter that the fake event had no noise or glitches in it?

Host 2Because it shows the pattern, oval at low frequencies, round-looking above twenty hertz and inflated spin under a circular template, can come from a genuinely oval signal alone. It does not depend on a particular stretch of detector noise, though it also cannot rule out that real noise contributed.

What it means

Host 1So what does this change about where these black holes came from?

Host 2It suggests a simpler story for the smaller one. They estimated the heaviest black hole a single earlier merger could produce from two ordinary star-made black holes, using a lower gap edge of about forty-three solar masses from an independent study that excluded this event, and five percent of the mass lost as gravitational waves. That comes to roughly eighty-two solar masses.

Host 1Does the smaller black hole fit under that limit?

Host 2Better than before: the share of its mass estimates below the limit is about two point two times larger under the oval model than the circular one. With a spin near zero point seven and a mass in the gap, it looks like a second-generation black hole, the product of one previous merger.

Host 1And the bigger one, can it be explained the same way?

Host 2No. Only about one in a thousand of its mass estimates fits under that limit, and its spin stays high. The authors conclude it formed either above the gap or through a more complicated chain of mergers, so the two black holes may have different histories even if they met in the same crowded place.

Host 1An earlier study said both came from about ten smaller black holes merging step by step. Does this paper say that is wrong?

Host 2It says the eccentric reading motivates a simpler, asymmetric picture instead of extended growth for both. That earlier work relied on circular-orbit estimates, so the two are not directly contradicting each other's data, only each other's assumptions.

Host 1Who besides black hole specialists should care whether this one orbit was oval?

Host 2People measuring where the mass gap lies. A recent study found this event essential for pinning the gap's upper edge; if its smaller black hole is a merger product rather than a star born above the gap, it cannot by itself set that edge. The general principle the authors draw is that when a boundary rests on a few unusual events, the modelling assumptions for those events directly shape the boundary.

Host 1And for an ordinary listener, is there any practical consequence here?

Host 2None directly; it is about how the universe builds its largest stellar black holes. The transferable lesson is methodological: a short, strange signal analysed with a too-simple template can return confident but wrong-looking numbers.

The honest limits

Host 1What should we be careful not to conclude from this?

Host 2That the orbit has been proven oval. A Bayes factor of ten is moderate, it rests on one event, and the eccentricity range runs from about zero point one one to zero point four six.

Host 1Since precession and eccentricity can mimic each other, could the reverse be happening, with tilted spins being misread as an oval orbit?

Host 2That risk is real and the paper cites work showing precession can be misinterpreted as eccentricity. Their template includes both effects at once, which is the right way to address it, but with so few cycles the separation still depends on how accurately one model captures both near the merger.

Host 1Is it a problem that everything hangs on one waveform model?

Host 2It is a limitation. At twenty hertz an independent model agrees with them, but nobody has yet run a second eccentric, precessing model on the eleven hertz Hanford data, and models are known to differ for this event.

Host 1And the low-frequency stretch that carries the evidence comes from one detector, using cleaned data?

Host 2Yes, the cleaned Hanford data, the one released and validated by the collaboration. The Livingston-to-eleven-hertz check was reassuring, but that is the part of the data a sceptic will examine first.

What to watch

Host 1What would settle this one way or the other in the next few years?

Host 2First, other groups repeating the eleven hertz analysis with different eccentric, precessing models. Second, more events like this one: as the wider field expects, detectors with better sensitivity at low frequencies will catch more cycles of each merger, which is exactly what separates eccentricity from precession.

Host 1And if more heavy mergers turn out to be oval, what would that tell astronomers?

Host 2It would strengthen the case that the heaviest black holes are built in crowded environments by repeated mergers. It would also push the field to include eccentricity as standard when estimating spins for short, massive signals.

Host 1So in one breath, what is the take-home from this paper?

Host 2Analysed from eleven hertz with a model that allows an oval orbit, G W two three one one two three moderately prefers an eccentricity of about zero point three four, and the smaller black hole's extreme spin softens towards the value expected from an earlier merger. It is a coherent story for one event, from one model, and it now needs independent checks.

Sign-off

Host 1That is the deep dive: Evidence for orbital eccentricity supports a hierarchical origin for G W two three one one two three, by Wang, Yang and colleagues, a preprint on arXiv. The link is in the show notes.

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.