WEBVTT 00:00.000 --> 00:02.880 Thank you to Incogni for supporting PBS. 00:02.880 --> 00:07.360 We spent the last decade listening to the  vanishingly faint vibrations in the fabric of 00:07.360 --> 00:14.240 spacetime from black holes colliding across the  universe. Now, hundreds of detections later, we 00:14.240 --> 00:19.520 finally thought we understood the story they were  telling. But then we spotted something that should 00:19.520 --> 00:24.960 be impossible. A black hole smaller than known  stellar astrophysics allows. A black hole that may 00:24.960 --> 00:31.120 have formed in the Big Bang itself. It may tell  us something incredible about our universe—but 00:31.120 --> 00:44.800 also equally importantly about what this young  field of gravitational wave astronomy has become. 00:39.612 --> 00:44.412 much as it's inspired us, check it out at the  Space Time merch store. Now on to the episode. 00:44.412 --> 00:48.492 Ten years ago, LIGO detected gravitational  waves for the first time. It was one of 00:48.492 --> 00:53.052 those discoveries that immediately felt like  the beginning of something much bigger. Yes, 00:53.052 --> 00:58.732 we’d proved Einstein right. Again. But we'd  also gained an entirely new way of observing 00:58.732 --> 01:03.612 the universe. What else would that lead to?  Entirely new types of collapsed object? Dancing 01:03.612 --> 01:10.332 cosmic strings? Echoes from the Big Bang? Even  cracks in the deeper depths of Einstein’s theory? 01:10.332 --> 01:15.692 Revolutionary observations are exciting because  we don’t know what the new field might evolve 01:15.692 --> 01:22.492 into. Now fast-forward ten years. We're finally  far enough away from that first detection to ask 01:22.492 --> 01:28.332 what gravitational-wave astronomy actually has  become. Has it lived up to those expectations? 01:28.332 --> 01:33.132 Has it changed astrophysics? Or has it mostly  confirmed what we already thought we knew? 01:33.132 --> 01:38.652 That question has been on my mind because  LIGO, Virgo and KAGRA released their latest 01:38.652 --> 01:44.172 gravitational-wave catalog—now nearly  400 detections of gravitational monsters 01:44.172 --> 01:49.052 colliding across the cosmos. But just as we  started to think we understood what we were 01:49.052 --> 01:55.292 seeing—building and refining our theories of  how they must have come to be… something new 01:55.292 --> 02:01.292 appeared. It's only a candidate for now, but  if it survives further analysis, it may contain 02:01.292 --> 02:08.252 a black hole that’s lighter than anyone thought  possible. A black hole that could NOT have formed 02:08.252 --> 02:14.892 from a star. It’s not confirmed, but people  are paying attention. We'll come back to it. 02:14.892 --> 02:20.492 Now, the first gravitational-wave detection was  a huge headline, and shortly thereafter a Nobel 02:20.492 --> 02:26.172 prize. Fair enough. The Laser Interferometer  Gravitational Wave Observatory and its cousins 02:26.172 --> 02:33.372 are, after all, are miracles of human ingenuity  and chutzpah. The kilometer-scale vacuum systems; 02:33.372 --> 02:37.932 the mirrors isolated from vibrations  smaller than the width of an atomic nucleus; 02:37.932 --> 02:44.172 sensitivity improving one observing run at a  time until esoteric tricks of quantum optics 02:44.172 --> 02:49.292 pushed the fundamental limits of measurement.  Thousands of scientists and engineers working 02:49.292 --> 02:54.492 together for decades to measure distortions  of spacetime of only about one part in 10²¹. 02:56.652 --> 03:01.852 Virgo and then KAGRA joined the two LIGO  detectors to improve our ability to localize 03:01.852 --> 03:09.212 sources on the sky. An ecosystem of fast-response  "traditional" observatories made ready to follow 03:09.212 --> 03:15.852 up gravitational wave hits. So much work by so  many people, even before the first detection—of 03:15.852 --> 03:23.132 black holes colliding a billion light years away. But then something more interesting happened. 03:23.132 --> 03:31.532 LIGO and its partners got to work. We’ve since  watched spacetime wobbles many hundreds of times 03:31.532 --> 03:38.732 and how have nearly confirmed 400 mergers of black  holes, with a few neutron stars in the mix. We 03:38.732 --> 03:46.332 quietly built something far more powerful than any  one measurement of black hole mergers. We built a 03:46.332 --> 03:54.972 catalog. And in that catalog lies a pattern.  And in that pattern we can infer a universe. 03:54.972 --> 04:00.732 An interesting thing about every successful  observatory is that it eventually fades away. 04:00.732 --> 04:06.332 And I mean that nobody asks anymore whether  the James Webb Space Telescope can see distant 04:06.332 --> 04:11.852 galaxies—we shifted to wondering at the galaxies  themselves—and then wondering what this tells us 04:11.852 --> 04:17.052 about structure formation in our universe.  We marveled that the Large Hadron Collider 04:17.052 --> 04:22.092 could make Higgs Bosons— but then we dug into  measuring the particle's properties—then asked 04:22.092 --> 04:28.252 what its existence tells us about the quantum  structure of reality. The instrument disappears 04:28.252 --> 04:32.412 and the universe takes center stage. The same thing has happened with 04:32.412 --> 04:36.812 gravitational-wave astronomy. Signals  that once would have been buried in the 04:36.812 --> 04:45.052 noise became routine detections. The catalog  grew—10 mergers, 50, 100—eventually there were 04:45.052 --> 04:50.092 too many to think about one at a time. And  I think that's the moment this field really 04:50.092 --> 04:56.252 crossed a threshold. The “discoveries” were  no longer individual events; the discovery 04:56.252 --> 05:03.212 was the population of events. And it’s in  this catalog that the real science happens. 05:03.212 --> 05:07.692 There's actually a name for the transition that  gravitational wave astronomy is going through. The 05:07.692 --> 05:13.212 philosopher of science Thomas Kuhn argued that  scientific revolutions eventually settle into 05:13.212 --> 05:18.652 what he called normal science. It’s a slightly  unfortunate name, because the most exciting 05:18.652 --> 05:25.052 science happens in the “normal” phase. In the  case of a revolutionary observational technique, 05:25.052 --> 05:30.972 Kuhnian normality is the moment a new instrument  becomes dependable enough that scientists stop 05:30.972 --> 05:35.692 asking whether it works, and start using  it to answer questions they couldn't even 05:35.692 --> 05:40.892 ask before. That's where gravitational-wave  astronomy is today. With a single detection, 05:40.892 --> 05:45.932 we can say that gravitational waves are real  and that black holes sometimes merge. And we 05:45.932 --> 05:50.892 can measure things about the masses and  the spins and their distance and whatnot. 05:50.892 --> 05:56.892 But with a catalog of detections we can  infer a huge hidden population of black 05:56.892 --> 06:01.372 hole pairs out there that spend billions  of years orbiting one another in complete 06:01.372 --> 06:06.892 darkness. And from the spacetime ripples  that wobble though the Earth from the 06:06.892 --> 06:12.572 last fraction of a second of their inspiral,  we can learn the properties of that whole 06:12.572 --> 06:18.012 hidden population. We learned that many  of these black holes are much heavier than 06:18.012 --> 06:22.572 we thought likely. That taught us that  there are unexpected growth mechanisms 06:22.572 --> 06:27.372 after black holes initial formation for  the collapse of massive stellar cores. 06:27.372 --> 06:29.532 From the black hole spins we inferred some really 06:29.532 --> 06:32.412 surprising things about the  way these merging black holes 06:32.412 --> 06:33.852 find each other. 06:33.852 --> 06:39.132 A little while ago, when we were talking about  the Dark Energy Survey, I said that reality 06:39.132 --> 06:44.412 often hides in the relationships between  data. That's what's happening here. The 06:44.412 --> 06:48.892 universe isn't revealing itself one merger  at a time. It's revealing itself through 06:48.892 --> 06:53.932 the patterns those mergers form together. The inference chain is pretty incredible 06:53.932 --> 07:00.652 if you think about it. Remember that LIGO and co  never measure any of the stuff we figure out. No 07:00.652 --> 07:06.732 ancient binary progenitor stars for example,  not even the binary black holes themselves. 07:06.732 --> 07:12.812 We measure a miniscule ripple in spacetime  lasting a fraction of a second. We get to binary 07:12.812 --> 07:18.972 black holes by asking “what sort of phenomenon  could have produced this?” And once we build the 07:18.972 --> 07:24.652 catalog—once we know the population properties—we  ask a different question: “What sort of universe 07:24.652 --> 07:31.132 could have given rise to such a population?”  The ripple remembers the merger and the merger 07:31.132 --> 07:36.972 remembers the parent stars. And with many ripples  we can infer all the way back to the universe that 07:36.972 --> 07:42.732 produced the stars that produced the merging black  holes and the faint signal that LIGO catches. 07:42.732 --> 07:47.692 All of astronomy works this way. It’s a  particularly awesome example of “reverse 07:47.692 --> 07:51.932 inference”, where we’re only given the  after-the-fact traces and need to infer 07:51.932 --> 07:56.652 the processes that led to them. Like a detective  figuring out a crime from a room full of subtle 07:56.652 --> 08:02.492 clues. And astronomy is especially dependent on  reverse inference because there’s no re-running 08:02.492 --> 08:08.012 the experiment. Other fields have the luxury  of regrowing the culture in a petri dish or 08:08.012 --> 08:14.092 re-smashing protons in the collider. Astronomy  gets a handful of ancient photons, neutrinos, 08:14.092 --> 08:18.812 gravitational waves, and has to figure out  the universe that created them. But before 08:18.812 --> 08:24.652 we can do that, we need this intermediate  step of the catalog. Of the population. Of 08:24.652 --> 08:32.092 what our universe normally makes. The work  then of normal science is to understand what 08:32.092 --> 08:37.612 is normal. So maybe it’s a good name after all. And the reason we care so much about the ordinary 08:37.612 --> 08:43.372 is that it allows us to prove ourselves wrong.  Attempting to falsify our models—to disprove 08:43.372 --> 08:48.972 our ideas about the universe is the highest  priority of science. And so a hallmark of a 08:48.972 --> 08:54.492 mature field of science is that it’s no longer  trying to construct the basic explanations—it’s 08:54.492 --> 09:01.212 trying to break them. Instead of asking “can I  explain this?”, a mature science asks "Does my 09:01.212 --> 09:07.772 explanation still survive the next observation?" Let’s take the example of the correlations between 09:07.772 --> 09:11.932 the spins of merging black holes. If we  assume that all merging black holes come 09:11.932 --> 09:16.732 from massive binary stars that formed from  the same cloud of interstellar dust then we 09:16.732 --> 09:21.692 have a prediction. The spins of those black  holes should both carry the signature of the 09:21.692 --> 09:28.012 original rotation of the birth cloud. The binary  black holes should have correlated spin. But when 09:28.012 --> 09:34.172 we look at hundreds of black hole mergers, we  find that indeed many have correlated spin, 09:34.172 --> 09:40.172 indicating formation by the expected process.  But there’s a second population in which the 09:40.172 --> 09:46.012 spins of the merging pair are uncorrelated with  each other—pointing in random directions. That 09:46.012 --> 09:52.732 falsifies the hypothesis that all mergers  come from binary stars. We now have a second 09:52.732 --> 09:58.732 population of black hole pairs that came together  after their progenitor stars died. In that case, 09:58.732 --> 10:04.092 they probably found each other after falling  towards the centers of dense star clusters. 10:04.092 --> 10:09.292 No individual merger points confidently to  one formation scenario or the other, but 10:09.292 --> 10:14.492 every detection nudges the weight of evidence in  a different direction. After hundreds of events, 10:14.492 --> 10:19.132 we start to be more confident that, across  the universe and back through its history, 10:19.132 --> 10:25.132 both of these processes played out. And that we  live in a universe in which both are possible. 10:25.132 --> 10:30.252 The growing catalog also hints that many black  holes are a bit too massive and spinning a bit 10:30.252 --> 10:38.412 too rapidly for either of the above scenarios,  allowing us to motivate another formation 10:38.412 --> 10:43.932 scenario. Some black holes probably sink  into the dense gas disks surrounding active 10:43.932 --> 10:49.052 supermassive black holes in galaxy centers.  There, the gas can shepherd them together, 10:49.052 --> 10:55.772 allowing repeated mergers that naturally produce  unusually massive, rapidly rotating black holes. 10:55.772 --> 10:59.212 This is still just a hypothesis—but one  that's increasingly consistent with the 10:59.212 --> 11:04.252 growing population of LIGO detections.  But again, the evidence is only in the 11:04.252 --> 11:09.292 population, not in any one observation. This is why gravitational wave astronomy 11:09.292 --> 11:14.732 is rapidly evolving into a mature science, The  individual discoveries are less important than the 11:14.732 --> 11:21.532 broader, statistical picture. The universe leaves  the most telling traces of its past in statistics, 11:21.532 --> 11:26.492 not in objects. A mature science interrogates  those statistics to interrogate its models 11:26.492 --> 11:33.612 of the universe that might explain them. And to  eliminate the models that fail. A mature science 11:33.612 --> 11:38.964 is in the business of eliminating explanations  until, gradually, the picture that remains starts 11:38.964 --> 11:42.092 looking less like one possible universe and  more like the universe we actually live in. 11:42.092 --> 11:48.892 Which brings us back to that one curious signal of  the impossibly light black hole. We don't directly 11:48.892 --> 11:54.092 measure the masses of each black hole—instead  we measure a particular combination of the two 11:54.092 --> 12:02.092 masses—the chirp mass. So on November 12th last  year, both LIGO detectors and Virgo quivered under 12:02.092 --> 12:08.332 an exceptionally faint spacetime ripple that  suggests an extraordinarily low chirp mass. 12:08.332 --> 12:13.612 If confirmed, it’s very likely that at least one  of the objects has less than the mass of the Sun. 12:13.612 --> 12:18.412 Such objects should be impossible by the  established modes of black hole formation. 12:18.412 --> 12:25.612 A dense stellar core of one solar mass does not  collapse into a black hole—it forms a white dwarf. 12:25.612 --> 12:30.332 But above the Chandrasekhar mass of 1.4  solar masses it becomes a neutron star, 12:30.332 --> 12:35.292 and only above 3 or so times the sun’s mass  can a dead stellar core collapse into a black 12:35.292 --> 12:40.732 hole. Now white dwarfs just aren’t compact  enough to produce a LIGO signal on merger, 12:40.732 --> 12:45.612 but neutron stars and black holes simply  can’t form at such a low mass—at least not 12:45.612 --> 12:52.572 via stellar evolution. But why did the field  have to be mature in order to spot this thing? 12:52.572 --> 12:57.772 Well imagine that the very first LIGO signal  had  indicated a sub-solar mass black hole merger. 12:57.772 --> 13:03.452 What would we have concluded? Well probably that  our instrument or our analysis method was flawed. 13:03.452 --> 13:09.452 We would have spent months or years tweaking  and recalibrating and hoping to catch another. 13:09.452 --> 13:15.612 Instead, this apparently-tiny black hole  merger came a decade in, when we know that 13:15.612 --> 13:21.452 LIGO reliably finds and measures appropriately  large masses for its black hole mergers. 13:21.452 --> 13:26.972 We know the instrument and analysis works.  So now, rather than stressing that we messed 13:26.972 --> 13:34.492 up somehow, we get to move forward with science.  LIGO has told us what is ordinary—and black 13:34.492 --> 13:40.732 hole pairs of 3 to 50 solar masses are routine.  Those are consistent with black holes formed in 13:40.732 --> 13:47.132 the collapse of the dead cores of massive stars. But a sub-solar mass black hole is NOT consistent 13:47.132 --> 13:53.452 with this picture. If confirmed, it will falsify  the hypothesis that all merging black holes come 13:53.452 --> 13:59.372 from stars. We’d need yet another formation  channel. If this thing is real, the best 13:59.372 --> 14:05.852 explanation seems to be that those black holes are  primordial. That they formed in the unimaginably 14:05.852 --> 14:11.532 dense conditions of the early universe, before  the first stars had even begun to shine. Finding 14:11.532 --> 14:17.292 primordial black holes with LIGO would allow  us to extend our reverse inference process all 14:17.292 --> 14:23.052 the way back to the moments after the Big Bang. That's an extraordinary possibility, and so also 14:23.052 --> 14:28.412 requires extraordinary evidence. More detailed  analysis is needed to ensure it isn’t a fluke of 14:28.412 --> 14:34.332 detector noise and that the inferred masses are  correct. This candidate may still disappear or 14:34.332 --> 14:40.252 never rise to the level of a confident detection.  That sounds disappointing, but it means the 14:40.252 --> 14:45.852 process of this now-mature science is working. Of course, if the sub-solar masses prove to 14:45.852 --> 14:51.212 be real then we have another mini-scientific  revolution. It starts with one detection—the 14:51.212 --> 14:56.732 first primordial black hole. Then we find another,  and another, and build the catalog and measure the 14:56.732 --> 15:02.012 population. From things like the number density  and the mass distribution we’ll learn how and when 15:02.012 --> 15:08.652 these PBHs formed, just as we did with the regular  black hole mergers. In the case of PBHs, we’d 15:08.652 --> 15:13.772 learn whether these things can explain dark matter  and derive powerful constraints on the conditions 15:13.772 --> 15:19.532 of the extremely early universe. Finding multiple  primordial black holes with LIGO would allow us to 15:19.532 --> 15:25.132 extend our reverse inference process all the  way back to the moments after the Big Bang, 15:25.132 --> 15:32.092 and PBH detection would become normal science.. Gravitational wave astronomy is right now 15:32.092 --> 15:37.852 transitioning into a mature field, at least  with respect to “normal” black hole mergers. 15:37.852 --> 15:44.972 Be excited when it uncovers extraordinary new  things, like this possible PBH. But be satisfied 15:44.972 --> 15:50.732 when it doesn’t, because this meticulous  mapping of the normal is the real work, 15:50.732 --> 15:56.492 and the key to an iteratively better  understanding of the whole of space time.