WEBVTT 1 00:00:03.439 --> 00:00:07.719 Welcome to Bedtime Astronomy. Explore the wonders of the cosmos 2 00:00:07.759 --> 00:00:12.279 with our soothing Bedtime astronomi podcast. Each episode offers a 3 00:00:12.359 --> 00:00:16.320 gentle journey through the stars, planets, and beyond, perfect for 4 00:00:16.399 --> 00:00:20.239 unwinding after a long day. Let's travel through the mysteries 5 00:00:20.239 --> 00:00:22.440 of the universe as you drift off into a peaceful 6 00:00:22.480 --> 00:00:23.800 slumber under the night sky. 7 00:00:26.960 --> 00:00:30.760 Welcome back today. We are strapping in for quite a journey. 8 00:00:30.960 --> 00:00:33.880 We're diving into a universe that doesn't actually exist in 9 00:00:33.920 --> 00:00:36.920 the sky, but inside a supercomputer. 10 00:00:37.119 --> 00:00:40.920 That's right, we're talking about the Flagship two Galaxy Mocks simulation. 11 00:00:41.479 --> 00:00:45.719 It's genuinely the largest, most detailed synthetic simulation of the 12 00:00:45.719 --> 00:00:47.039 Cosmos ever put. 13 00:00:46.880 --> 00:00:51.799 Together, an unprecedented scientific achievement, really, and it's absolutely foundational 14 00:00:51.840 --> 00:00:53.399 for what's coming next in astronomy. 15 00:00:53.719 --> 00:00:56.920 Absolutely, So our mission today is to unpack this thing, 16 00:00:56.960 --> 00:00:59.640 Flagship two. If you're say, prepping for a big meeting 17 00:00:59.679 --> 00:01:02.439 on where cosmology is headed, or maybe you just want 18 00:01:02.479 --> 00:01:05.239 to grasp how scientists are actually going to handle the 19 00:01:05.280 --> 00:01:06.879 insane amount of data coming. 20 00:01:06.680 --> 00:01:09.239 Our way, then this is the place to be. This simulation. 21 00:01:09.400 --> 00:01:11.439 It's not just some cool tech demo. 22 00:01:11.400 --> 00:01:14.159 No, not at all. It's the essential blueprint, the roadmap 23 00:01:14.359 --> 00:01:15.120 for discovery. 24 00:01:15.200 --> 00:01:17.920 And the scale of this blueprint, that's what really jumps 25 00:01:17.920 --> 00:01:20.799 out first. When we say largest, we mean well, utterly 26 00:01:20.879 --> 00:01:23.840 staggering numbers. We're talking about a digital universe with three 27 00:01:23.920 --> 00:01:27.439 point four billion simulated galaxies inside it. Just try to 28 00:01:27.480 --> 00:01:29.159 wrap your head around that amount of information. 29 00:01:29.319 --> 00:01:33.920 It's almost impossible. And scientists build this this simulated reality 30 00:01:34.120 --> 00:01:37.159 very deliberately to get ready for and then you know, 31 00:01:37.560 --> 00:01:41.680 ultimately interpret the flood of data that's coming from ESA's 32 00:01:41.680 --> 00:01:42.719 EUCLID mission. 33 00:01:42.480 --> 00:01:45.840 The European Space Agency's big eye in the sky exactly. 34 00:01:45.879 --> 00:01:48.200 And what's really critical here, you see, is that this 35 00:01:48.280 --> 00:01:52.319 simulation does two key jobs. It there's two masters if 36 00:01:52.359 --> 00:01:56.239 you like. Okay, On one hand, preparation, it lets the 37 00:01:56.280 --> 00:02:01.400 EUCLID team build and crucially test their analysis methods, their 38 00:02:01.400 --> 00:02:02.799 software pipelines. 39 00:02:02.439 --> 00:02:05.200 Right ironing out the bugs before the real data hits precisely. 40 00:02:05.680 --> 00:02:08.039 But on the other hand, there's a kind of prophecy aspect. 41 00:02:08.560 --> 00:02:11.680 Because the simulation is built on our current best understanding 42 00:02:11.719 --> 00:02:15.159 of the universe, the standard model of cosmology, it creates 43 00:02:15.199 --> 00:02:17.000 this perfect theoretical prediction. 44 00:02:17.199 --> 00:02:19.759 Ah, So it's like here's what the universe should look 45 00:02:19.800 --> 00:02:21.120 like according to our theories. 46 00:02:21.159 --> 00:02:24.240 See exactly, it builds the target that reality. The actual 47 00:02:24.319 --> 00:02:27.159 data from EUCLID has the other hit or miss. 48 00:02:27.439 --> 00:02:29.680 And as we dig into the sources talking about this work, 49 00:02:29.719 --> 00:02:31.879 you know, stuff from the EUCHID Consortium and especially these 50 00:02:31.919 --> 00:02:36.840 amazing algorithms from USh professor Yokenstatal in his team, we 51 00:02:36.960 --> 00:02:39.639 start to see the whole point is actually designed to 52 00:02:39.719 --> 00:02:44.000 find the flaws. They're actively looking for where reality might 53 00:02:44.080 --> 00:02:46.759 just shatter the theory. They call it searching for cracks 54 00:02:47.159 --> 00:02:48.199 in the standard model. 55 00:02:48.520 --> 00:02:51.400 It's a really sophisticated strategy. We're going to explore how 56 00:02:51.439 --> 00:02:56.560 they built this computational colossus, why it's basically the only 57 00:02:56.639 --> 00:03:00.479 way forward when you're facing this data avalanche, and maybe 58 00:03:00.520 --> 00:03:04.000 most excitingly, what kind of potential scientific revolution it could 59 00:03:04.039 --> 00:03:04.879 actually trigger. 60 00:03:05.319 --> 00:03:08.800 Okay, let's unpack this journey then, starting with just the 61 00:03:08.840 --> 00:03:11.560 mind bending scale and the cleverness needed to build this 62 00:03:11.680 --> 00:03:14.080 digital cosmos in the first place. You really have to 63 00:03:14.080 --> 00:03:17.240 start with the architect, don't we Behind this massive scale 64 00:03:17.319 --> 00:03:22.520 flagship two you mentioned Youakim Statle, the astrophysicist at UZH. 65 00:03:22.719 --> 00:03:25.919 He developed the algorithms that really underpin the whole simulation. 66 00:03:26.159 --> 00:03:28.759 That's the key, because when you're trying to model three 67 00:03:28.800 --> 00:03:32.719 point four billion objects, the complexity it just explodes. It's 68 00:03:32.719 --> 00:03:34.599 not linear, it's exponential. 69 00:03:34.759 --> 00:03:37.360 So it wasn't just about getting a bigger computer, not 70 00:03:37.439 --> 00:03:37.840 at all. 71 00:03:38.159 --> 00:03:41.360 It was about having the mathematical insight, the genius really 72 00:03:41.680 --> 00:03:45.800 to handle the gravitational interactions between all those objects efficiently. 73 00:03:45.319 --> 00:03:48.680 The famous n body problem, right, calculating how everything pulls 74 00:03:48.719 --> 00:03:50.199 on everything else exactly. 75 00:03:50.560 --> 00:03:54.319 And that's the core challenge in any cosmology simulation. Billions 76 00:03:54.479 --> 00:03:56.159 or in this case, trillions of particles. 77 00:03:56.199 --> 00:03:56.520 Wow. 78 00:03:56.560 --> 00:04:00.360 Previous singulations always had to compromise. Either you simulated a 79 00:04:00.360 --> 00:04:03.159 big volume with low detail or a small volume with 80 00:04:03.240 --> 00:04:03.879 high detail. 81 00:04:03.960 --> 00:04:05.159 You couldn't have both, right. 82 00:04:05.479 --> 00:04:09.479 Statle's work was about optimizing the calculation, finding clever ways 83 00:04:09.479 --> 00:04:12.560 to group particles and calculate forces so they could manage 84 00:04:12.560 --> 00:04:16.279 both scale and resolution. The raw computing power is kind 85 00:04:16.279 --> 00:04:19.879 of useless without the smart algorithms to distribute the work 86 00:04:19.959 --> 00:04:24.800 and calculate those dynamic interactions across huge cosmic distances effectively. 87 00:04:24.959 --> 00:04:27.720 And the result of that optimization is just wow. Let's 88 00:04:27.720 --> 00:04:30.759 say that number again, three point four billion galaxies. That 89 00:04:30.800 --> 00:04:33.240 gives you the size of the simulated patch of universey 90 00:04:33.279 --> 00:04:35.720 and the volume. But you pointed out the real complexity 91 00:04:35.759 --> 00:04:38.120 comes from the depths of the data for each one. 92 00:04:38.319 --> 00:04:41.079 What kind of details are we talking about for each galaxy? 93 00:04:41.399 --> 00:04:45.439 Oh, it's incredibly detailed. We're walking high resolution data design 94 00:04:45.839 --> 00:04:49.480 to mimic what the real EUCLID telescope will actually capture. 95 00:04:49.720 --> 00:04:52.519 Each of those three point four billion galaxies isn't just 96 00:04:52.560 --> 00:04:57.759 a point of light. Each one has four hundred modeled properties, four. 97 00:04:57.600 --> 00:04:59.879 Hundred four hundred properties. 98 00:04:59.399 --> 00:05:03.439 Per gal like what, well, the basics you'd expect brightness, 99 00:05:03.959 --> 00:05:06.959 it's position in three D space, it's velocity, how fast 100 00:05:06.959 --> 00:05:11.519 it's moving. But then crucial details like its metallicity, how 101 00:05:11.560 --> 00:05:14.560 many heavy elements it has. Okay, it's star formation history, 102 00:05:14.560 --> 00:05:17.160 when its stars were born, and critically for one of 103 00:05:17.160 --> 00:05:21.879 euclid's main goals, it's precise shape and orientation. It's ellipticity. 104 00:05:22.079 --> 00:05:25.279 Why is the shape the ellipticity so important that they'd 105 00:05:25.279 --> 00:05:28.759 spend all that computational effort modeling at three point four billion. 106 00:05:28.439 --> 00:05:31.199 Times ah because that's the bedrock of the dark matter 107 00:05:31.199 --> 00:05:34.120 mapping effort. We'll get more into this, but basically, EUCLID 108 00:05:34.199 --> 00:05:36.560 finds dark matter by looking for tiny distortions in the 109 00:05:36.560 --> 00:05:40.920 shapes of background galaxies caused by gravitational. 110 00:05:39.879 --> 00:05:42.759 Lenses right the like it's bent by unseen mass exactly. 111 00:05:43.079 --> 00:05:46.839 So, if your synthetic universe, your simulation doesn't have accurately 112 00:05:46.879 --> 00:05:50.680 modeled galaxy shapes to begin with, you can't properly train 113 00:05:50.800 --> 00:05:54.920 the software that's supposed to detect those lensing distortions in 114 00:05:55.000 --> 00:05:56.160 the real data, So. 115 00:05:56.120 --> 00:05:58.079 You wouldn't know if your software was working correctly. 116 00:05:58.160 --> 00:06:02.279 Precisely, you couldn't accurately and interpret the real observations. Those 117 00:06:02.319 --> 00:06:05.600 four hundred properties are what tie the simulation directly to 118 00:06:05.639 --> 00:06:09.439 the specific physics and the specific measurements EUCLID is designed to. 119 00:06:09.399 --> 00:06:12.319 Make okay and tracing back even further before they even 120 00:06:12.360 --> 00:06:16.199 put the galaxies in the underlying structure, the cosmic web, 121 00:06:16.279 --> 00:06:18.360 that came from an even bigger calculation, didn't it. 122 00:06:18.439 --> 00:06:22.279 Oh yes, before populating it with galaxies, the simulation first 123 00:06:22.279 --> 00:06:25.759 built the large scale structure, the scaffolding of the universe, 124 00:06:25.800 --> 00:06:29.279 you could say, the dark matter halos and the filaments. 125 00:06:28.839 --> 00:06:30.319 Connecting them, the cosmic web. 126 00:06:30.399 --> 00:06:32.879 That's it, and that structure was built by tracking the 127 00:06:32.879 --> 00:06:40.839 gravitational interactions. Wait for it, four trillion particles, four trillion trillon, 128 00:06:41.199 --> 00:06:45.079 four trillion individual points, each representing a chunk of mass 129 00:06:45.120 --> 00:06:47.920 pulling on every other chunk over billions of years of 130 00:06:47.959 --> 00:06:49.160 simulated cosmic time. 131 00:06:49.439 --> 00:06:53.000 That number just defines the sheer massive scope of that 132 00:06:53.120 --> 00:06:57.199 initial n body calculation. You're modeling every significant lump and 133 00:06:57.240 --> 00:07:00.480 bump in matter density across this huge volume of space, 134 00:07:00.879 --> 00:07:03.439 simulating gravity's effect over cosmic history. 135 00:07:03.519 --> 00:07:06.120 Think about the memory needed just to store the position 136 00:07:06.199 --> 00:07:10.000 and velocity of four trillion points at any given moment, let. 137 00:07:09.839 --> 00:07:12.920 Alone calculate all the forces between them constantly exactly. 138 00:07:12.959 --> 00:07:15.360 It's a staggering computational. 139 00:07:14.639 --> 00:07:18.839 Feat and managing that four trillion interacting particles in one 140 00:07:19.000 --> 00:07:24.800 giant calculation that's almost beyond imagining. It needed serious computing power, 141 00:07:24.879 --> 00:07:28.040 right institutional scale. Where did they actually do this? 142 00:07:28.600 --> 00:07:31.240 Yeah, this monster calculation was run back in twenty nineteen. 143 00:07:31.920 --> 00:07:34.600 They used the piz Dint supercomputer at the Swiss National 144 00:07:34.600 --> 00:07:37.680 Supercomputing Center CSCs and Lugano, Switzerland. 145 00:07:37.839 --> 00:07:39.560 PiZZ Date wasn't that one of the fastest in the 146 00:07:39.560 --> 00:07:40.759 world at the time It was. 147 00:07:40.800 --> 00:07:44.360 Indeed, in twenty nineteen, PiZZ Date was ranked the third 148 00:07:44.480 --> 00:07:46.879 most powerful supercomputer on the planet. 149 00:07:46.959 --> 00:07:48.959 Okay, and to give people context on what that means 150 00:07:48.959 --> 00:07:52.439 in terms of resources, the sources really highlight the commitment involved. 151 00:07:52.439 --> 00:07:55.319 This wasn't just like a big job they ran overnight. 152 00:07:55.199 --> 00:07:58.959 No far from it. Get this, more than eighty percent 153 00:07:59.160 --> 00:08:03.120 of PiZZ Dant's entire capacity was dedicated solely to this 154 00:08:03.160 --> 00:08:04.000 one calculation. 155 00:08:04.199 --> 00:08:09.040 Eighty percent. Imagine basically shutting down almost everything else happening 156 00:08:09.079 --> 00:08:12.800 at a major national supercomputing center just to run one simulation. 157 00:08:13.040 --> 00:08:15.720 It shows you how important this was and just how 158 00:08:15.800 --> 00:08:19.759 challenging it was computationally. Statle himself said, and I'm quoting here, 159 00:08:20.160 --> 00:08:22.639 it was a huge challenge to simulate such a large 160 00:08:22.680 --> 00:08:27.319 portion of the universe at this resolution in a single calculation. 161 00:08:26.839 --> 00:08:29.360 Because it had never been done before. They needed that massive, 162 00:08:29.480 --> 00:08:31.399 sustained dedication of resources. 163 00:08:31.439 --> 00:08:32.919 Absolutely, it pushed the limits. 164 00:08:33.000 --> 00:08:35.279 So it's important for people listening to understand this two 165 00:08:35.360 --> 00:08:40.159 step process. Right First, the pure physics track, those four 166 00:08:40.240 --> 00:08:44.279 trillion particles, let gravity shape the cosmic web over billions 167 00:08:44.320 --> 00:08:47.200 of years. That builds the underlying framework. 168 00:08:46.919 --> 00:08:49.960 Yes, the scaffolding, the dark matter distribution. That's step one. 169 00:08:50.240 --> 00:08:53.799 Step two is then more astrophysical and arguably more complex 170 00:08:53.840 --> 00:08:56.759 in some ways. Once you have those structures, the dark 171 00:08:56.799 --> 00:09:00.279 matter halos where galaxy should form the connecting filaments, you 172 00:09:00.320 --> 00:09:03.519 then have to populate them with synthetic galaxies. And these 173 00:09:03.519 --> 00:09:06.200 aren't just random points. They have to obey the known 174 00:09:06.240 --> 00:09:10.240 physical laws of how galaxies actually form and evolve, and critically, 175 00:09:10.480 --> 00:09:15.039 they have to mimic exactly what the EUCLID instruments will observe. 176 00:09:14.840 --> 00:09:17.600 So they look like real galaxies to the telescopes, cameras 177 00:09:17.600 --> 00:09:19.159 and spectrographs. 178 00:09:18.519 --> 00:09:21.919 Exactly complete with those four hundred properties for each of 179 00:09:21.919 --> 00:09:24.919 the three point four billion galaxies. This is what produces 180 00:09:24.919 --> 00:09:29.000 that realistic blueprint Statle talks about a simulation of what 181 00:09:29.080 --> 00:09:31.679 yuclind will actually see when it looks to the sky. 182 00:09:32.000 --> 00:09:36.279 So the gravitational physics creates the container, the large scale structure, 183 00:09:36.879 --> 00:09:39.480 and then the astrophysics fills that container with the right 184 00:09:39.559 --> 00:09:41.879 kinds of objects, the things EUCLID is actually looking for, 185 00:09:42.000 --> 00:09:46.080 the light sources. That interplay between modeling the invisible dark 186 00:09:46.120 --> 00:09:49.039 matter structures and the visible galaxies is really the genius 187 00:09:49.080 --> 00:09:50.000 of Flagship two. 188 00:09:50.240 --> 00:09:54.120 It bridges the gap between fundamental theory and actual observation, 189 00:09:55.000 --> 00:09:58.480 which brings us neatly to the crucial why why go 190 00:09:58.559 --> 00:10:01.600 to all this trouble? Why dedicate eighty percent of one 191 00:10:01.639 --> 00:10:04.639 of the world's fastest supercomputers to building a fake universe? 192 00:10:04.960 --> 00:10:07.639 And the answer really comes down to the sheer amount 193 00:10:07.679 --> 00:10:10.480 of data that the real EUCLID mission is generating. 194 00:10:10.679 --> 00:10:13.879 Exactly, let's quickly ground ourselves in what EUCLID is doing. 195 00:10:14.200 --> 00:10:17.519 It's an ESA space telescope launched in June twenty twenty 196 00:10:17.519 --> 00:10:20.600 three started its survey, and it's designed to look at 197 00:10:20.600 --> 00:10:23.200 a huge chunk of the sky over a third of 198 00:10:23.240 --> 00:10:27.840 the entire celestial sphere, with unprecedented resolution both in imaging 199 00:10:27.960 --> 00:10:28.960 and spectroscopy. 200 00:10:29.120 --> 00:10:33.840 That combination of vast area and sharp detail that's the 201 00:10:33.879 --> 00:10:36.720 source of the problem, isn't it the data logistics problem? 202 00:10:36.799 --> 00:10:39.559 It is. Julian Adamik, one of the key collaborators on 203 00:10:39.759 --> 00:10:43.039 Flagship two, puts it very clearly. He explains that EUCLID 204 00:10:43.039 --> 00:10:46.960 produces data in such sheer volume and speed that having 205 00:10:47.039 --> 00:10:51.240 humans analyze it manually is just completely impossible. 206 00:10:51.320 --> 00:10:53.320 Give us a sense of that impossibility. You mentioned three 207 00:10:53.360 --> 00:10:55.879 point four billion galaxies in the simulation right now. 208 00:10:55.799 --> 00:10:58.799 Imagine trying to process the real data for billions of 209 00:10:58.840 --> 00:11:02.799 galaxies and they have Hypothetically, a highly trained astrophysicist could 210 00:11:02.840 --> 00:11:05.679 analyze everything about one galaxy in just one. 211 00:11:05.519 --> 00:11:08.720 Second, which is ridiculously fast, totally unrealistic. 212 00:11:08.759 --> 00:11:11.960 Yeah, but even then it would take centuries of NonStop 213 00:11:12.000 --> 00:11:15.440 work just to get through the catalog once. And EUCLID 214 00:11:15.559 --> 00:11:18.960 is delivering this data continuously day after day. 215 00:11:19.120 --> 00:11:24.200 Wow, that's yeah, that's a staggering bottleneck. It almost feels 216 00:11:24.240 --> 00:11:27.320 a bit demoralizing. Scientists build this incredible instrument that produces 217 00:11:27.399 --> 00:11:29.720 data they know they can't possibly look at themselves. 218 00:11:30.039 --> 00:11:33.080 Well, it's just the focus. The human element moves away 219 00:11:33.159 --> 00:11:38.559 from clicking on individual galaxies towards designing the algorithms the 220 00:11:38.559 --> 00:11:41.559 automated systems that can handle the load. And this is 221 00:11:41.600 --> 00:11:45.759 why the mock data, the Flagship two simulation is absolutely crucial. 222 00:11:45.440 --> 00:11:48.559 Because you need something to test those algorithms on before 223 00:11:48.559 --> 00:11:51.200 the real, precious unique data arived. 224 00:11:51.279 --> 00:11:55.440 Exactly, you have to develop and rigorously test the methodology 225 00:11:55.840 --> 00:11:59.360 for interpreting these massive, complex data sets in advance. You 226 00:11:59.360 --> 00:12:01.399 get afford to wait for the real data stream to 227 00:12:01.519 --> 00:12:03.759 start and then begin figuring out how to process. 228 00:12:03.840 --> 00:12:05.200 We just fall hopelessly behind. 229 00:12:05.440 --> 00:12:08.919 Hopelessly you need to be ready to process it almost instantly, 230 00:12:09.120 --> 00:12:11.480 often within hours or days of it coming down from 231 00:12:11.480 --> 00:12:14.600 the telescope because the data pipelines have to run continuously 232 00:12:14.639 --> 00:12:15.200 to keep up. 233 00:12:15.600 --> 00:12:20.600 So the simulation Flagship two becomes the ultimate training ground. 234 00:12:21.200 --> 00:12:24.279 It's the perfect controlled environment for the AI and machine 235 00:12:24.320 --> 00:12:26.720 learning algorithms that have to do the heavy lifting. 236 00:12:26.879 --> 00:12:29.039 That's a great way to put it. If the simulation 237 00:12:29.200 --> 00:12:33.960 is this perfect theoretical universe based on known physics, scientists 238 00:12:33.960 --> 00:12:37.679 can feed that simulated data into their algorithms. 239 00:12:37.200 --> 00:12:40.679 And see if the algorithms correctly identify all three point 240 00:12:40.720 --> 00:12:45.240 four billion fake galaxies and measure their four hundred properties accurately. 241 00:12:45.399 --> 00:12:48.440 Yes, And just as importantly, see if the algorithms correctly 242 00:12:48.480 --> 00:12:51.759 identify and handle the errors and complexities they'll encounter in real. 243 00:12:51.720 --> 00:12:53.399 Data, like what kind of errors. 244 00:12:53.120 --> 00:12:56.759 Things like instrumental effects, noise cosmic rays hitting the detector, 245 00:12:57.320 --> 00:13:01.080 but also tricky astrophysical things like accurate measuring the shapes 246 00:13:01.080 --> 00:13:05.759 for lensing, or dealing with deeplending objects. Yeah, where two 247 00:13:05.919 --> 00:13:08.200 or more galaxies overlap in the image on the sky 248 00:13:08.600 --> 00:13:10.759 and the software has to figure out that they're separate 249 00:13:10.799 --> 00:13:13.120 objects and measure their properties individually. 250 00:13:13.440 --> 00:13:16.720 That's really hard, I can imagine. So if your mock 251 00:13:16.840 --> 00:13:21.559 data Flagship two is this perfect high resolution replica where 252 00:13:21.600 --> 00:13:24.240 you know the ground truth, you know exactly where every 253 00:13:24.279 --> 00:13:26.840 galaxy is and what its property should be, you know 254 00:13:26.879 --> 00:13:28.960 the right then you can run your software on it, 255 00:13:29.120 --> 00:13:31.879 see where it makes mistakes, and fine tune it until 256 00:13:31.879 --> 00:13:35.159 it reaches the accuracy and reliability needed for the real mission. 257 00:13:35.320 --> 00:13:38.320 Exactly, you calibrate your tools on the simulator before you 258 00:13:38.399 --> 00:13:39.399 use them on the real sky. 259 00:13:39.600 --> 00:13:42.399 And this also helps tackle that huge problem in astronomy 260 00:13:42.919 --> 00:13:46.960 systematic errors. When you look at the real universe and 261 00:13:47.039 --> 00:13:51.000 find something weird, how do you know if it's genuinely 262 00:13:51.080 --> 00:13:54.320 new physics or just some subtle bias in your telescope 263 00:13:54.440 --> 00:13:56.000 or your analysis software. 264 00:13:56.360 --> 00:14:00.200 Flagship two helps isolate that if your algorithm process is 265 00:14:00.200 --> 00:14:04.320 the perfect synthetic universe and consistently finds, say that it 266 00:14:04.399 --> 00:14:08.360 underestimates the distances to small, faint galaxies, then you know 267 00:14:08.399 --> 00:14:11.279 that bias comes from the algorithm itself or how it 268 00:14:11.360 --> 00:14:14.679 interacts with simulated instrument effects, not from some strange new 269 00:14:14.720 --> 00:14:17.879 property of the actual universe. You can then fix the 270 00:14:17.919 --> 00:14:20.759 code before you let it loose on the real euclid data. 271 00:14:20.960 --> 00:14:24.480 It really underscores how modern astronomy isn't just about pointing 272 00:14:24.600 --> 00:14:30.440 telescopes anymore. It's deeply intertwined with sophisticated data science, computer science, 273 00:14:30.559 --> 00:14:31.559 preemptive coding. 274 00:14:31.799 --> 00:14:35.399 Absolutely, the modern astronomer, especially on these big survey projects, 275 00:14:35.480 --> 00:14:38.159 is often as much a data pipeline manager and algorithm 276 00:14:38.159 --> 00:14:42.480 developer as they are a traditional observer. That simulation isn't illuxiery, 277 00:14:42.799 --> 00:14:46.000 it's a necessity. It's like a massive digital dress rehearsal, 278 00:14:46.360 --> 00:14:47.840 or maybe a digital safety net. 279 00:14:47.879 --> 00:14:50.399 Okay, now we get to what, for many people is 280 00:14:50.440 --> 00:14:54.320 the really exciting part, the potential scientific drama, because while 281 00:14:54.399 --> 00:14:59.320 Flagship two is fundamentally this incredible technical exercise in preparation. 282 00:14:59.360 --> 00:15:01.440 Data handling, algorithm testing, its. 283 00:15:01.360 --> 00:15:04.639 Ultimate scientific purpose is actually to challenge our current understanding 284 00:15:04.639 --> 00:15:07.759 of the universe, to actively seek out the flaws in 285 00:15:07.799 --> 00:15:08.720 our best theories. 286 00:15:09.039 --> 00:15:12.360 It really is a high stakes test, and it's essential 287 00:15:12.399 --> 00:15:15.559 for listeners to remember, as we discussed that Flagship two, 288 00:15:15.720 --> 00:15:19.120 the mock universe is built entirely on the foundation of 289 00:15:19.159 --> 00:15:21.240 the standard cosmological model, which. 290 00:15:21.080 --> 00:15:24.759 Is called Lambda CDM. Right, can we just quickly recap 291 00:15:24.799 --> 00:15:26.360 what that model actually assumes? 292 00:15:26.559 --> 00:15:29.639 Sure, land to CDM basically says the universe is made 293 00:15:29.679 --> 00:15:33.200 up of a few key ingredients governed by Einstein's general relativity. 294 00:15:33.799 --> 00:15:39.000 There's ordinary matter like us, stars, gas, but much more importantly, 295 00:15:39.039 --> 00:15:40.559 there's cold dark matter. 296 00:15:40.399 --> 00:15:43.039 CDM, invisible slow moving. 297 00:15:42.799 --> 00:15:47.039 Stuff exactly invisible, doesn't interact with light, moves relatively slowly, 298 00:15:47.399 --> 00:15:51.080 and its gravity dictates where structures like galaxies and galaxy 299 00:15:51.080 --> 00:15:54.519 clusters form. That's the CDM part, okay, and the lambda 300 00:15:54.679 --> 00:15:59.159 landis represents the cosmological constant. This is the simplest mathematical 301 00:15:59.200 --> 00:16:02.480 form of dark energy, an intrinsic energy of space itself 302 00:16:02.679 --> 00:16:05.159 that causes the expansion of the universe to accelerate. 303 00:16:05.320 --> 00:16:08.879 So dark mata pulls things together, dark energy pushes things apart, 304 00:16:08.919 --> 00:16:10.480 basically in simple terms. 305 00:16:10.559 --> 00:16:14.759 Yes, and this LAMB to CDM model, despite its weird ingredients, 306 00:16:14.799 --> 00:16:18.279 has been incredibly successful. It explains a vast range of 307 00:16:18.320 --> 00:16:22.159 observations from the cosmic microwave background radiation left over from 308 00:16:22.200 --> 00:16:25.080 the Big Bang right up to the large scale distribution 309 00:16:25.120 --> 00:16:26.480 of galaxies we see today. 310 00:16:26.720 --> 00:16:31.360 So Flagship two simulates a perfect LAMB to CDM universe, 311 00:16:31.919 --> 00:16:35.320 and the expectation as both statal and atomic mention is 312 00:16:35.320 --> 00:16:39.200 that euclid's actual observations will broadly speak and confirm the 313 00:16:39.240 --> 00:16:43.399 matter distribution predicted by Flagship two. That's the baseline, hope 314 00:16:43.639 --> 00:16:44.559 or expectation. 315 00:16:44.759 --> 00:16:48.519 That's the baseline. But science rarely progresses just by confirming 316 00:16:48.559 --> 00:16:51.480 what we already think we know. The real excitement lies 317 00:16:51.480 --> 00:16:55.000 in the potential for disagreement. Scientists are actively hoping to 318 00:16:55.039 --> 00:16:56.840 find places where the model breaks down. 319 00:16:57.120 --> 00:17:01.360 They explicitly anticipate surprises and on a expected discoveries. As 320 00:17:01.399 --> 00:17:04.079 the sources say, they seem quite convinced the model isn't 321 00:17:04.119 --> 00:17:04.680 the final word. 322 00:17:04.799 --> 00:17:08.359 Well, yes, because there are already hints tensions in existing data. 323 00:17:08.440 --> 00:17:12.279 Statle uses that striking phrase, we already see indications of 324 00:17:12.359 --> 00:17:13.680 cracks in the standard model. 325 00:17:13.799 --> 00:17:16.559 That sounds significant. What sort of cracks is he likely 326 00:17:16.599 --> 00:17:19.119 referring to? Where are these tensions showing up already? 327 00:17:19.599 --> 00:17:22.319 There are a few areas. Probably the most famous is 328 00:17:22.319 --> 00:17:25.759 the Hubble tension. Different methods of measuring the current expansion 329 00:17:25.839 --> 00:17:29.200 rate of the universe the Hubble constant are giving slightly 330 00:17:29.240 --> 00:17:31.079 but persistently different answers. 331 00:17:31.559 --> 00:17:35.079 Measuring it locally using supernovae gives one value. Measuring it 332 00:17:35.119 --> 00:17:37.480 based on the early universe light gives another. 333 00:17:37.400 --> 00:17:40.279 Exactly and they don't quite agree within their error bars. 334 00:17:40.680 --> 00:17:41.319 That's a crack. 335 00:17:41.519 --> 00:17:41.720 Yeah. 336 00:17:42.160 --> 00:17:45.720 There are also potential issues, though perhaps less statistically strong, 337 00:17:46.160 --> 00:17:49.559 with how dark matter seems to clump on smaller scales 338 00:17:49.599 --> 00:17:54.160 compared to LAMB to CDM predictions, or maybe slight inconsistencies 339 00:17:54.200 --> 00:17:55.960 in gravitational lensing measurements. 340 00:17:56.319 --> 00:18:00.079 So these are subtle discrepancies found by cobbling together data 341 00:18:00.319 --> 00:18:02.279 from different telescopes different methods. 342 00:18:02.440 --> 00:18:06.000 Right, And the hope or expectation is that EUCLID, with 343 00:18:06.119 --> 00:18:09.920 its huge uniform high precision survey covering billions of years 344 00:18:09.920 --> 00:18:11.319 of cosmic time in one. 345 00:18:11.119 --> 00:18:15.200 Go, might finally provide data clear enough and statistically powerful 346 00:18:15.279 --> 00:18:18.200 enough to show definitively whether these cracks are real and 347 00:18:18.240 --> 00:18:22.039 maybe reveal new phenomena that LAMB to CDM just cannot explain, 348 00:18:22.599 --> 00:18:26.200 things about the cosmic web structure or galaxy evolution that 349 00:18:26.279 --> 00:18:27.480 don't fit the simple. 350 00:18:27.319 --> 00:18:30.799 Rules that tension really defines the scientific knife edge. Here 351 00:18:31.200 --> 00:18:34.000 Atomic sums it up perfectly. It will be exciting to 352 00:18:34.079 --> 00:18:37.039 see whether the model holds up against euclid's high precision data, 353 00:18:37.279 --> 00:18:39.519 or whether we uncover signs of new shortcomings. 354 00:18:39.759 --> 00:18:43.519 So Flagship two acts as the perfect theoretical benchmark. It's 355 00:18:43.559 --> 00:18:48.519 the idealized prediction. EUCLID delivers the messy, complicated. 356 00:18:47.839 --> 00:18:52.160 Reality, and of that reality, the EUCLID data deviates significantly 357 00:18:52.200 --> 00:18:55.000 and systematically from the flagship two prediction. 358 00:18:55.160 --> 00:18:57.720 You can't just blame the simulation because you know exactly 359 00:18:57.759 --> 00:19:00.599 what physics went into it standard Lambda CDM precisely. 360 00:19:00.880 --> 00:19:03.400 The blame has to fall on the input physics. The 361 00:19:03.440 --> 00:19:06.200 assumptions of land to CDM must be wrong or at 362 00:19:06.279 --> 00:19:07.680 least incomplete. 363 00:19:07.119 --> 00:19:10.039 And that deviation that confirmed crack. That's where the scientific 364 00:19:10.079 --> 00:19:11.240 revolution would have to start. 365 00:19:11.519 --> 00:19:14.799 Absolutely. It would force cosmologists back to the drawing board. 366 00:19:15.119 --> 00:19:18.359 Do we need more complex form of dark energy? Does 367 00:19:18.440 --> 00:19:22.359 dark matter have some unexpected properties? Does gravity itself behave 368 00:19:22.400 --> 00:19:25.000 differently on cosmic scales than Einstein predicted? 369 00:19:25.240 --> 00:19:29.960 Maybe considering alternative theories like mond, modified Newtonian dynamics or 370 00:19:30.000 --> 00:19:31.599 other modified gravity ideas. 371 00:19:31.759 --> 00:19:37.200 Potentially Yes, the simulation, by providing that precise baseline, clarifies 372 00:19:37.319 --> 00:19:41.079 exactly where and by how much the standard model is failing. 373 00:19:41.480 --> 00:19:43.920 It points the way for the next generation of theories. 374 00:19:44.039 --> 00:19:47.960 Okay, let's zoom in on one of those potentially revolutionary areas, 375 00:19:48.119 --> 00:19:51.920 dark energy it's arguably the best mystery in cosmology. And 376 00:19:51.960 --> 00:19:54.720 within that standard LAMB to CDM model that Flagship two 377 00:19:54.799 --> 00:19:57.240 is built on, dark energy is represented in the simplest 378 00:19:57.240 --> 00:19:58.000 possible way. 379 00:19:57.880 --> 00:20:01.319 Right exactly as Statle puts it quite bluntly. In the model, 380 00:20:01.440 --> 00:20:05.680 dark energy is just a constant. It's the cosmological constant Lambda, 381 00:20:05.839 --> 00:20:10.359 meaning it's unchanging, unchanging in space, unchanging in time. It's 382 00:20:10.440 --> 00:20:14.279 treated as an intrinsic property of space itself, a constant 383 00:20:14.359 --> 00:20:17.880 energy density everywhere that doesn't dilute as the universe expands. 384 00:20:18.759 --> 00:20:22.000 It's the simplest mathematical term you can add to Einstein's 385 00:20:22.000 --> 00:20:24.200 equations to get accelerated expansion. 386 00:20:24.599 --> 00:20:27.279 But the whole point of EUCLID, one of its primary 387 00:20:27.359 --> 00:20:31.519 missions is to rigorously test whether that assumption holds true. 388 00:20:32.000 --> 00:20:34.240 Is it really just a constant? Why is finding that 389 00:20:34.279 --> 00:20:35.440 out so critical? 390 00:20:35.759 --> 00:20:37.759 Because if it's not a constant, if its strength or 391 00:20:37.799 --> 00:20:41.000 density has changed over cosmic history, then it's not just 392 00:20:41.079 --> 00:20:44.160 some background property of space time. It must be something dynamic, 393 00:20:44.200 --> 00:20:45.839 something physical that evolves. 394 00:20:45.519 --> 00:20:48.559 Which would completely change our picture of the universe's past 395 00:20:48.680 --> 00:20:50.279 and presumably its future fate. 396 00:20:50.480 --> 00:20:54.960 Absolutely, if it's a true constant Lambda, the acceleration continues relentlessly, 397 00:20:55.039 --> 00:20:57.359 leading eventually to a big freeze or maybe even a 398 00:20:57.359 --> 00:21:00.759 big rip where everything gets torn apart. But if dark 399 00:21:00.839 --> 00:21:04.680 energy is dynamic, if its strength changes, well, all bets 400 00:21:04.720 --> 00:21:07.160 are off. It could fade away, could strengthen, It could oscillate. 401 00:21:07.400 --> 00:21:11.000