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 Astronomie 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 --> 00:00:23.839 slumber under the night sky. 7 00:00:26.960 --> 00:00:29.600 Imagine standing in front of a mirror, just, you know, 8 00:00:29.920 --> 00:00:31.120 looking at your normal. 9 00:00:30.800 --> 00:00:33.119 Reflection, right, just checking your hair, yeah. 10 00:00:32.880 --> 00:00:36.920 Exactly, adjusted to your collar. But suddenly the glass shifts, 11 00:00:37.399 --> 00:00:41.119 and instead of just showing your face, the reflection starts playing. 12 00:00:40.799 --> 00:00:42.719 Backwards, like a movie on rewind. 13 00:00:43.000 --> 00:00:45.520 Yes, and not just by a few hours or like 14 00:00:45.640 --> 00:00:49.679 a few days, yep. Imagine the reflection rewinds faster and faster, 15 00:00:49.960 --> 00:00:51.600 dissolving the room right behind. 16 00:00:51.359 --> 00:00:53.320 You, just completely falling away, right. 17 00:00:53.240 --> 00:00:58.320 The walls fall away, the city outside dissolves into this wilderness, 18 00:00:58.880 --> 00:01:03.280 and the earth itself literally unforms. It turns back into 19 00:01:03.280 --> 00:01:05.840 a swirling disc of dust. The sun breaks apart. 20 00:01:05.959 --> 00:01:07.879 You're going back to the very beginning. 21 00:01:07.680 --> 00:01:10.959 All the way back. The mirror pulls back, accelerating through 22 00:01:11.040 --> 00:01:15.239 deep time, billions of years, just collapsing in seconds, retreating 23 00:01:15.239 --> 00:01:19.439 to a time of absolute, just unimaginable density. Everything you 24 00:01:19.519 --> 00:01:22.680 ever known shrinks into a blinding, searing point of heat, 25 00:01:22.879 --> 00:01:24.760 and then the mirror just pauses. 26 00:01:24.920 --> 00:01:26.239 It hits play, hits play. 27 00:01:26.719 --> 00:01:29.920 You are suddenly watching the entire thirteen point eight billion 28 00:01:29.959 --> 00:01:32.840 year history of the cosmos play out and fast forward 29 00:01:32.920 --> 00:01:33.760 right there in the glass. 30 00:01:33.920 --> 00:01:38.959 Which I mean, it sounds like pure unadulterated science fictionally, 31 00:01:39.200 --> 00:01:43.040 but the crazy thing is humanity is essentially building that 32 00:01:43.159 --> 00:01:44.400 exact mirror right now. 33 00:01:44.760 --> 00:01:47.799 It's wild to even think about. We are currently undertaking 34 00:01:47.840 --> 00:01:52.120 one of the most ambitious computational projects ever attempted in 35 00:01:52.159 --> 00:01:55.480 the history of our species. Oh, without a doubt, recreating 36 00:01:55.519 --> 00:01:58.359 the formation of the universe itself from its very first 37 00:01:58.359 --> 00:01:59.640 breath inside. 38 00:02:01.040 --> 00:02:03.359 The sheer audacity of the project is what makes it 39 00:02:03.400 --> 00:02:07.359 so compelling, honestly, because when we talk about these computational reconstructions, 40 00:02:07.359 --> 00:02:09.840 we really have to strip away the idea that these 41 00:02:09.840 --> 00:02:12.520 are just high definition animations. 42 00:02:12.560 --> 00:02:14.360 No, it's not a Pixar movie exactly. 43 00:02:14.639 --> 00:02:17.599 We aren't painting pretty pictures of stars and galaxies on 44 00:02:17.639 --> 00:02:19.960 a digital canvas just so someone can use them as 45 00:02:20.000 --> 00:02:21.039 a desktop wallpaper. 46 00:02:21.080 --> 00:02:22.520 Though they do make great wallpapers. 47 00:02:22.560 --> 00:02:28.039 They do, But we are constructing immense dynamic physical laboratories 48 00:02:28.080 --> 00:02:32.639 in silicon. These are mathematical reconstructions built rigorously, I mean 49 00:02:32.840 --> 00:02:34.439 line of code by line of code, from the. 50 00:02:34.439 --> 00:02:36.240 Ground up, driven by the actual math. 51 00:02:36.599 --> 00:02:42.400 Yes, they are governed by the precise, unrelenting equations of gravity, thermodynamics, 52 00:02:42.479 --> 00:02:47.199 radiation transfer, fluid motion, all of it cosmic expansion. 53 00:02:47.840 --> 00:02:51.960 So by taking the absolute limits of observational astronomy infusing 54 00:02:51.960 --> 00:02:55.919 them with the bleeding edge of supercomputing architecture, we're attempting 55 00:02:55.960 --> 00:02:59.439 to mathematically weave reality together just to see if we 56 00:02:59.479 --> 00:03:03.080 trulyunderstand how it works exactly. So today we are going 57 00:03:03.120 --> 00:03:07.400 to explore exactly how scientists are building these dynamic reconstructions 58 00:03:07.439 --> 00:03:11.000 of reality. You listening to this right now are about 59 00:03:11.000 --> 00:03:13.879 to see how the invisible forces that shape literally everything 60 00:03:13.879 --> 00:03:17.680 around us are being mapped and tested and ultimately understood. 61 00:03:17.800 --> 00:03:19.560 It's a huge undertaking, it really is. 62 00:03:19.680 --> 00:03:21.560 We're going to dive into the actual mechanics of how 63 00:03:21.560 --> 00:03:24.400 you take the raw, chaotic math of the cosmos and 64 00:03:24.439 --> 00:03:27.960 turn it into a breathing, evolving universe inside a machine. 65 00:03:28.319 --> 00:03:31.120 Because honestly, there's a profound philosophical weight. 66 00:03:31.000 --> 00:03:33.520 To this there is. It forces a real reckoning with 67 00:03:33.560 --> 00:03:35.439 our own place in the grand scheme of things. I mean, 68 00:03:35.479 --> 00:03:38.759 think about it. We are a temporary biological species residing 69 00:03:38.840 --> 00:03:41.879 on a tiny rock, a very tiny rock, attempting to 70 00:03:41.879 --> 00:03:46.120 mathematically model our own origins. We are building virtual universes 71 00:03:46.159 --> 00:03:49.039 to understand the singular physical universe that gave. 72 00:03:48.919 --> 00:03:50.879 Birth to us, which is just wow. 73 00:03:51.080 --> 00:03:53.639 It forces us to confront the deepest questions about the 74 00:03:53.719 --> 00:03:59.680 nature of reality and how unbelievable complexity like galaxies, stars, planets, biology, 75 00:04:00.039 --> 00:04:04.360 consciousness itself, how all of that emerges from incredibly simple 76 00:04:04.439 --> 00:04:05.479 initial conditions. 77 00:04:05.599 --> 00:04:07.960 But let's look at the immense roadblock right at the 78 00:04:08.000 --> 00:04:11.919 starting line. Because the fundamental challenge of understanding the cosmos, 79 00:04:12.000 --> 00:04:15.199 and therefore, you know, the very first challenge of simulating it, 80 00:04:15.240 --> 00:04:16.279 is deeply frustrated. 81 00:04:16.319 --> 00:04:17.399 It is a massive headache. 82 00:04:17.480 --> 00:04:20.519 It's the fact that the vast, overwhelming majority of the 83 00:04:20.600 --> 00:04:25.120 universe is completely utterly invisible to us. The stuff we know, 84 00:04:25.920 --> 00:04:29.480 ordinary matter, the atoms that make up the stars, the planets, 85 00:04:29.519 --> 00:04:31.920 the glowing gas clouds, the device you're listening to this 86 00:04:32.000 --> 00:04:35.680 on in our own biological bodies, that constitutes just a 87 00:04:35.800 --> 00:04:38.240 tiny almost negligible fraction of. 88 00:04:38.160 --> 00:04:40.040 Cosmic reality around five percent. 89 00:04:40.199 --> 00:04:41.519 Five percent. That is nothing. 90 00:04:41.839 --> 00:04:44.079 It really isn't, which means if you want to build 91 00:04:44.079 --> 00:04:48.079 a universe in a computer, you absolutely cannot start with ordinary. 92 00:04:47.639 --> 00:04:49.279 Matter because it just wouldn't work right. 93 00:04:49.519 --> 00:04:52.560 If you build a universe using only atoms, your simulation 94 00:04:52.680 --> 00:04:56.000 will fail catastrophically. It won't look anything like the night sky. 95 00:04:56.639 --> 00:05:00.879 The universe is completely dominated by two deeply mysterious, totally 96 00:05:00.920 --> 00:05:05.439 invisible components, dark matter and dark energy, the Big Two. 97 00:05:05.639 --> 00:05:10.199 The Big Two. Dark matter acts as the vital gravitational scaffolding. 98 00:05:10.879 --> 00:05:14.240 It is the structural foundation that actually allows galaxies and 99 00:05:14.360 --> 00:05:16.759 massive galaxy clusters to form in the first place. 100 00:05:16.800 --> 00:05:17.680 And then dark energy. 101 00:05:17.800 --> 00:05:21.560 Dark energy is the enigmatic force driving the accelerating expansion 102 00:05:21.560 --> 00:05:22.560 of space itself. 103 00:05:22.800 --> 00:05:24.199 And we can't see either of them. 104 00:05:24.360 --> 00:05:26.360 No, we cannot hold either of these in a jar. 105 00:05:26.600 --> 00:05:29.079 We don't even have a standardized particle model for dark 106 00:05:29.120 --> 00:05:34.480 matter yet. But their silent, invisible influence dictates absolutely everything, 107 00:05:34.680 --> 00:05:37.839 both the structure and the ultimate destiny of the cosmos. 108 00:05:38.040 --> 00:05:40.000 So I want to focus on the mechanics of getting 109 00:05:40.040 --> 00:05:44.639 this started. If you are a computational astrophysicist, sitting at 110 00:05:44.639 --> 00:05:48.439 a supercomputer terminal and you want to simulate this invisible universe. 111 00:05:48.920 --> 00:05:49.959 Where do you even begin? 112 00:05:50.199 --> 00:05:52.680 You definitely don't just write a command prompt that says 113 00:05:52.800 --> 00:05:54.959 run universe dot xe and walk away wish. 114 00:05:55.120 --> 00:05:55.600 Is that easy? 115 00:05:55.720 --> 00:05:57.800 Me too? But no, you have to give the computer 116 00:05:57.879 --> 00:06:02.399 a highly specific starting state, initial conditions, and those initial 117 00:06:02.439 --> 00:06:05.720 conditions are drawn from something called the cosmic microwave background 118 00:06:05.879 --> 00:06:06.600 or the CMB. 119 00:06:07.040 --> 00:06:09.800 Right, So, the CMB is what like the baby picture 120 00:06:09.839 --> 00:06:10.480 of the universe. 121 00:06:10.519 --> 00:06:15.680 Basically, it is the Rosetta stone of cosmology. Massive cosmological 122 00:06:15.680 --> 00:06:19.519 simulations must begin with the conditions inferred from that early universe. 123 00:06:20.600 --> 00:06:25.240 The cosmic microwave background is the ancient relic radiation left 124 00:06:25.240 --> 00:06:28.000 over from the Big Bang. So it's light, yes, yes, 125 00:06:28.240 --> 00:06:31.920 specifically it is the light emitted roughly three hundred and 126 00:06:32.000 --> 00:06:35.519 eighty thousand years after the Big Bang during an event 127 00:06:35.560 --> 00:06:36.800 called recombination. 128 00:06:37.079 --> 00:06:38.519 Recombination, what's happening there? 129 00:06:38.680 --> 00:06:41.240 That's when the universe finally cooled down enough for the 130 00:06:41.279 --> 00:06:45.519 first stable atoms to form, which meant light could finally 131 00:06:45.639 --> 00:06:49.439 travel freely through space without just smashing into a dense 132 00:06:49.519 --> 00:06:50.600 soup of particles. 133 00:06:50.839 --> 00:06:52.800 Ah, so it cleared up exactly. 134 00:06:53.040 --> 00:06:56.600 It's the oldest light we can possibly see, and preserved 135 00:06:56.639 --> 00:06:59.920 within that ancient light kind of like insects trapped in amber, 136 00:07:00.439 --> 00:07:05.399 are tiny, almost imperceptible density fluctuations from when the universe 137 00:07:05.480 --> 00:07:06.759 was extremely young. 138 00:07:06.959 --> 00:07:09.720 Let's talk about the scale of those fluctuations, because it's 139 00:07:09.759 --> 00:07:12.120 really hard to wrap your head around. When we say tiny, 140 00:07:12.439 --> 00:07:13.560 we really mean miniscus. 141 00:07:13.600 --> 00:07:14.519 It's absolutely tiny. 142 00:07:14.759 --> 00:07:17.759 The differences in density across that early universe were smaller 143 00:07:17.800 --> 00:07:19.480 than one part in one hundred thousand. 144 00:07:19.560 --> 00:07:20.600 Yeah, just barely there. 145 00:07:21.040 --> 00:07:25.879 Imagine a perfectly smooth, vast ocean and the only waves 146 00:07:25.879 --> 00:07:27.480 on it are microscopic ripples. 147 00:07:27.600 --> 00:07:28.399 Great analogy. 148 00:07:28.959 --> 00:07:32.160 Yet those incredibly faint quantum fluctuations, which were blown up 149 00:07:32.199 --> 00:07:35.120 to macroscopic size during an early period of rapid inflation, 150 00:07:35.519 --> 00:07:38.279 those are the seeds of absolutely everything that exists today. 151 00:07:38.079 --> 00:07:40.079 Every galaxy, every star, You and me. 152 00:07:40.439 --> 00:07:43.519 We all trace our physical lineage back to those microscopic 153 00:07:43.560 --> 00:07:44.839 density irregularities. 154 00:07:45.079 --> 00:07:48.399 And the physics of why those specific tiny seeds grew 155 00:07:48.759 --> 00:07:52.680 is the entire foundation of the simulation. Gravity is the 156 00:07:52.879 --> 00:07:57.000 ultimate patient sculptor here, how so well, under the influence 157 00:07:57.000 --> 00:08:00.680 of gravity, those slightly denser regions of that prime soup. 158 00:08:01.199 --> 00:08:03.319 The regions that were just a fraction of a percent 159 00:08:03.399 --> 00:08:07.399 denser than the average, they exerted a slightly stronger gravitational 160 00:08:07.480 --> 00:08:10.040 pull than the less dense regions around them. 161 00:08:10.079 --> 00:08:12.879 Because more mass equals more gravity exactly. 162 00:08:13.360 --> 00:08:16.879 So what happens. They gradually attract more matter. It is 163 00:08:16.920 --> 00:08:20.560 a compounding runaway effect. Over hundreds of millions and then 164 00:08:20.759 --> 00:08:25.079 billions of years, those one in one hundred thousand irregularities 165 00:08:25.160 --> 00:08:28.160 undergo what we call nonlinear gravitational collapse. 166 00:08:28.279 --> 00:08:30.439 Nonlinear gravitational collapse, right. 167 00:08:30.360 --> 00:08:32.559 They draw in more and more dark matter, pulling it 168 00:08:32.600 --> 00:08:34.840 out of the surrounding boyds and weaving it into these 169 00:08:35.039 --> 00:08:36.399 enormous DNSE structures. 170 00:08:36.559 --> 00:08:39.600 I always think of the CMB as this blurry static 171 00:08:39.720 --> 00:08:42.799 ultrasound of the infant universe. It gives us that initial snapshot, 172 00:08:43.120 --> 00:08:46.639 and the dark matter acts like an invisible, ghostly skeleton 173 00:08:47.360 --> 00:08:48.639 growing out of that snapshot. 174 00:08:48.720 --> 00:08:50.840 I like that, the skeleton holding it all together. 175 00:08:51.399 --> 00:08:54.879 Yeah, we can't see the bones directly, but we know 176 00:08:55.000 --> 00:08:57.799 exactly how they are shaping the glowing flesh of the 177 00:08:57.799 --> 00:09:00.879 galaxies resting on top of them. We know the skeleton 178 00:09:00.960 --> 00:09:03.600 is there because we see the exact shape of the 179 00:09:03.639 --> 00:09:05.000 body it supports. 180 00:09:04.759 --> 00:09:07.559 And historically this was a massive revelation back in the 181 00:09:07.639 --> 00:09:11.200 nineteen seventies and eighties when cosmologists first started running these 182 00:09:11.320 --> 00:09:13.320 primitive N body simulations. 183 00:09:13.480 --> 00:09:15.080 Wait, what's an N body simulation? 184 00:09:15.200 --> 00:09:18.759 Ah, there are computer algorithms that calculate the gravitational forces 185 00:09:18.799 --> 00:09:22.440 between thousands or millions of individual particles. So when they 186 00:09:22.480 --> 00:09:25.320 started running these, they realized that without dark matter, there 187 00:09:25.360 --> 00:09:27.960 simply hadn't been enough time since the Big Bang for 188 00:09:28.120 --> 00:09:31.120 ordinary matter to clump together and form the galaxies we see. 189 00:09:31.200 --> 00:09:34.080 It was too slow, way too slow. Ordinary matter is 190 00:09:34.120 --> 00:09:36.440 coupled to radiation in the early universe, so it just 191 00:09:36.440 --> 00:09:39.440 gets pushed around. It couldn't start collapsing until that three 192 00:09:39.559 --> 00:09:40.960 hundred and eighty thousand year. 193 00:09:40.879 --> 00:09:42.360 Mark, but dark matter could. 194 00:09:42.799 --> 00:09:47.120 Yes, dark matter doesn't care about light. It is collisionless. 195 00:09:47.639 --> 00:09:52.440 It started collapsing much earlier, creating these really deep gravitational wells. 196 00:09:52.759 --> 00:09:54.840 So by the time ordinary matter was free to collapse, 197 00:09:54.879 --> 00:09:57.759 the dark matter foundations were already laid out, just waiting 198 00:09:57.799 --> 00:09:58.759 for the gas to fall in. 199 00:09:58.840 --> 00:10:01.039 Wait. I want to pause there, because you just mentioned 200 00:10:01.039 --> 00:10:03.519 that dark matter is collisionless. 201 00:10:03.720 --> 00:10:05.879 I did, Yes, what does. 202 00:10:05.759 --> 00:10:08.799 That actually mean in the context of coding a simulation, 203 00:10:09.360 --> 00:10:11.120 how do you code a collisionless fluid? 204 00:10:11.480 --> 00:10:15.480 It is a brilliant mathematical abstraction. Honestly, in a computer, 205 00:10:15.799 --> 00:10:19.919 we don't simulate individual dark matter particles because if dark 206 00:10:19.960 --> 00:10:22.759 matter is composed of sub atomic particles, there would be 207 00:10:22.799 --> 00:10:26.279 an impossibly large number of them in even a cubic 208 00:10:26.399 --> 00:10:27.759 millimeter of space. 209 00:10:27.559 --> 00:10:29.879 Right the computer would just explode. 210 00:10:29.279 --> 00:10:32.840 Instantly, So instead we use macro particles. A single particle 211 00:10:32.840 --> 00:10:35.440 in the simulation might actually represent the mass of a 212 00:10:35.480 --> 00:10:39.080 million suns worth of dark matter, And because it is collisionless, 213 00:10:39.120 --> 00:10:41.960 it means these particles don't experience friction or pressure when 214 00:10:41.960 --> 00:10:42.960 they pass through each other. 215 00:10:43.080 --> 00:10:44.159 They just phase through. 216 00:10:44.480 --> 00:10:48.279 Exactly. If two clouds of ordinary gas collide, they heat up, 217 00:10:48.480 --> 00:10:51.399 they slow down, they emit X rays. If two clouds 218 00:10:51.440 --> 00:10:55.200 of dark matter collide, they literally pass right through each other, 219 00:10:55.320 --> 00:10:59.159 like ghosts, interacting only via their mutual gravitational pull. 220 00:11:00.360 --> 00:11:00.679 Eerie. 221 00:11:00.720 --> 00:11:03.240 It is so the code simply has to solve Quasslan's 222 00:11:03.240 --> 00:11:07.000 equation for gravity. It calculates the gravitational force every particle 223 00:11:07.000 --> 00:11:10.279 exerts on every other particle, moves them a tiny step 224 00:11:10.279 --> 00:11:13.360 forward in time, and then recalculates. 225 00:11:12.679 --> 00:11:15.120 But calculating the gravity of every particle pulling on every 226 00:11:15.120 --> 00:11:17.600 other particle, if you have billions of particles, doesn't the 227 00:11:17.639 --> 00:11:19.679 math get exponentially out of control? 228 00:11:19.840 --> 00:11:21.480 Yes, highly out of control. 229 00:11:21.559 --> 00:11:24.200 I mean I remember from high school physics. The gravitational 230 00:11:24.240 --> 00:11:26.960 force between two objects is easy. But if you have 231 00:11:27.039 --> 00:11:31.000 ten billion objects, calculating the distance and pull between every 232 00:11:31.000 --> 00:11:34.159 single combination of two objects sounds like it would melt 233 00:11:34.159 --> 00:11:34.600 a computer. 234 00:11:34.840 --> 00:11:38.080 You've hit on the exact computational bottleneck that plagued all 235 00:11:38.080 --> 00:11:41.759 those early simulations. If you use a naive brute force calculation, 236 00:11:41.840 --> 00:11:44.759 what we call an O N squared algorithm, the time 237 00:11:44.799 --> 00:11:46.840 it takes to compute grows with the square of the 238 00:11:46.919 --> 00:11:49.879 number of particles. So for billions of particles, for billions 239 00:11:49.879 --> 00:11:52.919 of particles, the universe would end before the simulation finished. 240 00:11:53.000 --> 00:11:57.279 That's a problem, a huge one. So computational astrophysicists had 241 00:11:57.279 --> 00:12:00.799 to invent these brilliant shortcuts, and one of the most 242 00:12:00.840 --> 00:12:04.480 famous is the tree algorithm, which was originally adapted from 243 00:12:04.559 --> 00:12:06.200 methods in plasma physics. 244 00:12:06.399 --> 00:12:07.200 How does that work? 245 00:12:07.360 --> 00:12:11.000 Well, imagine you are looking at a forest from say 246 00:12:11.559 --> 00:12:15.440 ten miles away. Do you need to calculate the gravitational 247 00:12:15.519 --> 00:12:18.039 pull of every individual leaf on you. 248 00:12:18.480 --> 00:12:19.679 No, obviously not right. 249 00:12:20.039 --> 00:12:23.120 You group the entire forest into one massive object and 250 00:12:23.159 --> 00:12:26.360 calculate the pull from its center of mass. The tree 251 00:12:26.399 --> 00:12:30.679 algorithm divides the simulated universe into a hierarchical grid. Okay, 252 00:12:30.960 --> 00:12:33.519 if a group of dark matter particles is far enough 253 00:12:33.519 --> 00:12:36.360 away from the particle you are currently calculating, the computer 254 00:12:36.480 --> 00:12:39.759 just treats that distant group as a single combined mass. 255 00:12:39.960 --> 00:12:40.799 Oh, that's smart. 256 00:12:41.000 --> 00:12:44.960 It drastically reduces the number of calculations required, changing the 257 00:12:44.960 --> 00:12:48.639 computing time from o n square to oen logan. And 258 00:12:48.759 --> 00:12:52.120 later they combine this with mesh grids, where you calculate 259 00:12:52.159 --> 00:12:56.080 the gravitational field on a three D grid and interpolate 260 00:12:56.120 --> 00:12:59.279 the forces resulting in three PM or tree particle mesh 261 00:12:59.320 --> 00:13:00.639 algorithm article mesh. 262 00:13:00.679 --> 00:13:00.960 Got it. 263 00:13:01.279 --> 00:13:03.840 That breakthrough is exactly what allowed us to scale up 264 00:13:03.879 --> 00:13:07.120 from simulating a few thousand particles in the eighties to 265 00:13:07.879 --> 00:13:09.759 literally trillions of particles today. 266 00:13:10.000 --> 00:13:13.279 That's incredibly elegant. We're using computational geometry to sort of 267 00:13:13.679 --> 00:13:16.399 cheat the math just enough to make it run without 268 00:13:16.440 --> 00:13:19.480 losing the macro accuracy precisely. Okay, so we have this 269 00:13:19.600 --> 00:13:23.879 optimized collisionless ghost skeleton of dark matter forming in the computer. 270 00:13:24.399 --> 00:13:26.840 But if we run this simulation, what does it actually 271 00:13:26.840 --> 00:13:29.519 look like? Because it's not just a random scattering of 272 00:13:29.639 --> 00:13:34.039 spherical blobs, right, it has a very specific, almost biological architecture. 273 00:13:34.120 --> 00:13:36.120 It does, and it is arguably one of the most 274 00:13:36.159 --> 00:13:40.759 stunning discoveries verified by these simulations. The dark matter organizes 275 00:13:40.799 --> 00:13:43.639 into what we call the cosmic web cladma. We imagine 276 00:13:43.759 --> 00:13:49.159 vast interconnected filaments of dense matter stretching across unimaginable distances 277 00:13:49.200 --> 00:13:52.919 of space. Where these filaments intersect, you get these massive, 278 00:13:53.080 --> 00:13:57.399 hyperdense knots called halos halos okay. And between these luminous 279 00:13:57.440 --> 00:14:02.879 threads light enormous, terrifying void regions of space millions of 280 00:14:02.960 --> 00:14:06.000 light years across that contain almost nothing at all, just 281 00:14:06.159 --> 00:14:10.879 dead space exactly. And this weblike structure wasn't programmed into 282 00:14:10.879 --> 00:14:15.279 the computers directly. It arises naturally, inevitably, just from the 283 00:14:15.279 --> 00:14:18.240 gravitational evolution of those tiny CMB fluctuations. 284 00:14:18.360 --> 00:14:20.120 I want to make sure I'm visualizing this correctly. It's 285 00:14:20.200 --> 00:14:24.960 like watching water trickle down a bumpy windshield. It naturally 286 00:14:25.000 --> 00:14:27.919 pools into streams and droplets leaving empty spaces between. 287 00:14:28.000 --> 00:14:30.159 That's a really good way to picture it. The process 288 00:14:30.200 --> 00:14:33.840 is often described by the Zeldovitch approximation. Initially, a matter 289 00:14:33.960 --> 00:14:39.039 collapses along one axis, forming vast flattened structures we call pancakes. 290 00:14:38.519 --> 00:14:41.120 Pancakes cosmologists are great at naming things. 291 00:14:40.960 --> 00:14:44.480 They really are. Then it collapses along a second axis, 292 00:14:44.720 --> 00:14:49.039 forming the long, stringy filaments. Finally, matter flows along those 293 00:14:49.080 --> 00:14:52.759 filaments and collapses along the third axis into the dense nodes, 294 00:14:52.840 --> 00:14:53.879 those dark matter. 295 00:14:53.720 --> 00:14:55.240 Halos, and the voids. 296 00:14:55.639 --> 00:14:58.399 The voids are just the regions that lost the gravitational 297 00:14:58.440 --> 00:15:02.320 tug of war, slowly empty out as all their material 298 00:15:02.399 --> 00:15:03.840 is pulled toward the filaments. 299 00:15:04.120 --> 00:15:08.559 So in our supercomputer, we now have this beautiful, twisting 300 00:15:08.799 --> 00:15:12.080 skeleton of dark matter. We have these dense halos acting 301 00:15:12.120 --> 00:15:16.279 as massive gravitational sinks. But we are missing the stars. 302 00:15:16.519 --> 00:15:18.320 We are missing the light, right. 303 00:15:18.200 --> 00:15:20.679 We are missing the galaxies. We need to introduce ordinary 304 00:15:20.759 --> 00:15:24.919 baryonic matter the gas. But wait, if gravity alone creates 305 00:15:24.919 --> 00:15:28.120 this beautiful dark matter web, why can't we just stop there? 306 00:15:28.240 --> 00:15:30.559 What goes wrong in the simulation if we hit play 307 00:15:30.799 --> 00:15:32.919 with only gravity and no ordinary gas. 308 00:15:32.960 --> 00:15:35.159 Well, if you only simulate gravity you get a highly 309 00:15:35.200 --> 00:15:37.440 accurate map of where galaxies should be, but you don't 310 00:15:37.440 --> 00:15:40.759 actually get galaxies. You just get cold, dark, gravitational wells, 311 00:15:41.039 --> 00:15:44.559 empty bowls. Exactly to understand the actual universe, you have 312 00:15:44.639 --> 00:15:48.039 to throw gas into those wells. You have to introduce hydrodynamics. 313 00:15:48.240 --> 00:15:51.080 And this is where the simulation goes from being a 314 00:15:51.159 --> 00:15:57.799 relatively clean, elegant mathematical exercise into an absolute nightmare of complexity. 315 00:15:58.080 --> 00:16:02.559 I find this historically fascinating. It's so counterintuitive that the 316 00:16:02.600 --> 00:16:05.480 stuff we understand best, ordinary matter, the gas, the stuff 317 00:16:05.519 --> 00:16:07.720 you and I are made of, is actually the hardest 318 00:16:07.720 --> 00:16:10.879 thing to mathematically simulate, oh without a doubt. Meanwhile, the 319 00:16:11.000 --> 00:16:13.720 dark matter, that we fundamentally don't understand at a quantum 320 00:16:13.840 --> 00:16:17.080 level is computationally much simpler because it's a one trick pony. 321 00:16:17.120 --> 00:16:20.639 It just does gravity. But ordinary gas gas is a diva. 322 00:16:20.720 --> 00:16:24.200 A diva is putting it mildly. Gas dynamics involve massive 323 00:16:24.440 --> 00:16:28.600 chaotic turbulence. When ordinary gas falls into a dark matter halo, 324 00:16:28.759 --> 00:16:31.120 it accelerates, and as it accelerates, it heats up to 325 00:16:31.279 --> 00:16:35.320 millions of degrees, creating supersonic shock waves that just ripple 326 00:16:35.360 --> 00:16:36.120 out through the space. 327 00:16:36.200 --> 00:16:38.159 That sounds intense. It is. 328 00:16:38.679 --> 00:16:43.120 You have incredibly complex magnetic fields twisting, snapping, and reconnecting. 329 00:16:43.759 --> 00:16:47.080 You have to calculate the cooling functions because gas can't 330 00:16:47.120 --> 00:16:50.360 form stars unless it cools down, and it cools by 331 00:16:50.360 --> 00:16:54.840 emitting radiation, which requires tracking this specific atomic transitions of 332 00:16:54.840 --> 00:16:57.080 elements like hydrogen, helium, and oxygen. 333 00:16:58.000 --> 00:17:01.720 That sounds computationally You can't just use those dark matter 334 00:17:01.799 --> 00:17:04.039 macro particles for gas, can you? 335 00:17:04.039 --> 00:17:07.119 You can, but you have to modify them entirely. There 336 00:17:07.160 --> 00:17:09.920 are two main approaches to handling the gas. One is 337 00:17:09.960 --> 00:17:13.119 grid based, where you divide the universe into millions of 338 00:17:13.160 --> 00:17:16.720 tiny cubic cells and calculate the flow of gas between 339 00:17:16.759 --> 00:17:20.480 the cells using the oiler equations of fluid dynamics DEMI other. 340 00:17:20.599 --> 00:17:24.519 The other is particle based. It's called smooth particle hydrodynamics 341 00:17:24.640 --> 00:17:28.519 or SPH. In SPH, the gas is represented by particles 342 00:17:28.519 --> 00:17:32.200 that carry properties like temperature, density, and entropy, and the 343 00:17:32.240 --> 00:17:35.559 computer essentially blurs them together to calculate fluid pressure. 344 00:17:35.720 --> 00:17:36.559 And both are hard. 345 00:17:36.799 --> 00:17:39.319 Both methods are computationally brutal, and it's. 346 00:17:39.240 --> 00:17:41.640 Not just the fluid dynamics right, because once the gas 347 00:17:41.680 --> 00:17:44.599 does cool down and form stars, those stars start fighting back. 348 00:17:44.880 --> 00:17:50.319 Yes, that is the crucial concept of baryonic feedback. In 349 00:17:50.359 --> 00:17:53.799 the late nineties and early two thousands, cosmologists ran these 350 00:17:53.839 --> 00:17:57.440 early hydrodynamics simulations and they ran into a massive problem 351 00:17:57.480 --> 00:18:00.119 called the overcooling problem. 352 00:18:00.119 --> 00:18:01.240 Cooling, what did that look like? 353 00:18:01.359 --> 00:18:05.000 Basically, the gas and the simulations cooled too efficiently. It 354 00:18:05.039 --> 00:18:07.559 fell into the center of the dark matter halos and 355 00:18:07.680 --> 00:18:11.440 formed stars at a ridiculous rate. The simulated galaxies ended 356 00:18:11.519 --> 00:18:15.279 up being way too massive, way too compact, and formed 357 00:18:15.440 --> 00:18:18.319 far too many stars compared to what we actually observed 358 00:18:18.319 --> 00:18:19.000 through telescopes. 359 00:18:19.039 --> 00:18:21.400 There were just star factories on overdrive exactly. 360 00:18:21.440 --> 00:18:25.279 The simulations were completely failing to match reality because. 361 00:18:25.039 --> 00:18:28.039 They were missing the explosions. They were modeling the stars 362 00:18:28.079 --> 00:18:30.680 being born, but they weren't modeling what happens when the 363 00:18:30.720 --> 00:18:31.759 massive ones die. 364 00:18:31.799 --> 00:18:35.480 Precisely. They were missing supernova and they were missing super 365 00:18:35.519 --> 00:18:38.519 massive black holes. A galaxy is not just a passive 366 00:18:38.559 --> 00:18:43.039 collection of stars. It is a highly volatile, finely balanced ecosystem. 367 00:18:43.480 --> 00:18:46.839 When a massive star dies in a supernova, it doesn't 368 00:18:46.880 --> 00:18:51.839 just disappear quietly. It detonates with unimaginable energy, violently heating 369 00:18:51.880 --> 00:18:54.839 the surrounding gas and physically blowing it out of the galaxy. 370 00:18:54.920 --> 00:18:56.440 In massive galactic. 371 00:18:55.960 --> 00:18:58.079 Winds, It's like the galaxy is an engine and the 372 00:18:58.119 --> 00:19:01.039 supernova is an exploding pistol, and that blows the fuel 373 00:19:01.039 --> 00:19:03.759 line apart. If the gas gets blown away, the galaxy 374 00:19:03.799 --> 00:19:06.279 is suddenly out of fuel, so it stops making new stars. Yeah, 375 00:19:06.319 --> 00:19:07.720 star formation shuts. 376 00:19:07.400 --> 00:19:10.759 Down perfect analogy. And if supernovae weren't enough, you have 377 00:19:10.799 --> 00:19:15.480 to model agn feedback AGM active galactic nuclei. At the 378 00:19:15.519 --> 00:19:18.640 center of almost every galaxy is a super massive black hole. 379 00:19:19.200 --> 00:19:21.759 As gas falls into the black hole, it forms an 380 00:19:21.759 --> 00:19:24.680 accretion disk that gets so incredibly hot it outshines the 381 00:19:24.799 --> 00:19:25.599 entire galaxy. 382 00:19:25.799 --> 00:19:26.200 Wow. 383 00:19:26.240 --> 00:19:29.680 It generates colossal jets of high energy particles that blast 384 00:19:29.680 --> 00:19:31.920 out into space, heating up the gas in the dark 385 00:19:31.960 --> 00:19:34.480 matter halo and preventing it from ever cooling down to 386 00:19:34.519 --> 00:19:39.559 form stars. Injecting that kind of localized, extreme violent energy 387 00:19:39.599 --> 00:19:44.160 into a massive, large scale cosmological simulation requires a level 388 00:19:44.200 --> 00:19:47.119 of computational finesse that we have only recently begun. 389 00:19:46.920 --> 00:19:50.440 To master, which brings us to the hardware. Because capturing 390 00:19:50.440 --> 00:19:53.880 the sheer explosive chaos of a single galaxy is hard enough. 391 00:19:54.559 --> 00:19:57.440 Trying to mathematically calculate the shock wave of a dying 392 00:19:57.519 --> 00:20:01.279 star or the jet of a black hole requires solving 393 00:20:01.319 --> 00:20:06.279 differential equations over incredibly tiny time steps, very tiny, but 394 00:20:06.400 --> 00:20:10.599 doing that simultaneously for tens of thousands of galaxies across 395 00:20:10.680 --> 00:20:14.680 a massive chunk of the observable universe, we are talking 396 00:20:14.720 --> 00:20:19.799 about true computational leviathans. I mean, my laptop fan sounds 397 00:20:19.839 --> 00:20:21.599 like a jet engine taking off just because I have 398 00:20:21.680 --> 00:20:22.920 too many browser tabs open. 399 00:20:23.039 --> 00:20:23.880 We've all been there. 400 00:20:23.960 --> 00:20:27.319 I literally cannot fathom the it infrastructure required to run 401 00:20:27.359 --> 00:20:28.279