Latest / Star Trails: From Backyard Astronomy to Cosmic Wonder / From Gears to Code: Computing the Cosmos

Transcript

From Gears to Code: Computing the Cosmos · Star Trails: From Backyard Astronomy to Cosmic Wonder

• 0:07
  Howdy Star Gazers and welcome to this episode
• 0:10
  of Star Trails. My name is Drew and I'll be your
• 0:13
  guide to the night sky for the week of April
• 0:16
  the 19th to the 25th. This week we're continuing
• 0:20
  our look at astronomy behind the scenes with
• 0:23
  a discussion on the role of computers and how
• 0:26
  we learn about the universe. From brass gears
• 0:30
  and ancient devices to programming, supercomputers,
• 0:34
  linked radio telescopes, and artificial intelligence,
• 0:38
  these machines have enabled us to see our reality
• 0:41
  more clearly. We'll even go retro with a quick
• 0:45
  foray into Fortran, a programming language that
• 0:49
  began life on punch cards, but remains deeply
• 0:52
  embedded in scientific computing even today.
• 0:56
  Later in the show, we'll take a look at this
• 0:58
  week's night sky. Whether you're tuning in from
• 1:01
  the backyard or the balcony, I'm glad you're
• 1:04
  here. So grab a comfortable spot under the night
• 1:06
  sky and let's get started. We tend to think of
• 1:12
  astronomy as something passive. You step outside,
• 1:16
  you look up, and the universe reveals itself.
• 1:20
  But that's not really how it works. The sky doesn't
• 1:23
  label its stars, it doesn't tell you where a
• 1:26
  planet will be tomorrow, and it doesn't hand
• 1:29
  you an image of a distant galaxy fully formed.
• 1:33
  To understand the universe, we've always had
• 1:36
  to compute it. And for as long as we've been
• 1:39
  doing astronomy, we've been building tools to
• 1:42
  help us think. Long before electricity, silicone,
• 1:46
  and modern technology, there were machines built
• 1:49
  to model the sky. One of the most remarkable
• 1:52
  is the Antikythera mechanism. Recovered from
• 1:56
  a shipwreck off the coast of Greece in 1901 and
• 1:59
  dated to around 100 BCE, this device is a dense
• 2:04
  arrangement of bronze gears. Intricate, precise,
• 2:08
  and astonishingly sophisticated. When scientists
• 2:11
  used X -rays and high -resolution scanning in
• 2:14
  the early 2000s, its use finally revealed itself.
• 2:19
  It could predict eclipses, track the motions
• 2:22
  of the sun and the moon, and even model the wandering
• 2:26
  paths of the known planets. The Antikythera mechanism
• 2:30
  is the oldest known analog computer, and it was
• 2:34
  a machine that embodied the cosmos, turned the
• 2:38
  crank, and the sky itself would move. And it
• 2:41
  wasn't alone. Ancient astronomers built astrolabes.
• 2:46
  beautiful handheld devices that could tell you
• 2:49
  the time, your latitude, and the position of
• 2:52
  stars. They constructed armillary spheres, rings
• 2:56
  within rings, representing the celestial coordinate
• 2:59
  system. These early computers didn't run code.
• 3:03
  They were the code. As astronomy became more
• 3:07
  precise, the sky became more demanding. Positions
• 3:11
  had to be measured, brightness had to be recorded,
• 3:14
  spectra had to be analyzed. And before electronic
• 3:18
  machines existed, there was only one way to do
• 3:22
  this kind of work. You hired people, entire teams
• 3:26
  of them, and they were called computers. At places
• 3:30
  like the Harvard College Observatory, groups
• 3:33
  of women were tasked with analyzing photographic
• 3:36
  plates. These are glass images of the night sky,
• 3:40
  captured through telescopes. Night after night,
• 3:44
  the observatory would gather data. And day after
• 3:47
  day, these human computers would process it.
• 3:51
  They measured star positions. They classified
• 3:53
  spectra. They turned raw observation into usable
• 3:57
  knowledge. Among them was Henrietta Swan Leavitt.
• 4:02
  Working quietly and methodically, she discovered
• 4:05
  a relationship between the brightness and period
• 4:08
  of Cepheid variable stars. A discovery that would
• 4:12
  become one of the most important tools in measuring
• 4:15
  the scale of the universe. Another was Annie
• 4:19
  Cannon, who developed the system we still use
• 4:22
  today to classify stars. O, B, A, F, G, K, and
• 4:28
  M. Stars arranged from hot and blue to cool and
• 4:32
  red. That sequence came from a room full of people
• 4:37
  looking at tiny points of light on glass plates
• 4:40
  and trying to make sense of them. Eventually,
• 4:43
  the limits of human computation became clear.
• 4:47
  There was simply too much to calculate, too many
• 4:50
  orbits, too many corrections, and too many tables
• 4:53
  of data to produce. So, we began building machines
• 4:57
  to take over. Early mechanical calculators, descendants
• 5:02
  of ideas like Charles Babbage's Difference Engine,
• 5:06
  were used to compute astronomical tables. These
• 5:10
  machines could perform repetitive calculations
• 5:12
  far more quickly and reliably than a human could,
• 5:16
  but they were still limited, slow, and bound
• 5:19
  by physical motion. Then came the Electronic
• 5:23
  Age. The demands of World War II accelerated
• 5:26
  the development of electronic computing, machines
• 5:30
  capable of performing calculations at unprecedented
• 5:33
  speeds, even if they took up entire rooms. Machines
• 5:38
  like ENIAC come to mind. Initially used for things
• 5:42
  like ballistics and code breaking, some of these
• 5:44
  computers soon found a new role in astronomy,
• 5:48
  calculating orbital mechanics, modeling gravity,
• 5:52
  and predicting the motion of celestial bodies
• 5:54
  with increasing precision. For the first time,
• 5:59
  astronomers could offload the heavy lifting of
• 6:01
  calculation to machines that operated at the
• 6:04
  speed of electricity. As our questions about
• 6:08
  the universe grew more complex, so did the machines
• 6:11
  we built to answer them. By the latter half of
• 6:14
  the 20th century, we entered the era of supercomputing.
• 6:18
  Machines like the Cray -1 and later the Cray
• 6:21
  -2 represented a leap forward not just in speed
• 6:25
  but in ambition. Astronomers began using supercomputers
• 6:29
  to model entire systems, the life cycles of stars,
• 6:34
  the dynamics of galaxies, and the evolution of
• 6:37
  structure and the universe itself. If you've
• 6:40
  ever seen a simulation of galaxies forming, streams
• 6:44
  of matter collapsing under gravity, swirling
• 6:47
  into vast spirals of light, you've seen the output
• 6:51
  of these machines. They allow us to recreate
• 6:54
  the universe, rather than just observe it. We
• 6:57
  can take the laws of physics, encode them into
• 7:00
  equations, and let a computer run the universe
• 7:03
  forward in time, from the Big Bang to the formation
• 7:07
  of galaxies, to the stars we see today. At the
• 7:10
  heart of all this is code, the languages we use
• 7:13
  to write the software to perform these calculations.
• 7:18
  For decades, the backbone of scientific computing
• 7:21
  was Fortran, short for formula translation. Designed
• 7:27
  by IBM in the 1950s for numerical computation,
• 7:31
  it became the language of physics, engineering,
• 7:35
  and astronomy. owing to its ability for users
• 7:38
  to translate formulas into code. NASA's earliest
• 7:42
  rockets were even engineered with Fortran. More
• 7:46
  recently, languages like Python have taken center
• 7:50
  stage. With libraries like NumPy, SciPy, and
• 7:54
  AstroPy, astronomers can analyze data, run simulations,
• 8:00
  and process images with relatively accessible,
• 8:03
  readable code. Another language, Julia, is gaining
• 8:08
  traction in the scientific community, offering
• 8:10
  the speedy number -crunching capabilities of
• 8:13
  Fortran with Python's ease of use. Even modern
• 8:18
  telescopes are computational systems now. When
• 8:21
  astronomers point a telescope at the sky, they're
• 8:24
  not manually guiding it. Computers calculate
• 8:28
  the exact position of an object, adjust for Earth's
• 8:31
  rotation, compensate for atmospheric distortion,
• 8:35
  and track that object with incredible precision.
• 8:39
  Many observatories today are fully automated.
• 8:42
  Robotic systems scan the sky night after night,
• 8:46
  capturing vast amounts of data without direct
• 8:49
  human intervention. And instead of photographic
• 8:52
  plates, we now use digital sensors, CCDs. Every
• 8:57
  photon captured is immediately converted into
• 9:00
  data. stored, processed, and ready to be analyzed.
• 9:05
  Even amateur astronomers control their scopes
• 9:08
  with computers, and recent years have ushered
• 9:11
  in a new generation of so -called smart scopes
• 9:14
  that can auto -align, track objects, photograph
• 9:18
  them, and send the results to a smartphone or
• 9:21
  tablet for integration. Which brings us around
• 9:25
  to astrophotography. We don't shoot film anymore,
• 9:28
  of course. Nowadays, images of the cosmos are
• 9:31
  constructed, often from hours of information
• 9:35
  digitally captured over hundreds of frames, possibly
• 9:39
  over different observation nights. We run these
• 9:42
  images through programs to stack multiple images
• 9:46
  to build very long exposures, reduce noise, and
• 9:50
  stretch the tonal range of high contrast deep
• 9:53
  space captures to reveal fine details and color.
• 9:57
  Astronomers have a term for this process, data
• 10:01
  reduction. That makes it sound like we're throwing
• 10:04
  something away, but that's not really what's
• 10:06
  happening. Data reduction is about taking something
• 10:10
  overwhelming, noisy, messy, incomplete, and refining
• 10:15
  it until the underlying signal begins to emerge.
• 10:18
  Removing distortion, correcting errors, combining
• 10:22
  fragments. Not reducing the data, but focusing
• 10:26
  it. This becomes even more dramatic in radio
• 10:29
  astronomy. Arrays of telescopes, sometimes spread
• 10:33
  across continents, can work together to observe
• 10:36
  the same object. Each one collects a piece of
• 10:40
  the signal and then computers combine those signals
• 10:43
  using techniques like Fourier transforms to reconstruct
• 10:47
  an image. I talked about this at length in our
• 10:50
  last episode. As data grew, so did the need for
• 10:55
  processing power. And at one point, astronomers
• 10:58
  tried something clever. They asked for help.
• 11:02
  The project was called SETI at Home. The idea
• 11:06
  was simple. Install a program on your home computer,
• 11:10
  and when your system was idle, it would analyze
• 11:13
  radio signals collected by telescopes, looking
• 11:16
  for patterns, searching for signs of extraterrestrial
• 11:20
  intelligence. Millions of people participated.
• 11:24
  Together, they created one of the largest distributed
• 11:27
  computing systems ever built. This was computation
• 11:30
  at a planetary scale. And even though it didn't
• 11:34
  find aliens, it revealed something important.
• 11:37
  That the search for knowledge doesn't have to
• 11:39
  be confined to observatories or laboratories.
• 11:43
  It can be shared. Even with all this, we started
• 11:47
  to run into a problem. Modern astronomy produces
• 11:51
  an overwhelming amount of data. Sky surveys,
• 11:55
  like the Legacy Survey of Space and Time at the
• 11:58
  Rubin Observatory, map the entire sky over and
• 12:02
  over again. Telescopes monitor millions of stars
• 12:06
  continuously. Radio arrays are generating constant
• 12:10
  streams of information. We've reached a point
• 12:13
  where no human could examine it all, and even
• 12:17
  traditional programs struggle to keep up. So
• 12:21
  astronomers are starting to do something new.
• 12:23
  They're building systems that don't just follow
• 12:26
  instructions. They look for patterns. This is
• 12:29
  the AI era of astronomy. Artificial intelligence
• 12:34
  and machine learning are becoming critical for
• 12:36
  discovery. Instead of telling a computer exactly
• 12:40
  what to look for, we can give it data and let
• 12:42
  it learn what matters. Systems can now classify
• 12:46
  galaxies automatically, detect the subtle dimming
• 12:50
  of a star as a planet passes in front of it.
• 12:53
  and identify unusual signals that might otherwise
• 12:56
  go unnoticed. AI has been used to analyze a backlog
• 13:01
  of Hubble data looking for astrophysical anomalies.
• 13:06
  According to a paper published last year in Astronomy
• 13:09
  and Astrophysics, from nearly 100 million image
• 13:13
  cutouts, at least 86 new candidate gravitational
• 13:17
  lenses, 18 jellyfish galaxies, and 417 mergers
• 13:23
  or interacting galaxies have been discovered
• 13:25
  by AI combing through this data. Impressively,
• 13:30
  the entire archive was scanned in just two to
• 13:33
  three days. I'll include a link to this paper
• 13:36
  in the show notes. For most of the history of
• 13:39
  computing and astronomy, we told machines what
• 13:41
  to do. Now, we're asking them to help us decide
• 13:45
  what's worth looking at. In a way, this is the
• 13:48
  next step in data reduction. The final step,
• 13:52
  the interpretation, the meaning, still belongs
• 13:55
  to us, for now. It's quite a journey when you
• 13:59
  think about it. We've progressed from armillary
• 14:02
  spheres to using machines as collaborators for
• 14:05
  new discoveries. The tools have changed, but
• 14:08
  the goal hasn't. Astronomy is the study of things
• 14:12
  unimaginably distant. but the work itself is
• 14:16
  intensely human, even when we're using computers.
• 14:23
  Before we move off this topic, I wanted to experience
• 14:26
  some astronomy computing from days gone by. In
• 14:31
  past episodes, I've dabbled in Python to experiment
• 14:34
  with the Drake equation, simulate star populations,
• 14:38
  and we've even used it to demonstrate how to
• 14:41
  encode and decode the Arecibo signal. So, I'm
• 14:45
  going to compile some Fortran code for a simple
• 14:48
  in -body simulation, then I'll try and modernize
• 14:52
  the output of that code by visualizing the results
• 14:55
  using Python. An in -body simulation is a way
• 15:00
  of asking a computer a very old astronomical
• 15:03
  question. If you put some objects into space
• 15:06
  and let gravity act on all of them at once, what
• 15:10
  happens next? In this case, n is a variable that
• 15:14
  refers to the number of objects in our sim. You
• 15:18
  may have heard the term three -body problem,
• 15:21
  maybe because of the Netflix series of the same
• 15:24
  name. But in astrophysics, a two -body problem,
• 15:28
  say the Earth and Moon, is fairly routine to
• 15:31
  deal with mathematically, because one object
• 15:34
  orbiting another is predictable. But the three
• 15:38
  -body problem refers to the moment when physics
• 15:41
  stops being neatly solvable, and requires some
• 15:44
  heavier computation. Motion becomes complex as
• 15:49
  the three bodies interact with one another. In
• 15:52
  our code, we're going to track the motions of
• 15:55
  25 bodies. We'll spin up a small star cluster
• 15:59
  and assign each body a mass, position, and velocity.
• 16:03
  Then Fortran will handle the calculations, letting
• 16:07
  gravity do what gravity does. Every object pulls
• 16:12
  on every other one, and the computer keeps track
• 16:15
  of it all, step by step. The first thing I needed
• 16:18
  to do was install a Fortran compiler on my Linux
• 16:22
  machine. In this case, we're using GFortran on
• 16:26
  Fedora Linux. This language has been around since
• 16:30
  1957, and its last major update was just in 2023,
• 16:36
  so it's fairly modern, despite its early roots
• 16:39
  in punch card computing. Not bad for a language
• 16:42
  nearly 70 years old. Next, I tweaked the code
• 16:47
  for the in -body SIM, compiled it, and ran the
• 16:50
  resulting executable. This only took a fraction
• 16:53
  of a second. The program produced a text file
• 16:57
  formatted as comma -separated values, containing
• 17:01
  lines indicating each object name, its mass,
• 17:05
  its position in 2D space, and its direction and
• 17:08
  velocity. This is the type of information astronomers
• 17:12
  collect, data sets indicating some sort of change
• 17:16
  over time. Initially, it isn't visual. We need
• 17:20
  to bring it to life. We can compose a little
• 17:23
  Python script to draw the star locations frame
• 17:27
  by frame and output a six -minute moving visualization
• 17:31
  of the objects. We'll bounce out the resulting
• 17:34
  animation as an MPEG -4 movie for easy viewing.
• 17:38
  And there it is, our clusters swirling around
• 17:41
  hypnotically, stars influencing one another according
• 17:45
  to Newton's laws, all being tossed asunder. It
• 17:50
  kind of looks like a computer simulation from
• 17:52
  the early 80s. If you'd like to tinker with this
• 17:56
  code yourself, I'll make all my scripts and the
• 17:59
  movie of the simulation available on our website.
• 18:02
  Just check the link in the show notes. After
• 18:21
  a quick break, we'll be back with this week's
• 18:24
  night sky. Stay with us. Welcome back. As we
• 18:40
  step outside this week, the sky greets us in
• 18:43
  a quiet, reflective mood, still recovering from
• 18:46
  the darkness of the new moon. just a few days
• 18:49
  prior on April 17th. Now a delicate waxing crescent
• 18:54
  begins to return to the evening sky. Look to
• 18:57
  the western horizon just after sunset. The moon
• 19:01
  will be low and subtle. You'll likely be able
• 19:04
  to spot Earth's shine on the moon's shadowed
• 19:07
  side. That's sunlight reflected off our own planet
• 19:11
  and back onto the moon's surface. While the moon
• 19:14
  quietly returns, something more dynamic is unfolding
• 19:17
  overhead. The Lyrid meteor shower is active this
• 19:21
  week, peaking around the night of April 21st
• 19:24
  into the early hours of April 22nd. Under good
• 19:27
  conditions, you might see 10 to 20 meteors per
• 19:30
  hour. Fast, bright streaks that sometimes leave
• 19:34
  glowing trails behind. The radiant point lies
• 19:37
  in the constellation Lyra, near the bright star
• 19:41
  Vega. But as always, you don't need to look directly
• 19:44
  at the radiant. Just get comfortable, look up,
• 19:47
  and let the sky come to you. With the moon still
• 19:50
  in a thin crescent phase, conditions this year
• 19:53
  are especially favorable. If you head out just
• 19:56
  after sunset, one object will immediately grab
• 19:59
  your attention. That brilliant point of light
• 20:02
  low in the west is Venus, still dominating the
• 20:05
  evening sky. On the 19th, Venus appears near
• 20:09
  the young crescent moon forming a beautiful pairing
• 20:12
  in twilight. Jupiter is still high in the sky.
• 20:16
  Just look almost straight up and search for the
• 20:19
  brightest point of light. Mars, Mercury, and
• 20:23
  Saturn are in a very tight grouping, but only
• 20:26
  visible in the moments just before sunrise, low
• 20:29
  on the eastern horizon. You'll likely have a
• 20:32
  hard time seeing them in the sun's glare. With
• 20:35
  the moon still modest in brightness, this is
• 20:38
  a good week to explore a few deeper targets,
• 20:41
  especially if you give yourself an hour or two
• 20:43
  after sunset. If you're using binoculars or a
• 20:46
  small telescope, consider these. Messier 3, a
• 20:51
  dense ancient globular cluster rising in the
• 20:54
  east and one of the finest of its kind. Messier
• 20:57
  13, the great cluster in Hercules, is still climbing
• 21:01
  higher each night. And Messier 81 and 82 are
• 21:07
  a beautiful galaxy pair in Ursa Major, especially
• 21:10
  under darker skies. And for something a bit more
• 21:14
  delicate, look for the Ring Nebula, a faint smoke
• 21:17
  ring suspended in space near Vega. Leo is a prominent
• 21:22
  constellation of the season, and around April
• 21:25
  24th the moon will pass near it. Look for its
• 21:28
  backward question mark shape, known as the sickle,
• 21:31
  anchored by the bright star, Regulus. That's
• 21:39
  going to do it for this week. If you found this
• 21:41
  episode interesting, please share it with a friend
• 21:44
  who might enjoy it. The easiest way to do that
• 21:47
  is by sending folks to our website, StarTrails
• 21:50
  .Show. And if you'd like to support the show,
• 21:54
  use the link on the site to buy me a coffee.
• 21:56
  That really helps. Be sure to follow Star Trails
• 22:00
  on Blue Sky and YouTube. Links are in the show
• 22:03
  notes. Until we meet again beneath the stars,
• 22:07
  clear skies everyone!