It opens with a magic trick. A bottle, a laser, and a video of light itself crawling through the glass at 250 billion frames per second. Then the camera starts to move, sweeping across the scene faster than the light pulse it is filming, and so the camera appears to outrun light, which should be impossible. That illusion is the hook for the whole video, which sets out to do one thing three different ways: keep slowing down time until you can see what is normally invisible.
Published Jan 19, 202630:09 video27 min readAdded Jun 14, 2026Open on YouTube →
At a glance
It opens with a magic trick. A bottle, a laser, and a video of light itself crawling through the glass at 250 billion frames per second. Then the camera starts to move, sweeping across the scene faster than the light pulse it is filming, and so the camera appears to outrun light, which should be impossible. That illusion is the hook for the whole video, which sets out to do one thing three different ways: keep slowing down time until you can see what is normally invisible. The journey runs from a century old strobe that still out resolves a 2020 research camera, through a one pixel camera that records a literal trillion frames per second, to a 3.2 kilometer accelerator at SLAC that builds a quadrillion frame per second movie of electrons rearranging themselves inside a molecule.
This is a Veritasium production, written and presented by Gregor Čavlović with Casper Mebius and Derek Muller, and it is built as a single escalating argument rather than a list of gadgets. Each technique trades something away to gain something else, and the through line is that every method, from the 1930s flash lamp to the attosecond x-ray laser, is fundamentally the same trick: control a brief enough burst of light, repeat it perfectly, and you can freeze anything. Below is the whole thing rebuilt in order, with every number, every name, and every aside intact.
How a strobe works: Edgerton freezes the motor
The story starts in the 1920s, when electric motors had just become the standard for powering factories and mills. They had a flaw: they were sensitive to fluctuations in the electrical grid. A power surge, say from a lightning strike, made them behave unpredictably. An MIT engineer named Harold "Doc" Edgerton set out to find the cause. He had a rig that could induce these power surges in the lab, but he could not actually see what the motors were doing, because the machines spun too fast for the human eye, and cameras at the time were no help either. Exposure times were too slow, so any photo of a running motor came out a blur.
Then came the accident that started everything. Every time Edgerton triggered a power surge, his equipment gave off a bright flash of light, and when that flash hit the spinning motor, the moving parts appeared to stand perfectly still, frozen in time. That gave him the idea the whole field is built on. Turn off every light in the room, set up a camera, and leave the shutter open. With no light, no image forms on the film. But fire one very brief, very bright flash at the motor, and you get one sharp photograph, with the brevity of the flash, not the speed of the shutter, doing the freezing.
All he needed was a way to make those flashes on demand, and the circuit he built to do it is worth following exactly, because the same idea scales all the way up to SLAC. A high voltage source loads electrons onto a capacitor, piling them onto one plate. An insulator between the two sides stops the electrons jumping straight across, so the only path back to balance is the long way around the circuit, through a glass tube filled with a non conducting gas like argon or xenon. On their own the electrons cannot punch through that gas. So Edgerton added a trigger: a high voltage pulse through a wire wrapped around the tube, whose electric field rips electrons off the gas atoms, ionizes the gas, and turns it into a conductor. In that instant the stored charge surges through and heats the gas to around 10,000 Kelvin, nearly twice the surface temperature of the Sun. The result is a flash of light lasting just 10 microseconds. Then the electrons recombine with the gas atoms, the current stops, and the tube goes dark. That is the strobe.
Figure 1. The strobe circuit. Charge stacks up on the capacitor, an insulator forces it the long way round, and a trigger pulse ionizes the gas tube so the stored charge dumps through all at once, heating the gas to roughly 10,000 K and producing a single 10 microsecond flash. Every imaging method in the video is a variation on this: make one perfect burst of light, then repeat it.
By the early 1930s Edgerton wanted to test the strobe outside the lab, so he packed one into the car and hit the road with his wife. When he passed a factory he would pull over, find the nearest phone booth, call the factory president, and ask something like, "Do you happen to have any motors in there that don't work right? I'd like to show you something." More often than not he ended up inside, setting up next to a motor, freezing it in time so the workers could finally take a sharp picture of gears in motion.
The video is careful to give credit where it is due. Edgerton did not invent the strobe; he took bits of existing technology and built a better one, brighter and with a shorter flash. Plenty of electrical engineers at the time could have done that. What Edgerton uniquely brought was his eye for photography. He photographed synchronous motors obsessively, partly because he just thought they were cool, until the day he showed his wife the 300th photo of a synchronous motor and she said, "Harold, can't you take a picture of something a little more interesting?" So he did: tennis balls pancaked against a racket, hummingbirds frozen mid wingbeat. He was among the first to use strobes to communicate what happens at timescales we cannot see, publishing in Life Magazine and National Geographic, which in the 1930s and 1940s were, as the video puts it, essentially the social media influencers of the day.
The million dollar question: how do you time the flash?
A strobe that fires for half a millionth of a second is impressive, but useless if it fires at the wrong half millionth of a second. As the team at the MIT Edgerton Center frames it, that is the million dollar question. The answer is sound.
To prove it the crew recreates one of Edgerton's photos: popping a balloon and freezing it. The setup is a checklist worth keeping. First you set up the experiment, or in this case the performer holding the balloon. Then you frame the image and lock focus before anything happens, because, in a line that doubles as photography advice, if you cannot get a good photo with nothing happening, adding the motion will not help. The strobe is wired to a trigger unit with a microphone, so a sharp sound sends the signal to fire. Lights out, shutter open, no image yet because the room is dark. Three, two, one, pop. When the sound of the pop hits the mic, after a tiny delay the strobe fires, and the camera records the scene for the one hundred thousandth of a second it is lit. The result is a balloon caught mid burst, the rubber peeled back so you can see inside.
A second photo shows a hovering white orb that becomes, a moment later, a tiny sombrero. It is a drop of milk hitting a plate, the splash crown caught so crisply that the rim is translucent and you can see straight through it.
Strobe versus modern: the 1930s wins
Edgerton's photos look impossibly sharp, sharper than they have any right to be for the 1930s, so the video runs a head to head. On one side, a research grade slow motion camera from 2020 shooting 20,000 frames per second. On the other, Edgerton's single flash method. The test: a bullet through a playing card.
The high speed camera does fine; you can watch the top half of the card levitate after the bullet passes. But Edgerton's single frame is the clear winner on sharpness. The focus is flawless, the edge of the card is clean, and the only flaw is a faint ghost of the card from stray light leaking in during the second or two the shutter sat open before the gun fired.
The timing trick for the bullet is even cleverer than the balloon's, because a bullet outruns sound. So you cannot trigger on the muzzle blast. Instead you use the fact that a supersonic bullet drags its own sonic boom through the air. Pick up that boom with a microphone, slide the microphone physically along the bullet's path, and you control exactly where the bullet will be when the strobe fires. The Edgerton Center expert is generous about it: "I think it's a brilliant way to solve the problem and I get to say that because I did not invent it."
Edgerton himself was relentlessly inventive across many fronts at once. He taught at MIT, ran companies for things like underwater cameras, made movies, and even won an Academy Award. If there was something he wanted to do, he just did it. That sets up a personal aside from the presenter, who joined Veritasium in 2023 as a researcher, fact checking videos and setting up shoots, kept being told by Derek and the other writers to make a video of his own, kept putting it off with a "maybe one day," and finally did it and fell in love with it. The takeaway, which the video later loops back to: if there is anything you are putting off, just go do it. The strobe story even has a wartime chapter. In 1939 a US major named George Goddard walked into Edgerton's lab unannounced; he worked in the Army's photographic lab on night reconnaissance. The old method was to fly high, drop a flare on a parachute, then fly under it, where the plane was silhouetted and could be shot at. Goddard wanted something safer. He asked whether Edgerton could build a strobe bright enough to light the ground from a mile up. Edgerton did a few calculations and said, "We can do that." The flash released about 60,000 joules in a single millisecond, a peak power near 60 megawatts, comparable to the output of a large solar farm. It saw service in World War II and let the allies photograph Normandy the night before D-Day, confirming German troops were unprepared for the attack.
Spatial versus temporal: the trade you cannot escape
Why does a 2020 camera lose to one from decades ago? Because every imaging system juggles two resolutions, and hardware almost always forces you to trade one for the other. Spatial resolution is how many pixels the image has. Temporal resolution is how finely you slice time, whether you capture one frozen frame like a strobe or a smooth progression of frames like high speed video.
The hard limit is how fast you can pull pixels off the sensor. There is a maximum rate to read every pixel, and the only way to go faster is to stop reading all of them. As the video puts it, a camera might give you a million frames a second, but at only 16 by 128 pixels, which is barely an image. So there is always this trade. Crank up pixel count and the frame rate falls. Crank up frame rate and the resolution collapses. Edgerton sits at one extreme: maximum spatial detail, exactly one frame in time. The next two methods push hard the other way.
Figure 2. The trade no camera escapes. The faster you slice time, the fewer pixels you can read off the sensor, so the methods slide down one curve. Edgerton lives top left, every pixel and a single frozen frame; the trillion FPS single pixel camera lives bottom right, one pixel and a trillion slices a second. Beating the trade means cheating it, which is exactly what the next method does.
One trillion frames per second, from a single pixel
Push the trade all the way to the temporal extreme and you arrive at the opening trick: light moving through a bottle, captured at a trillion frames per second, by a camera that sees only one pixel.
The single pixel camera measures just one quantity, how many photons land on the sensor, and the sensor is sensitive enough to register even a single photon. It counts those photons about a trillion times a second, so each frame, each time bin, is roughly one picosecond long. In one picosecond, light itself travels only 0.3 millimeters. The video notes this is not exotic hardware; the same kind of single photon timing is the LIDAR already in many phones, where you fire a pulse, time the bounce back, and read off distance. Here you use it not to measure distance but to film light in flight.
The demo, run by MIT's Camera Culture Group, is a scaled down room holding a cone, a sphere, a mirror at the back, and the Veritasium and Camera Culture logos. To film light crossing it, you fire a very short laser pulse, packed with photons, at a single point. The photons scatter everywhere, and you want to see that scattering across the whole scene. The single pixel detector starts aimed at, say, the top left corner. A pack of photons hits a spot on the wall, reflects to that corner, and bounces into the camera, which times the arrival at a trillion frames per second, though the signal is faint. To beat the faintness you fire the identical pulse many times and stack the measurements until the timing of light reaching that one point becomes clean.
Then you nudge the detector to the next point and repeat: fire, scatter off the same spot, record, a couple hundred times per pixel. Two steering mirrors swing the line of sight left, right, up and down so the one pixel sweeps a grid over the whole scene. As the team says, you are literally going one pixel at a time. The catch, and it is the same catch that returns later at SLAC, is that the scene must play out essentially identically every single time you move the detector. If it did not, each pixel would tell a different story and you would get a garbled mess, like recording one section four times and using a quarter of each to fill the frame. Because a laser pulse scatters predictably, the scene is repeatable, and that repeatability buys you unlimited spatial resolution: want 4K, scan a 4K grid, it just takes longer. Light is fast, so the limit is really how fast the mirrors move, and within a couple of minutes the sensor logs millions of pulses across the grid. Stitched together, you watch a laser pulse sweep the room and light up the Camera Culture logo, with the whole show lasting under eight nanoseconds.
Figure 3. A camera that sees one pixel and a trillion frames a second. A laser pulse scatters through a fixed, repeatable scene; two mirrors sweep a single photon counting detector across a grid; the identical pulse is fired hundreds of times per point and the faint signals are stacked. Because each frame is about a picosecond, light crosses only 0.3 mm per frame, so you literally watch the pulse travel, the whole event under eight nanoseconds.
Then it gets stranger. By rotating the scene and recording many viewpoints, the University of Toronto Computational Imaging Group and MIT can reconstruct fly throughs, watching the light pulse propagate from any angle. In one Coke bottle fly through the virtual camera moves rightward faster than the light it is filming, so a wavefront that is genuinely moving right appears frozen in place. That is the opening illusion explained: "That is kind of breaking physics. I mean, you are moving the camera faster than light." Nothing physical breaks; the camera is virtual, reconstructed after the fact, so it can be flown at any speed. A fish tank shows a pulse entering, bouncing off a mirror, and hitting a diffuse reflector; a diffraction grating shows the pulse splitting into separate modes. It looks so clean it could be Unreal Engine 5, but it is real data. The presenter compares it to the bullet time effect in The Matrix, and the researcher admits The Matrix really was part of the motivation. As a final note, these were built at Toronto and MIT, but Brian from AlphaPhoenix built one of these speed of light cameras in his garage, which the video calls mad and worth checking out.
Combining both extremes: a movie of electrons at SLAC
So far, two opposite corners of the trade: all spatial detail (Edgerton) and all temporal speed (the trillion FPS single pixel). The video's last act argues that if you combine the spirit of both, a perfectly repeatable scene plus an absurdly fast strobe, you can image what electrons are doing. The video is honest that even saying what that means is contested. Do electrons act like waves? Not exactly. Like particles? Not exactly. We can write the math and predict what the thing will do without being able to picture it. Asked point blank whether electrons even exist, the researcher answers, "How truthful do you want me to be?" There is still no photograph of an electron, but this may be the closest thing yet, and pulling it off meant building very big.
The setting is SLAC, a US national lab housing a 3.2 kilometer long, perfectly straight electron accelerator, which until 2017 was considered the world's straightest object. Driving its length takes five minutes and you are only halfway. A constant 120 hertz hum fills the tunnel; that is the rate at which electron pulses are generated, 120 pulses a second, accelerated to over 99.9999992 percent of the speed of light. The point of all this is to watch electron clouds move around molecules, and the reason that matters is foundational: electrons create the fields in which everything else happens, molecular bonds break and form because electrons push them, so electrons are responsible for everything you see in nature, and watching their motion is the most fundamental way to study materials and matter.
To strobe at that scale you need a nanoscopic flash, and the way SLAC makes one is a beautiful relativity argument. The relativistic electron pulses first run through devices called undulators, stacks of magnets spaced only a few millimeters apart with alternating poles, north over south, then flipped, then flipped again. Because the electrons move through a magnetic field, the Lorentz force pushes them perpendicular to both their velocity and the field lines, following the left hand rule, so they curve clockwise at one magnet pair and counterclockwise at the next. The electrons wiggle. A wiggling charge radiates, so they emit electromagnetic radiation, and although the wiggle period is set by the centimeter scale magnet spacing, the wavelength of the radiation comes out vastly smaller. Why? Relativity. To the electron, traveling near light speed, all that magnet spacing length contracts, so the structure it actually experiences is microscopically tight and it oscillates extremely fast, pushing the radiation toward x-ray wavelengths. That gets you only part of the way. The rest comes from the observer at the far end: the electrons are racing toward you at over 99 percent of light speed, so their light is blueshifted on top, producing x-rays as short as 50 picometers.
Figure 4. How SLAC builds a nanoscopic strobe. Alternating magnet poles a few millimeters apart wiggle a near light speed electron via the Lorentz force. To the electron that spacing length contracts, so it oscillates fast; to the observer the light blueshifts; together they push the emission to roughly 50 picometer x-rays. Microbunching then locks the electrons into sheets one wavelength apart so they radiate coherently, like a laser, in pulses a few femtoseconds down to a few hundred attoseconds long.
There is one more amplification step. At first the x-rays appear randomly along the undulator, an incoherent jumble. But the x-ray electric fields start nudging the electrons, speeding some up and slowing others, so faster electrons catch up to slower ones and the beam self organizes into parallel sheets spaced exactly one x-ray wavelength apart. This is microbunching. Those sheets radiate as unified fronts, so the x-rays come out coherently, like a laser, with a huge jump in intensity. The pulses are only a few femtoseconds long and can be squeezed to a couple hundred attoseconds, that is ten to the negative eighteen seconds. The video gives the scale that lands hardest: the attosecond is to one second what one second is to the entire age of the universe. On that timescale, you can watch electrons zip around atoms and molecules.
Reading an electron's neighborhood: the x-ray strobe
The coherent x-ray pulses travel to experimental stations at the end of the tunnel, focusing into what the team calls an interaction point, which you fill with the molecules whose electrons you want to study. When the x-ray pulse hits a molecule it ionizes it, knocking electrons predominantly out of the inner, core shells. The clever part is selectivity: core electrons of different elements have different ionization energies. Ejecting a core electron from nitrogen takes around 400 electron volts; from oxygen, around 550. So by tuning the x-ray energy to a specific ionization energy, you choose which atom in the molecule to ionize.
Any energy left over after ejecting the electron is carried off as that electron's kinetic energy, and that number is the whole measurement. Electrons are not independent; they talk to each other through their negative charge. Where electron density is high around an atom, the core electrons feel all those neighbors and are bound a little less tightly, so their ionization energy drops slightly. Where density is low, the core electrons are bound more tightly and ionization energy rises. So by measuring the kinetic energy of an ejected electron and subtracting it from the known input x-ray energy, you infer the local electron density. You have, in effect, photographed the electronic structure around that atom.
Method
Flash / frame
Speed
The trade
Edgerton strobe
10 µs flash, ~1/100,000 s exposure
1 frame
all spatial detail, one moment in time
Single pixel camera
~1 picosecond per frame
~1 trillion FPS
1 pixel swept on a grid; scene must repeat exactly
SLAC x-ray strobe
few fs to ~hundreds of as
over 1 quadrillion FPS
infers density, not a photo; scene must repeat exactly
A quadrillion frames per second: the molecular movie
A single snapshot is a still. To make a movie you add a trigger and a clock. Above the experimental hall sits an entire laser hall generating traditional infrared laser light, piped down through tubes. Rows of boxes on the optical table condition each pulse, changing its color, polarization, and duration, sculpting it precisely. That sculpted laser pulse is sent co-propagating with the x-rays into the target. The laser pulse fires first and kicks the molecule into a non equilibrium state, driving some dynamics. Then, after a chosen delay t, the attosecond x-ray pulse probes it by ejecting a core electron, and the electron's kinetic energy tells you how the electron density looked at that instant. That is one attosecond snapshot, a strobe of the molecule mid change.
Repeat with a fresh, identical molecule but bump the probe delay to t plus delta t, and you see the density a little later. Keep stepping the delay and you build a sequence of snapshots tracing how the electron density evolves. The same assumption the trillion FPS camera needed returns here, and the team flags it directly: this only works if the trigger drives the same dynamics every single time. A new process on each shot and the technique fails. But if the scene is repeatable, you can stitch the snapshots into a molecular movie, and because the smallest delay step is around 300 attoseconds, the frames sit only a few hundred attoseconds apart, which works out to a movie running at over a quadrillion frames per second.
The example molecule is para-aminophenol, a small system. Collaborators in Madrid (Gilbert Grell, Alicia Palacios, and Fernando Martín at the Universidad Autónoma de Madrid) computed how it responds to losing an electron: simulate an x-ray pulse removing an electron, then watch the charge redistribute. In the visualization red marks an increase in electron density and blue a decrease, and you watch a charge distribution begin migrating across the molecule. The simulation looks polished, but it was validated against the SLAC measurements, which is where the real payoff lives. The method tracks the simulation well for the first few femtoseconds, then somewhere between five and ten femtoseconds prediction and measurement start to diverge. Rather than a failure, the researcher calls this the most exciting time in science: "When you have a prediction, and then you have a measurement, and they don't agree, that's when you get really excited, because you just found something you didn't know ahead of time." The closing reflection ties it back to the channel itself: nearly every Veritasium video animates electrons moving in some way, so actually seeing those electron densities move around, in real data, is, in the presenter's word, spectacular.
Key takeaways
Brevity, not shutter speed, freezes motion. Edgerton's whole insight was darkness as a canvas: open the shutter in a dark room and let one 10 microsecond flash, heating gas to ~10,000 K, do the work.
Every camera trades spatial resolution against temporal resolution. A million FPS camera drops to 16 by 128 pixels; Edgerton chose one frame with full detail; the single pixel camera chose one pixel with a trillion frames a second.
Timing a strobe is the hard part, and the trick is sound. A microphone catches the balloon pop, or a supersonic bullet's sonic boom, and slides along the trajectory to fire the flash exactly on target.
A one pixel detector can film light itself. Counting single photons a trillion times a second, each frame is one picosecond, in which light moves only 0.3 mm; sweep one pixel across a perfectly repeatable scene and reconstruct it under eight nanoseconds.
Relativity turns a kilometers long machine into an attosecond strobe. Length contraction and blueshift convert centimeter magnet spacing into 50 picometer x-rays; microbunching makes them coherent, down to a few hundred attoseconds.
Electron density is read from leftover kinetic energy. Tune the x-ray to an element's ionization energy (≈400 eV nitrogen, ≈550 eV oxygen), measure the ejected electron's energy, and infer the local electron density.
Repeatability is the hidden price of speed. Both the trillion FPS camera and the quadrillion FPS molecular movie demand a scene that plays out identically on every shot.
Disagreement is the prize. When the para-aminophenol simulation and the SLAC measurement diverge past five to ten femtoseconds, that gap is where new physics hides.
Chapters
Timestamps are clickable. Click one and the player jumps there and keeps playing while you read.
00:00 How does a strobe work?
03:28 The Man Who Stopped Time
08:43 Spatial vs Temporal Resolution
13:09 1 Trillion FPS
16:56 Watching Light Move
19:04 The Straightest Building in the World
24:19 How to get 1 Quadrillion FPS
28:25 Seeing Electrons
Notable quotes
Do you happen to have any motors in there that don't work right? I'd like to show you something.
Derek Muller, 03:50
Harold, can't you take a picture of something a little more interesting?
Derek Muller, 04:31
If you can't get a good photo with nothing happening, adding the motion will not help.
Derek Muller, 05:13
I think it's a brilliant way to solve the problem and I get to say that because I did not invent it.
Derek Muller, 07:12
So if there's anything you're putting off, you should just go do it.
Derek Muller, 07:34
That is kind of breaking physics. I mean, you are moving the camera faster than light.
Derek Muller, 18:13
The attosecond is to the second what the second is to the age of the universe.
Derek Muller, 27:48
When you have a prediction, and then you have a measurement, and they don't agree, that's when you get really excited, because you just found something you didn't know ahead of time.
Derek Muller, 29:50
Electrons are responsible for everything that you see in nature.
Derek Muller, 20:48
Do electrons exist? How truthful do you want me to be?
Derek Muller, 19:32
Resources mentioned
Veritasium, Derek Muller's science channel, which produced and presented the video.
Harold "Doc" Edgerton, MIT engineer, inventor of the modern electronic strobe, Academy Award winner.
The MIT Edgerton Center, where the crew recreated the balloon, milk drop, and bullet through a card demonstrations. Thanks in the video to Jim Bales, Kim Vandiver, and Jamie Chelel.
Hostinger, the video's sponsor, an AI powered website and hosting platform.
Snatoms, Derek Muller's magnetic molecular modelling kit, mentioned in the video description.
Closing
The video never says "relativity is the secret," but it is. Edgerton's strobe and SLAC's attosecond x-ray laser are the same idea separated by ninety years and twelve orders of magnitude: make a single clean burst of light, repeat it perfectly, and the frozen world reveals itself frame by frame. What changes down the scale is what you must trade away, all your time resolution for Edgerton's detail, all your pixels for the trillion FPS camera, and a perfectly repeatable molecule for the quadrillion FPS movie of electrons. The most quietly radical claim is the last one: when the simulation and the measurement finally part ways past ten femtoseconds, that disagreement is not an error to fix but a door to walk through. Keep slowing down far enough and you stop watching the world and start watching the math of the world come apart, which is exactly where the next discovery lives.
Full transcript
This is a video of light traveling through a bottle at 250 billion frames per second. And here's that same video. But now with the camera moving. You can see it sweeps across the scene faster than the laser pulse itself, which means this camera must be traveling faster than light. So how is this possible? Well, in this video, I wanna show you three unusual ways of stopping time and what you can see if you just keep slowing down. From a century old technique that still beats modern slow-mo cameras, all the way to a massive quadrillion frames per second camera that captures electrons whizzing around molecules.
By the 1920s, electric motors were the new standard for powering factories and mills, but many of these motors also came with a flaw. They were sensitive to fluctuations in the electrical grid. A power surge, like from a lightning strike, made them behave unpredictably. So one MIT engineer named Harold "Doc" Edgerton set out to find a solution. He had a setup that could induce these power surges in a lab, but no matter what he tried, Edgerton just couldn't see what was going on with the motors because the machines would spin too fast for the human eye to see. And cameras at the time offered no help. Their exposure times were too slow, so any photograph of a running motor would come out blurry.
But one day, Edgerton noticed that every time he triggered a power surge, his equipment gave off a bright flash of light. And when that flash hit the motor, the moving parts appeared to stand perfectly still as if frozen in time, which gave him an idea. He could turn off all the lights in the room, set up a camera, and leave the shutter open. And since there was no light, no image would form on the film. But then if he could illuminate the motor with a very brief and very bright flash, like the ones his equipment gave off, well then he would get a sharp photograph.
All Edgerton needed was a way to reliably create these flashes. So he started by using a high voltage power source to load electrons onto a capacitor where they piled up onto one of the plates. But because there was an insulator slotted between the two sides, the electrons couldn't just jump to the positive side to balance out the charges. The only way for them to get there would be to travel through the rest of the circuit. And the circuit was intentionally designed so that electrons would have to cross a glass tube filled with a non conducting gas like argon or xenon. On their own they would not have the energy to get through that gas. So Edgerton added a trigger that sent a high voltage pulse through a wire wrapped around the tube, and the electric field from that pulse would rip electrons off the gas atoms inside the chamber, ionizing the gas and turning it into a conductor. In that instant, the charge stored in the capacitor would surge through, heating the gas to around 10,000 Kelvin, nearly twice as hot as the surface of the Sun. This would produce a very bright, very brief flash of light lasting just 10 microseconds. Then the electrons would recombine with the gas atoms, stopping the current, and the circuit would go dark again. This was Edgerton's strobe.
By the early 1930s, he was eager to test it outside the lab, so he packed up a strobe and hit the road with his wife. When he saw a random factory, he pulled over, got inside the nearest phone booth and called up the factory's president and asked him something like: "Do you happen to have any motors in there that don't work right? I'd like to show you something." More often than not, he ended up inside setting up the strobe next to one of the motors. The workers would watch as Edgerton froze the motor in time, allowing them to take sharp pictures of the gears in motion.
Edgerton isn't the first to make a strobe. Rather he took new bits of technology that existed so he could make a better strobe. A strobe that was brighter, shorter flash duration. But he was not unique in that. There were lots of electrical engineers in the world at that time who could have done that. No, what Edgerton uniquely brought to the table was his eye for photography. He took photos of synchronous motors, and I think in part because he just thought synchronous motors were cool. One day he showed his wife the 300th photo of a synchronous motor. And she said, "Harold, can't you take a picture of something a little more interesting?" And so he did. Tennis balls, pancaked against the racket, hummingbirds frozen in time.
He was one of the first to really start using strobes to communicate what's happening at these timescales we can't see. He would do this through like "Life Magazine," "National Geographic," magazines. These magazines in the '30's, '40's, were essentially the social media influencers of the day. He just had this eye for composition.
Most of these pictures were taken in the 1930s and yeah, it seems easy enough to swing a racket and it's easy enough to press a button on the strobe, but how do you time the strobe to go off exactly as the racket hits the tennis ball? That is the million dollar question, right? You have a strobe that turns on and off in half a millionth of a second, that's nice. How do you get it to go off at the right half millionth of a second? 'Cause there are a lot of them, right? And the answer is we use sound.
So we're gonna try and recreate one of Edgerton's photos, popping a balloon and freezing it in time. Is it okay if we walk through the setup? I think that would make perfect sense. Why don't you blow up a balloon. Step one is you set up the experiment, or in this case, the performer. And the next thing you wanna do is frame the image. So you're framing now before the balloon pops to get focus and things? Exactly, if you can't get a good photo with nothing happening, adding the motion will not help. And so the next step will be to get the strobe in the right spot. Now the strobe is set up with a trigger unit with a microphone. And when a sharp sound hits the microphone, the trigger unit sends a signal to set off the strobe.
We're gonna turn the lights out and then we're gonna open the shutter of the camera, but there won't be an image because the room will be dark. And I'll say three, two, one, pop. And when I say pop, pop the balloon with an upward motion. And when the sound from the pop hits the mic, after a minor delay, the strobe will fire. The camera will capture that image for the one 100,000th of a second it's lit. Alright, we ready? Lights out please. And three, two, one, pop. Lights. Oh, there you go. Oh, you look hmmmm. Do not mess with this man. Nope. Oh, it's awesome. You can see inside the balloon. That's really cool.
This is another image we took. Can you guess what this is? This hovering white orb. Here's another photo just a moment later. It's like a little sombrero! It is, yes. That orb is a drop of milk falling onto a plate. That's a pancake. But you'll notice the little drops are all spreading out. Yeah, it's so, so crisp. Come around and have a look here. Right, it's translucent here. You're seeing through it. Oh, wow.
Now, once Edgerton showed the world how powerful strobe photography was, he attracted some unexpected attention. In 1939, a US major named George Goddard walked into Edgerton's lab unannounced. He was working in the Army's photographic lab, developing ways to photograph enemy movements from a plane during the nighttime. The old way of doing a night reconnaissance photograph was to fly over the site at high altitude and drop a flare on a parachute. And then the reconnaissance plane had to fly in under the flare where it would be silhouetted, where you could shoot at it. Big problem... Right? So totally exposed. Goddard wanted a safer way. So he asked Edgerton whether he could develop a strobe powerful enough to illuminate the ground from a plane that was a mile or so up in the sky. A strobe that would be bright enough to take a reconnaissance photo. Edgerton pulled out some paper, did a few calculations and said: "We can do that."
The flash released about 60,000 joules in a single millisecond. A peak power of roughly 60 megawatts, which is comparable to the output of a large solar farm. One, two, three, push. The flash lamp was quickly utilized in World War II and it allowed the allies to take pictures of Normandy the night before D-Day. This way they could confirm that German troops were unprepared for the attack.
It's hard to ignore just how sharp these strobe photos are, especially the ones Edgerton took in the 1930s. So we got a research grade slow-mo camera from 2020 that shoots at 20,000 FPS, and we're gonna compare its quality to Edgerton's technique by shooting a bullet through a playing card. So let's do the slow-mo camera first. Three, two, one. Let's see the video. Oh, it's great to see how long the top part, which is now levitating, it stays up. Is now, yeah. Okay, and now let's do the same with Edgerton's method. Light's going out. Shutter. Shutter open. Three, two, one. Okay, I think I saw it! That's cool. The focus is amazing. The edge of the card is beautiful. You see this ghost effect? There's like card. Ah, you do see the ghost effect and that's because you open the shutter and it was a second or two before I actually fired the gun. There's enough stray light in the room to give you a faint exposure there.
Also, we still used a microphone to time the bullet, even though it's faster than sound. So here's the gun, it fires, the sound of the gun comes out, but the bullet is coming out ahead of the sound, but the bullet is supersonic and a supersonic object moving through the air creates a sound, a sonic boom. Now you can pick up that sonic boom with a microphone, and by moving the microphone physically along the trajectory, you get the time where the bullet is gonna be when the strobe goes off. I think it's a brilliant way to solve the problem and I get to say that because I did not invent it.
Edgerton was very inventive and had projects all over the place. He was teaching at MIT, but then he had companies for things like underwater cameras and he was making movies, and oh, he even won an Oscar. So if there's anything he wanted to do, he sort of just did it. And I feel like I'm quite the opposite. You know, when I joined Veritasium back in 2023, I started off as a researcher. I was fact checking videos and setting up shoots, but then Derek and the other writers suggested I should make a video of my own. And I remember thinking, yeah, I don't know, maybe one day. But they kept pushing for me to do it. And I'm so glad that they did because when Derek and I made my first video, I fell in love with it. So if there's anything you're putting off, you should just go do it.
And if that something is a project you want to set up as an entrepreneur or a creator, well, today's sponsor Hostinger, makes that first step really simple. Say you need a website or a store or any kind of online presence, Hostinger is an all-in-one AI-powered ecosystem that will get you up and running within minutes. With Hostinger you don't need any kind of crazy tech skills. It's all frictionless and affordable. So you can go from an idea to execution without needing 10 apps on your computer. If there's a project you've been putting off for that 'one day', turn that into your 'day one' with Hostinger. Head to hostinger.com/Veritasium10 and use our code Veritasium to get 10% off your subscription. You can also scan the QR code here. So I want to thank Hostinger for sponsoring this part of the video.
And now let's go look at pictures of that card we shot. So here are the two techniques side-by-side, and here's Edgerton's original as well. So why does a research grade camera from 2020 struggle to get the same resolution as a camera from decades ago? That's because we are really working with two resolutions here. A spatial resolution, or how many pixels your image has, and a temporal resolution which dictates whether you only capture one frame, like a strobe photo, or a progression of frames like our high speed video. The problem is that more often than not, the hardware is limited so that you really have to trade one resolution for the other. High pixel count or high frame rate.
The fundamental limit you hit is how fast you can get pixels off the sensor. And that's why there's a maximum speed to read every pixel, and then to go any faster, you have to not read out all the pixels. So this camera will give you a million frames a second, but you're like 16 by 128 pixels. And that's not much of an image. Okay, so there's always this trade off. Either you go for very high pixel counts and bring the frame rate down. At the Edgerton Center we pushed this as far as it goes with one frame and that's it. But you can also push it the other way, one pixel and very high FPS. 1 trillion FPS in fact.
But wait, it's just one pixel. What can you do with one pixel? Great question. The cameras that I can show you today are cameras that, you're right, they can really only see one pixel at a time, but they can see close to a trillion frames per second. And what that lets you do is ultra slow motion videos showing light actually traveling. Here's that video of light traveling through a bottle. You can see the wavefronts that form below the bottle and even how the light bounces off the cap. And even though this looks like a normal video, you can take it with a camera that only sees one pixel.
Here's how you do it. A single pixel camera is one that captures just one thing, how many photons land on the sensor. And the sensor here is typically sensitive enough to register whether even a single photon hits it. And it can count those incoming photons around a trillion times a second. So each bin, and technically each frame, is roughly one picosecond long. In that time, even light itself travels only 0.3 millimeters. Sounds impressive, but we've actually had this tech in many phones for years now. It's just LIDAR. You shoot out a pulse of light, it bounces off and you time how long it takes for that pulse to come back. And from that you get the distance to the object that it bounced off of. But this is all you need to take a speed of light video.
We have a setup here that is basically a scaled down room. And we just have some different shapes in it. We have a cone, a sphere, we have a mirror in the back. And finally we have the Veritasium and the Camera Culture logos. We wanna see light propagating through a scene. So the way we do that is we shoot out a really short laser pulse that hits just one point in the scene and that laser pulse has a ton of photons in it. Those photons will hit an object, scatter everywhere, and we want to see what that scattering looks like at all these portions of the scene.
To start, our single pixel camera will point at the top left corner. So when a pack of photons from the laser pulse hit this random spot on the wall, reflect to the corner and finally bounce into the camera, the sensor is gonna pick up their signal at a trillion frames per second, but that signal will be pretty faint. So the problem here is that we're exposing for such a short time that you actually just don't get that much photon return. What we do is we actually take a bunch of measurements and then group them all together, and that gives us actual usable information about, you know, how far away the light traveled in a scene.
Then you move the camera slightly and repeat the experiment. So you shoot out a laser pulse, you let it scatter off the same spot and record the signal from this new position. You can do that a couple hundred times and move the camera again until you get a grid of points across the whole scene. You're literally going just one pixel at a time. One pixel at a time. That's the caveat. Yeah, exactly. We actually have two mirrors here that let us steer the beam, you know, left and right and up and down. So by turning where the mirror is, we can turn where the sensor looks.
The most important thing for this technique to work is that the scene has to play out pretty much exactly the same every time you move the sensor, because if it didn't, then every pixel will tell a different story. It's like if I try to record this section four separate times and use a quarter of each to fill in the screen, I would get a garbled mess. Thankfully the laser pulse in our scene scatters pretty predictably. That's what lets us get unlimited resolution. So we can basically say, let's scan as many points in the scene as we need. That gives us good spatial resolution. The more points you scan along this grid, the higher your final resolution. If you want 4K, you simply scan a 4K grid of pixels. It's just gonna take more time. The nice thing is light is fast, so we can do this as fast as the mirrors can move.
Within just a couple of minutes, the sensor captures millions of laser pulses across the whole grid. So this is now everything compiled together for a time of flight? Exactly, this is everything put together. So what we're looking at now is going to be again for a fixed laser spot. So I'm gonna click play. Oh, Camera Culture logo! Yeah. All of this was less than eight nanoseconds of time. And here's another scene under the same setup.
Now you can also take this a step further by rotating the scene and recording multiple points of view. We have this algorithm that kind of takes this Coke bottle video capture from different views and is able to create these fly-throughs, being able to see the light propagate from any direction and like flying through the scene as it's happening. Oh, that's so cool. So some things that are interesting to note is, because we're moving towards the right and the light is propagating towards the right, but we're moving kind of faster in this visualization, this wavefront appears stationary, as you can see. Oh yeah. That is kind of breaking physics. I mean, you are moving the camera faster than light. Yeah, exactly, yeah, it's kind of mind boggling almost.
So this is a fish tank that we put a mirror into and this diffuse reflector. So you'll see a pulse of light will enter the fish tank, it will reflect off the mirror and hit the diffuse reflector. That's crazy. This is a diffraction grating. So it kind of defracts the light into different modes. So again, you see the light enter the tank and it separates into these different modes. Those are the different modes? That's insane. My first impression was, oh, these are just simulations from I don't know, Unreal Engine 5. But this is, like, real data. It's like the bullet time video in "The Matrix." You've probably seen that. No way! Is your motivation "The Matrix?" By the way, these videos were created at the University of Toronto and MIT, but Brian from AlphaPhoenix actually built one of these speed of light cameras in his garage, which is mad. You should go check that out.
So those are the two extremes. Strobe photography on one side and a trillion FPS on the other. But if you combine both, you actually get to see what electrons are doing. Even though what that means exactly is debated. You say electrons act like waves? No, they don't exactly. They act like particles. No, they don't exactly. We can write mathematical expressions and calculate what the thing is going to do without actually being able to picture it. Do electrons exist? How truthful do you want me to be?
Now we still don't have a video or photo of electrons, but this might be the next best thing. And to pull it off, we had to build big, really big. Okay, we're driving down what up until 2017 was the world's straightest object. So we've been driving for like what, five, five-ish minutes? Yeah, and we are wanna say about halfway there. Oh, I thought we were… Oh no, there's a lot more there. Let's do it. Let's go see it.
This is SLAC, a US national lab that houses a 3.2-kilometer-long, perfectly straight electron accelerator. Whoa. That is so long down there! And it just continues like this all the way down it. The noise you hear is exactly 120 hertz. That's the frequency at which electron pulses are generated underneath this building, 120 pulses a second. And this is the sound of the equipment that accelerates them to over 99.9999992% the speed of light.
And this lets you see electron clouds move around molecules, essentially? Right? So why would you care? Because essentially electrons create the fields in which everything else happens. Molecular bonds break and form because the electrons essentially give them a push to do so. Right, so the electrons are responsible for everything that you see in nature, and being able to look into their motion is the most fundamental way of studying materials and matter.
Now, to achieve this, you need a nanoscopic equivalent of a strobe. So you first feed these relativistic electron pulses through a set of devices called undulators. They're stacks of magnets spaced only a few millimeters apart, with alternating poles. So the first pair has the north pole facing the electron pulse from above and the south from below. Then the second pair flips and so on. Now because the electrons are traveling through a magnetic field, a force called the Lorentz force will push them in a direction perpendicular to both their velocity and the magnetic field lines, in accordance with the left hand rule. So at one magnet pair, the electrons will curve clockwise, and at the next pair counterclockwise and so on. This causes the electrons to wiggle. Now since the electrons carry a charge, this wiggling motion causes them to emit electromagnetic radiation. And even though they oscillate at these millimeter wavelengths because of the magnet spacing, the wavelength of the resulting EM radiation is much smaller.
This is the fun thing about the theory of relativity. If you would travel at near the speed of light, the length scales contract. While that periodic structure is macroscopic for us, right? We see each magnet, these are centimeter scales. For the electron, because it's traveling so fast, all of that space contracts. And so it's actually oscillating really, really fast. And these oscillation periods are compressed. And it means that if you compare that to wavelengths, that is in the x-ray domain.
Now that actually only gets your wavelength part of the way to the true x-ray regime. But if as an observer you stand at the far end of the accelerator, all those electrons will be coming at you at more than 99% the speed of light. So in your reference frame, any light those electrons emit will additionally be blueshifted, producing x-rays as small as 50 picometers in wavelength.
Initially, these x-rays are created randomly along the undulator, producing an incoherent light pattern. But soon after, the electric fields from the x-rays start to interact with the electrons, speeding some of them up and slowing others down. This causes faster electrons to catch up with slower ones. So they get bunched up into periodic structures, parallel sheets that are spaced at distances exactly equal to the wavelength of the X-rays. This is called microbunching. Now these sheets of electrons emit light as unified fronts. So the resulting X-rays come out coherently as a laser pulse. This dramatically increases their intensity. And the pulses come out incredibly tightly packed, being only a few femtoseconds long, and they can get as short as a couple hundred attoseconds. That's 10 to the power of negative 18. An absurdly quick pulse. To put it another way, the attosecond is to the second what the second is to the age of the universe. On an attosecond scale, you can see electrons zip around essentially atoms and molecules. That's insane. Yeah.
After the undulators the x-ray pulses are sent to experimental stations at the end of the tunnel. So where's the main x-ray beam? The main x-ray beam is coming through this tube over here. Okay. If you wanted to... It's a little harder to see, but coming in through this tube over here. So this is the main place where the x-rays come into the hutch. And so the x-rays focus into what we call an interaction point. Now you fill this interaction point with the molecules whose electrons you want to study.
Now we shine that x-ray pulse on a molecule, and when it hits the molecule, it will ionize the molecule, and it ionizes predominantly from these inner shells, from the very core parts. The thing is, core level electrons from different elements have different ionization energies. So if you want an x-ray to eject a core level electron from a nitrogen, it needs around 400 electron volts. Whereas for an oxygen, it needs around 550. So by tuning the x-ray energy to match these ionization energies, you get to choose which of these atoms within the molecule the x-ray is going to ionize. And any excess energy left after an x-ray has ejected an electron will be taken by that electron as kinetic energy.
Now, once you ionize this molecule, the kinetic energy will tell you something about what's going on around that electron. Well, electrons are not independent particles, they talk to each other, right? They have a negative charge. So if you have a high electron density around a particular atom, the core level electrons will be bound less tightly to the nucleus because of the presence of all these other electrons around the atom. So its ionization energy will actually be slightly lower. Whereas if you have a lower electron density than average around an atom, those core level electrons will be bound more tightly with a higher ionization energy. Therefore, when you measure the kinetic energy of the electrons you eject, you can use the difference between the input x-ray energy and the output kinetic energy to infer what that electron density was.
Now, once we can take a picture of an electron density, we can change what the molecule is doing, make it do some process in time, and then we can look at how electron densities change. We have above us, we actually have an entire laser hall. So we generate laser light, we bring it down through tubes like here behind you or over here on the ceiling. Traditional laser light. These are infrared lasers. We bring them onto this table. And you see all these boxes on this table? These boxes are to condition the laser, to give it the properties that we want. We can change the color of the laser, we can change the polarization state, we can change the duration of the pulse. So we sculpt these pulses, then we have them go co-propagating with the x-rays into our target.
And so, now our first laser pulse will create some non-equilibrium state in the molecule, will drive some dynamics, and then our x-ray pulse will probe it. This attosecond x-ray pulse ejects an electron from the molecule after a time delay t. And by measuring the kinetic energy of the electron, you can study how the electron density of the molecule reacted to the trigger laser. So you get an attosecond snapshot, like a strobe, of how the molecule changed. Then you can get another sample of the same molecule, shoot it with a laser again, but this time increase the probe delay slightly to t plus delta t. This will tell you how the electron density changes a little later. And you can keep increasing this delay each time to get a sequence of snapshots of how that electron density evolves over time.
Isn't there a big assumption here that every time you do it, you're expecting a repeatable result from the molecules? Yes, absolutely, so you need for your initiator to drive the same dynamics over and over again. If you have a new process happening every time, this technique will fail. But if the scene is repeatable, then like the trillion FPS camera, you can use this technique to create a molecular movie. And here, the smallest amount by which you can tweak the time delay for this x-ray strobe is around 300 attoseconds. So you can get frames that are only a few hundred attoseconds apart. And if you stitch those together, you get a movie that technically runs at over a quadrillion frames per second.
So this is a movie of the dynamics that we might like to image. So this is a small molecular system. This is called para-aminophenol. This calculation was done by some of our collaborators in Madrid. They had calculated what is the response of this molecule to the removal of an electron. So they simulate an x-ray pulse coming in and removing an electron. So the red color here represents an increase in the density of the electron, and the blue represents a decrease. And so we see that when we've shined this x-ray pulse onto this molecule and we've removed an electron, we initiate some charge distribution that starts to move across the molecule. And so we wanna image this charge motion.
The video is a simulation, but it's been validated by the experiments done at SLAC. Our method for probing these seems to work. We can compare to a handful of points and say, oh, these look broadly similar. Much after this few femtosecond, as we approach five to 10 femtoseconds, we start to diverge. Our prediction starts to diverge from our measurement. And actually this is the most exciting time in science, right? When you have a prediction, and then you have a measurement, and they don't agree, that's when you get really excited, because you just found something you didn't know ahead of time. You couldn't have predicted that.
I think the most powerful thing for me here is we animate a lot of electrons, right? And pretty much every Veritasium video has electrons moving in some way. So the fact that we're actually seeing these electron densities move around… I don't know, I think it's spectacular. Absolutely.
