To watch atmospheric and astronomical phenomena that normally unfold too slowly for convenient observation—from moving clouds and stars to variable stars and novae—we often use time-lapse photography, compressing a long event into seconds or minutes. For phenomena that occur too quickly for human eyes to follow in detail, such as lightning, we instead need high-speed imaging that lets us watch the event in slow motion. [source] [source]
 |
One of Harold Edgerton's best-known photographs shows a bullet passing through an apple. Taken in 1964 with an exposure of roughly one millionth of a second, it became an iconic example of high-speed photography. Edgerton, an MIT electrical-engineering professor, pioneered techniques using repeated flashes of light to freeze extremely rapid events. [source]
Modern slow-motion cameras record large numbers of frames per second. Consumer products were already becoming capable of around 240 frames per second when this article was written. But to study something as fast as a bullet, ordinary consumer cameras are insufficient. A camera capable of at least 10,000 frames per second, the source explains, can make a projectile travelling near 300 metres per second appear to move at only a few centimetres per second when played back slowly. [source]
What about light itself? [source] [source]
Now imagine trying to observe something vastly faster: light. A conventional high-speed camera is nowhere near sufficient, and photographing propagating light differs fundamentally from photographing an ordinary moving object. A bullet is illuminated and the light reflected from it is recorded. To visualize a light pulse itself, photons arriving from repeated laser pulses must instead be detected and reconstructed.
The article gives an illustrative calculation that to slow the apparent motion of light to something like the bullet example would require on the order of hundreds of millions of frames per second. Fortunately, specialized technology can go vastly beyond that. [source] [source] [source]
In 2011, roughly half a century after Edgerton's pioneering work at MIT, a team at the MIT Media Lab led by Ramesh Raskar developed a method capable of reconstructing observations at an effective rate of approximately one trillion frames per second.
Femto-photography
The researchers called the method femto-photography, or ultrafast photography. The system combines an ultrafast pulsed laser, a streak camera capable of resolving extremely short time intervals along one spatial dimension, and mathematical reconstruction that converts repeated one-dimensional measurements into a two-dimensional representation.
 |
| Image from the original article |
A titanium-sapphire laser produces a very narrow light pulse, roughly a millimetre across and lasting only about a trillionth of a second in the description given by the source. Photons returning from the scene are detected using a specialized streak camera, an instrument more commonly used in chemistry and biology than for ordinary outdoor photography. [source] [source]
Light enters the streak camera through a narrow slit, limiting the incoming image to one spatial dimension. The photons are then deflected so that photons arriving earlier strike a different position on the detector from photons arriving later. Thus the device encodes arrival time as position. [source] [source]
Because each exposure contains only a one-dimensional slice, the researchers repeat the experiment while changing the viewing geometry with a mirror. A numerical algorithm then combines many such slices into a two-dimensional reconstruction. In the resulting visualization, the light pulse appears as a band travelling through the scene. [source]
The article describes a nanosecond-scale demonstration in which viewers can watch light travel through a bottle and see a tomato's shadow develop on a wall as illumination reaches it.
Seeing around corners
Raskar also discussed another potential use: cameras capable of reconstructing objects hidden around a corner and outside direct line of sight. The article refers to this approach as CORNAR. Light can bounce from a visible surface into the hidden region, reflect from an unseen object, and return indirectly; precise time-of-flight measurements can then be used to infer the hidden geometry. The source also mentions possible future biomedical applications, such as viewing internal structures without conventional X-ray imaging.
 |
| Image from the original article |
Even faster imaging [source]
The article then notes that researchers at two Japanese universities had developed a system capable of recording the equivalent of 4.4 trillion frames per second at about 0.2-megapixel resolution (450 × 450 pixels). At such effective rates, light could be visualized as though moving at less than a millimetre per second. [source]
The researchers used the technique to image periodic atomic-scale vibrations in crystalline material and to record heat transport—phenomena occurring at extraordinarily high speeds. The author ends playfully by imagining borrowing the Japanese camera for a few days and pairing it with a powerful atomic microscope. [source] [source]