How do you harmlessly reveal bombs that can be triggered by a single photon? With the Elitzur-Vaidman bomb tester. It's a hypothetical tester meant to test some highly hypothetical bombs. In theory, it makes a point about quantum mechanics. In practice, it actually works.
Imagine a group of bombs that blow up with the detection of a single photon, and now imagine the photon detectors on these bombs are not particularly good. Some work, and some don't, making some of the bombs dangerous explosives and some duds. If you had no further use for the bombs, simply getting to a safe distance and shining light on the detectors would be enough. The duds would just sit there and the live bombs would explode.
But what if you wanted to save some of those live bombs? The task would seem impossible. You can't know if the detector works without testing the trigger, and you can't test the trigger without exploding each and every live bomb. At least, that would be how it would work with conventional bomb triggers. It's not quite the same with single photons.
Single photons obey the rules of quantum mechanics, and the rules of quantum mechanics say that, until the photon hits a working detector, it goes through all possible paths. The bombs have working detectors, and the duds don't. The Elitzur-Vaidman bomb tester makes use of this.
This is a diagram of the tester. Light is emitted from source A. It's bounced around by a series of mirrors. The mirrors in front of A, and in front of C and D, are half-silvered, which are often used in lasers. They bounce half the photons that hit them backwards, and allow half the photons that hit them to travel on. In the diagram, you can see the red path of the light travelling through both of the half-silvered mirrors. Finally, there are three possible photon detectors: C and D are permanent photon detectors, while the red B represents the placement of the bomb. This might be a photon detector, but only if the bomb is not a dud.
Here's the catch. If B weren't there, the path traveled by the light is slightly different depending on if it takes the high "road" toward C and D, or if it takes the low "road." When A emits light, the wavelengths of light traveling along the low road and going up towards C, through the half-silvered mirrors, will be slightly out of phase with the wavelength taking the high road up towards C. They will be out of phase just enough that the peaks from one wave will sync up with the troughs of another, and the waves will interfere with each other and cancel each other out. The detector at C will not detect the light. However, the light that travels toward D, using both the high and the low road, will be in sync, so that the peaks of the wave line up, emphasizing each other, and the detector at D will detect a photon.
And this is exactly what happens if the bomb is a dud. When the bomb is a dud, then the B detector isn't working. Because in quantum mechanics, an undetected photon can and does take all possible paths, even a single photon emitted by A takes both the high and low road. It both goes through and is bounced back by both half-silvered mirrors. The single photon causes its own interference pattern, and is detected at D. The bomb is a dud.
Of course, things get a little more complicated if the bomb is not a dud. If there's actually a working detector on the bomb, then the photon can't take both the high and low roads — it hits the half-silvered mirror in front of A and can either take the low road or the high one. If it takes the low road, it is detected by B and the bomb explodes.
On the other hand, each photon has a fifty percent chance of taking the high road. But unlike the scenario in which the bomb was a dud, there is no photon taking the low road. Because there is a working detector at B, there is only one light wave, and it's only on the high road, meaning it can't interfere with itself. The peaks and troughs can't emphasize or destroy each other when the photon gets to the half-silvered mirror in front of detectors C and D. The photon can only hit the detectors at C or at D. Since, if the bomb is a dud, a photon cannot be detected at C due to the photon interfering with itself, anytime a photon is detected at C, we know that B is a live, working bomb.
Through this process, the Elitzur-Vaidman tester can save 25% of the live bombs in the first round of testing. In 1994, a simulation of this — not using actual bombs — showed that it did work. Because of the rules of quantum mechanics, and the fact that photons take every possible path until they are detected, we can prove that certain bombs are live without ever actually interacting with them.