Simple mod turns diode into photon counter

Avalanche photodetection

Counting photons is a tricky business. They’re slippery beasts that arrive silently, often and in packs, in ways that are almost impossible to count.

One of the most widely used of devices that can spot the arrival of a single photon is the avalanche photodiode. These cheap and easy to use devices rely on the ability of diodes to allow the flow of electrons when the voltage across them is in one direction but prevent that flow when the bias is reversed. But  if the reverse bias is increased beyond a specific threshold then a breakdown occurs and a reverse current suddenly starts to flow.

Choose the right material for your photodiodes and this breakdown can be triggered by a the arrival of a single photon  smashing into an electron which goes on to hit other electrons causing a chain reaction. The result is an avalanche of current that signals the arrival of your photon.

Avalanche photodiodes are widely use to detect single photons but have an important limitation: they cannot distinguish between the arrival of a single photon and the arrival of two or more photon’s simultaneously.

But that is set to change. Today, our old friend Andrew Shields, at Toshiba’s research labs in Cambridge UK, explains how to soup up a bog-standard avalanche photodiode so that it can count photons as they arrive. That’s like turning a Fiat 500 into a Ferrari.

He says that the trick is to measure the characteristics of the avalanche current in the very first instants that it forms. At this early stage, say Shields and friends, the avalanche current  is proportional to the number photons that have struck.

Simple really but with enormous potential. The ability to count photons is one of the key enabling technologies for optical quantum computing. A number of schemes are known in which it is necessary to count the arrival of 0,1 or 2 photons at specific detectors.

Various people, including Shields himself, have  come up with complex, cooled devices that can count photons. But this is a major step forward. Avalanche photodiodes are cheap, widely available and easy to use. With such a cheap detector now available (as well as decent photon guns), we could see dramatic progress in this field in the coming months.

If you haven’t quite seen the significance of this, imagine overclocking your calculator and matching the performance of a workstation. Or polishing up the 3 inch reflector in your attic and outclassing Hubble with your images.

Impressive stuff.

Ref: arxiv.org/abs/0807.0330: An Avalanche-Photodiode-Based Photon-Number-Resolving Detector

5 Responses to “Simple mod turns diode into photon counter”

  1. George Myers says:

    I remember reading about these in infra-red tacheometers (or surveyor’s “total stations”) which time the arrival of photons from a corner cube prism use adjustments for the speed of light (altitude and temperature) and with the recorded angles from the circles in the theodolite portion calculate x,y,z coordinates of the target prism, today without a target or from reflective tape. I think of it as more like a “night vision” collector than what they were being called “lasers” though the LED diode was restricted to a specific frequency in infra-red sent through the telescope to the prism and back.

  2. RawThinkTank says:

    This will pave way clear for the photon CPU.

    Photons per second ?

  3. Marc Cabot says:

    This is an excellent advance in detector technology. Not having to deal with delicate crystals or high voltage photo multiplier tubes would sure help out a lot when building such high real-estate detectors such as gamma cameras. The only problem still remaining is the obvious: Avalanche detectors can’t provide energy discrimination.

  4. [...] copper, this discovery makes it cheaper to put glass fibre in your home. And as mentioned in the abstract it makes optical quantum computing more of a [...]

  5. Kent says:

    @RawThinkTank: I have never used avalanche diodes before but if I remember correctly, they have a very slow failing trail on the order of 5-10 microseconds. Judging by the description of the article, even with the new technique, single photons that arrive while a long trailing edge is still there won’t be detected properly. Hence you are looking at a maximum data rate of 10E4 photons/sec or 1kB/sec per diode. I can’t see a way of using several diodes in parallel without splitting single photons…