Tutorial : Avalanche Photodiodes Theory And Applications

By: Tim Stokes
General Sales Manager Hamamatsu Photonics UK Ltd.

Avalanche Photodiodes ( APDs ) are high sensitivity, high speed semi-conductor "light" sensors. Compared to regular PIN construction photodiodes, APDs, have an internal region where electron multiplication occurs, by application of an external reverse voltage, and the resultant "gain" in the output signal means that low light levels can be measured at high speed. Incident photons create electron – hole pairs in the depletion layer of a silicon photodiode structure and these move towards the respective PN junctions at a speed of up to 105 metres per second, depending on the electric field strength. If the external bias increases this localised electric field to above about 105 V / cm then the carriers in the semi-conductor collide with atoms in the crystal lattice, and the resultant ionization creates more electron – hole pairs, some of which then go on to cause further ionization giving a resultant gain in the number of electron – holes generated for a single incident photon (See schematic below).

Most commonly available APDs are fabricated from silicon and employ a so called "reach through" structure where light is incident from the N-side of the silicon. These devices show useful sensitivity in the 450 nm to 1000 nm wavelength range, such as the S6045 series from Hamamatsu Photonics. It is possible to fabricate devices where light is incident from the P-side, such as the S8664 series from Hamamatsu Photonics, and these then exhibit high sensitivity to UV – blue light and operate in the range from 200 nm to 800 nm.

APD gain is typically in the range from x10 to x300 for most commercial devices, but there are APDs available from specialist manufacturers with gains of thousands. This then can give a significant advantage over regular PIN photodiodes for applications which are short of photons and where it is not possible to integrate these low signals. As with regular photodiodes the maximum wavelength than can be detected is determined by the semi-conductor band gap energy using the formula: which is 1.12 eV for silicon at room temperature, giving a cut-off at 1100 nm. At longer wavelengths then an alternative semi-conductor material with smaller band gap is required, such as Germanium, or much more commonly these days due to its higher performance, InGaAs is chosen. Avalanche Photodiodes fabricated from these materials are then available in the market for operation in the 900 nm to 1700 nm wavelength range. A wide range of silicon APDs are commercially available, in sizes from <100 microns diameter to several cm diameter, and these days in a variety of packages, from TO metal cans, to carriers and now even on surface mount substrates such as the new Hamamatsu Photonics S9717 series. The range of commercial Infrared APDs available is however much smaller than for silicon; InGaAs APDs, such as the Hamamatsu Photonics G8931, having small area ( 30 micron diameter ) since they are used predominantly for fibre applications such as telecommunications.

As it is a relatively thin layer within the APD structure that gives rise to the "gain", the peak wavelength for silicon APDs tends to be from 600 nm to 800 nm, somewhat shorter than the 900 nm to 1000 nm peak wavelength for a regular photodiode. Deeper depletion silicon APD structures are then available for operation in the 900 nm to 1100 nm waveband range, such as the S8890 series from Hamamatsu Photonics, but these generally have the disadvantage of requiring a much higher reverse voltage to create the high electric fields needed and consequently they have much higher dark currents. All semi-conductor devices have such an associated dark current caused by thermal ( rather than optical ) generation of electron – holes. In practice then the shot noise associated with this dark current ultimately will limit the minimum amount of light that any device can detect. Thermo-electric cooling can then reduce the dark current and thus improve the range of incident light that can be measured. In an APD dark current is generated both from leakage at the surface of the diode and also from electron – holes thermally generated within the bulk of the silicon which are then multiplied in the gain region. Consequently increasing the gain of the APD, by increasing the external bias, also increases this dark current.

The APD multiplication process also produces an additional noise component, known as "excess noise" since the ionization of any individual carrier has a certain probability of occurance, the overall gain from the device being the statistical average of all of these individual ionization events. APD noise is given by the formula:

The consequence of this is twofold.

As the APD gain increases the output signal increases linearly, but the noise increases as shown in the graph below. This means for any APD there is an optimum operating gain, usually well below the actual maximum gain for that APD, where the maximum signal to noise performance can be obtained. Manufacturers then supply APD modules where the performance of each individual APD is optimised and set-up at the factory prior to supply, such as the Hamamatsu C5331 and C5460 devices.

It is apparent that the shot noise of an APD is higher than that for a comparable performance photodiode, so even though the APD gives an amplified output the overall signal to noise performance ( SNR ) is not necessarily improved. In order for a regular photodiode to detect lower light levels it is usual to increase the gain in the operating circuit by increasing the feedback resistor value. This has the unwanted consequence of reducing the speed of response and increasing the thermal noise associated with the operating circuit. In contrast, operation with an APD allows for the gain to be increased to improve the SNR whilst maintaining the speed of response, until the shot noise reaches a level equivalent to the thermal noise.

Due to their performance advantages APDs are then used widely in applications such as distance measurement, data transmission ( over fibre or through free space ), range finding, high speed industrial inspection ( including colour measurement ) and in various other medical and scientific instrumentation.

Providing the noise of the APD device is low enough, then it is also possible to operate an APD is Geiger mode as opposed to analogue operation, described above, to detect individual incident photons. The APD has to be operated at a few volts above its breakdown voltage with extremely stable operating conditions such as the APD power supply, temperature, etc. else the noise of the detector will "run away". This means that for some applications such photon counting APDs are these days also starting to be used over more established Photomultiplier Tube ( PMT ) technology, due to the higher quantum efficiencies of the semi-conductor device. We should add a note of caution here however as such highly stable, highly sensitive APD systems are often more expensive than a comparable PMT based system, and such low noise APDs are generally only hundreds of microns ( or smaller ) in size, thus very often more light is lost in the optical collection system than may be gained from the higher quantum efficiency of the detector itself ! For the majority of instrumentation based applications, the larger detection area, higher gain and superior SNR of the PMT make it still the detector of choice for many years to come.

SOURCE: Hamamatsu Corporation