This article was originally featured in the magazine:

Building Better Low-light Detectors


A stack of alternating layers of GaN and AlN creates an avalanche photodiode with high sensitivity and excellent controllability


If you need to measure incredibly low light signals, you have to make a compromise. If you want the best performance, you must select a photomultiplier tube. However, it is fragile and bulky. The alternative, addressing these weaknesses, is the avalanche photodiode (APD), but it is let down by its poor controllability.

The good news, however, is that this lack of controllability will not be plaguing the APD for much longer, thanks to a recent breakthrough by our group at Tsinghua University, working in partnership with researchers at CNRS-CRHEA, France, and the Chinese Academy of Sciences. Together we have developed a novel APD that delivers a record-breaking stable linear gain of over 104 under constant bias – that’s almost two orders of magnitude higher than that of silicon APDs.

With conventional APDs, producing high multiplication gain requires operation under extremely high bias. When controlled in this manner, there is a drastic rise in the response current, known as ‘breakdown’. This breakdown will not stop until the bias is reduced below its breakdown value.

What this means in practice is that when using an APD, it needs to be periodically quenched below its breakdown bias, with photons only detected when these diodes are under breakdown. It is a compromise between high gain and sustainable detection, and requires the use of complex control circuits. This intrinsic limitation has been plaguing researchers for years, who want an APD that works like a PMT, with a controllable avalanche and the opportunity to operate the device under a simple constant bias with high gain. This is what you get with our device.

It is important to note that avalanche and breakdown are actually two completely different concepts. Breakdown is a compromise, used to ensure that a conventional APD delivers high gain. In this type of device, a significant proportion of the carrier energy that’s drawn from the electric field is thermalized by intense scattering. As a result, these diodes have to work under extremely high bias, because this enables electrons and holes to both trigger ionization (see Figure 1(a)). Multiplication proceeds along two opposite directions, forming positive feedback chains that lead to a drastic rise in the response current.

A better approach would be for just one kind of carrier to trigger ionization, leading to an avalanche march in just one direction (see Figure 1 (b)). If that were the case, the avalanche could stop by itself when all the carriers move out of multiplication region. Thanks to this, the APD could provide a suitable response to the signal when it is operated under constant bias, simplifying its use and the accompanying circuitry.

Our device works on this principal, as do several others. In fact, in general, there are three conventional approaches to improving the sensitivity of linear-mode APDs: using materials with a great difference in conduction band and valence band profiles, narrowing the ionization region, and turning to impact ionization engineering.