Avalanche Photodiode

A reverse-biased silicon junction so hungry for current that a single photon sparks an avalanche of thousands of electron collisions — turning light into an electrical shout.

Converts photons into electrical current with built-in gain, making it sensitive enough for detecting weak light in laser rangefinders, autonomous-vehicle LIDAR, fiber-optic receivers, and quantum cryptography systems. Unlike a standard photodiode that produces one electron per absorbed photon, an APD uses a high reverse-bias voltage to accelerate the initial carrier into a high-field multiplication region, triggering a controlled avalanche of impact ionizations that delivers 50–500 electrons per photon in linear mode. Silicon APDs cover 400–1000 nm at gains up to several hundred, while InGaAs/InP APDs extend coverage to 1300–1550 nm telecom wavelengths at lower gains of 10–30.

The reach-through APD structure (RAPD) is built as a p⁺ / π / p / n⁺ vertical stack. Light enters through the p⁺ contact and is absorbed in the wide, lightly doped π region where the electric field is low (~3×10⁴ V/cm) — just strong enough to sweep carriers apart without triggering multiplication prematurely. Photo-generated electrons drift down into the narrow p multiplication layer where the field exceeds 2×10⁵ V/cm, surpassing the impact-ionization threshold. There, a drifting electron collides with the lattice, generating a new electron-hole pair; each daughter carrier also ionizes, producing a cascade. The gain M is described by McIntyre's 1966 formula M = 1/(1−⟨α(E)⟩·W), where α is the electron ionization coefficient and W is the multiplication-layer width. Because silicon has α_e ≫ α_h (electrons ionize far more readily than holes), the avalanche is efficiently initiated by electrons alone, giving a lower excess noise factor F(M) ≈ 2 + k_eff·M with k_eff < 0.1. InGaAs/InP APDs use a separate absorption–charge–multiplication (SACM) structure to confine the high field to the InP multiplication region while absorbing photons in the lower-gap InGaAs layer.

In plain terms

Imagine rolling a marble down a hill covered in sticks. The marble knocks one stick loose, that stick knocks two more, those knock four, and so on — suddenly you have a whole avalanche from one little marble. An APD does the same thing with electrons: one photon starts the chain reaction, and thousands of electrons come tumbling out the other end.

Also called: APD, linear-mode APD, reach-through APD, RAPD, InGaAs APD, Si APD.

Why it matters

Where it lives

Key numbers

Surprising facts

Open questions

A short history

The theoretical framework for avalanche multiplication noise was established by R. J. McIntyre at RCA Laboratories in 1966, deriving the excess noise factor F(M) that still governs every APD design today. The first practical silicon reach-through APD was developed at RCA in the early 1970s, capitalising on the low hole-ionization rate in silicon identified by McIntyre. Webb, McIntyre, and Conradi published the foundational properties paper in RCA Review in 1974. InGaAs/InP APDs for the 1550 nm telecom band were pioneered in the 1980s by Bell Labs and Japanese groups as single-mode fiber networks expanded, leading to the separated-absorption-charge-multiplication (SACM) structure that isolates the high-field multiplication region from the absorbing InGaAs layer. By the 2000s, InGaAs APD receivers were standard in all long-haul fiber systems. The 2010s saw APD arrays integrated into solid-state LIDAR modules for autonomous vehicles — transitioning the APD from a telecommunications component to a key sensor in self-driving car perception stacks. John Campbell (University of Virginia) summarised the state of the art in his 2007 Journal of Lightwave Technology review.

Sources

Related