IEEE study improves design of avalanche photodiodes for photodetection in the ultraviolet wavelength

Geiger-mode avalanche photodiodes (GM-APDs) are highly sensitive light detectors, capable of detecting single photons. Photons of certain wavelengths, when absorbed by photodiodes, generate electron-hole pairs in a process called impact ionization which can result in a multiplication of charges when occurring in an electric field. An avalanche photodiode is biased above their “breakdown voltage” at which point impact ionizations reach a self-sustaining rate, resulting in a distinct electrical pulse that is readily detectable. To detect single photons in the presence of other mechanisms that generate impact ionization, the avalanche diode must simultaneously have a high probability to absorb incident photons of the desired wavelength, known as the unity-gain quantum efficiency (QE). Both being able to support high fields and having good QE at the desired wavelength are critical factors in determining the device’s sensitivity.

Certain GM-APDs based on 4H-silicon carbide (4H-SiC) have high single-photon detection efficiency in the deep-ultraviolet (DUV) wavelengths around 280 nanometers. To reliably detect photons at higher wavelengths where absorption is weaker, SiC GM-APDs need to improve their baseline photon capture efficiency, as indicated by its unity-gain QE. To accomplish this, researchers often employ APDs with much thicker absorber layers. However, this can often lead to design challenges.

In a study that was recently published in the IEEE Journal of Quantum Electronics, Dr. Jonathan Schuster from DEVCOM Army Research Laboratory, United States, and his team have developed a numerical model with a calibrated 4H-SiC material library for the development of APDs in higher wavelength ranges.

“APDs with much thicker absorber layers (10’s of microns) must be utilized to improve the NUV response, which necessitates switching from a conventional PIN architecture (usually <3μm thick) to a separate-absorption charge-multiplication (SACM) architecture. However, this involves unique challenges such as deviating from existing front-side absorber SACM architectures to a very thick backside one,” explains Dr. Schuster.

Using the newly developed numerical model, the researchers designed SACM APDs that are predicted to have high single-photon detection efficiency in the NUV range. The researchers considered two architectural designs—non-reach-through (NRT) and reach-through (RT), with each having distinct design considerations. Notably, they have designed NRT-SACM APDs with unity gain QE of up to 32% and RT-SACM with unity gain QE of up to 71% for photons with a wavelength of 340 nm. These designs also maintain a large electric field in the multiplication layer for Geiger-mode operation. The improvements in QE can translate to more diverse application of APDs in high wavelength photodetection.

“For the NRT-SACM case, it was determined that the doping profiles must be engineered such that two competing mechanisms are balanced: maximizing the minority carrier diffusion length in the absorber layer (AL) while minimizing the corresponding potential barrier at the AL/charge layer (CL) interface,” notes Dr. Schuster. He further adds, “Conversely, in a RT-SACM architecture, it was determined that a narrow range of total charge in the CL properly modulated the electric field to be non-zero in the AL and sufficiently large in the multiplication layer to operate above avalanche breakdown.”

The study has identified several design rules that can be leveraged to design GM-APDs for single photon counting applications in the NUV wavelength range. Crucially, the researchers have determined that CL designs in APDs are inflexible with respect to deviations in layer thickness or doping which increases the difficulty of fabricating them.

4H-SiC avalanche photodiodes have a wide variety of applications in ultraviolet photon detection such as solar-blind ultraviolet detection, combustion monitoring, and environmental ultraviolet monitoring. In the future, the numerical model developed here can be instrumental in designing more sensitive and efficient APDs, significantly advancing their applications.