Modeling and Optimization of Single Photon Avalanche Photodiodes for X-Ray Detection

Avalanche Photodiodes (APDs) are electronic devices that transduce a photon flux into an electrical current and provide internal amplification of this current exploiting the impact ionization mechanism. APDs are used as receivers in optical fiber communication links as well as detectors in physics experiments and medical imaging. According to the needs of the target application, they can either be operated below (Linear mode) or above (Geiger mode) their breakdown voltage. For X-ray detection, APDs fabricated in III-V compound semiconductors, such as GaAs, offer an interesting alternative to SiAPDs, thanks to the short attenuation length at high energies offered by these materials. However, to improve the poor noise performance of APDs fabricated in III-V compounds given by similar electron’s and hole’s impact ionization coefficients, structures alternative to p-i-n APDs have to be employed. A possible solution is the use of staircase APDs, where heterojunctions between III-V compound semiconductors and their alloys with metals are exploited to enhance the electron to hole impact ionization probability by creating an artificial superlattice. This thesis aims at proposing models to compute the figures of merit of APDs fabricated in III-V compound semiconductors and operating in Linear mode for the detection of X-rays. Since accurate modeling of impact ionization is key to obtain reliable data from simulations, we present the development of a suite of simulations tools that includes finite difference and a Random Path Length algorithm implementation of a newly derived nonlocal history dependent impact ionization model and a Full Band Monte Carlo transport simulator. All these models have been validated against experimental results and are thus powerful tools in support of the interpretation of single photon APDs electrical measurements and for the optimization of their performance. These simulation tools have been used to compute the gain, the excess noise factor, the response time, the bandwidth and the jitter of different APD structures, including staircase APDs. In addition, the Full Band Monte Carlo transport simulator has been employed to assess the basic assumptions, identify the limitations and improve the calibration of nonlocal history dependent impact ionization models. We have found that, even though nonlocal history dependent models give results that are in a satisfactory agreement with experiments, they neglect that after an impact ionization event secondary carriers are generated with non null kinetic energy and that carrier-phonon scattering may lead to electrons and holes that travel for few free flights with velocities that are opposite to the direction of the electric fields. These aspects may become relevant and yield misleading results, in particular for short devices.