We report on a suite of modeling approaches for the optimization of Avalanche Photodiodes for X-rays detection. Gain and excess noise are computed efficiently using a non-local/history dependent model that has been validated against full-band Monte Carlo simulations. The (stochastic) response of the detector to photon pulses is computed using an improved Random-Path-Length algorithm. As case studies, we consider diodes consisting of AlGaAs/GaAs multi-layers with separated absorption and multiplication regions. A superlattice creating a staircase conduction band structure is employed in the multiplication region to keep the multiplication noise low. Gain and excess noise have been measured in devices fabricated with such structure and successfully compared with the developed models.
