Atomistic Origins of Compound Semiconductor Synthesis with Computational Neuromorphic Engineering

Abstract

We propose the usage of multi-element bulk materials to mimic neural dynamics instead of atomically thin materials via the modeling of group II–IV compound semiconductor growth using vacancy defects and dopants by creating and annihilating one another like a complex artificial neural network, where each atom itself is the device in analogy to crossbar memory arrays, where each node is a device. We quantify the effects of atomistic variations in the electronic structure of an alloy semiconductor using a hybrid method composed of a semiempirical tight-binding method, density functional theory, Boltzmann transport theory, and a transfer-matrix method. We find that the artificial neural network resembles the neural transmission dynamics and, by proposing resistive switching in small areas with low energy consumption, we can increase the integration density similar to the human brain.