Finite Element Modeling of Synaptic Plasticity in Mushroom-type Phase Change Memory Devices for Application in Neuromorphic Systems

Abstract

Neuromorphic Computing is part of a broader subfield called in-memory computing, targeting to achieve post-von Neumann architectures, under which computing is done insitu, with the strengths of the synaptic weights stored and adjusted directly in memory analogous to a biological Synapse. Because of their reversible and rapid phase transitions across different conductance levels in a tunable way, Phase change materials have exhibited this kind of plasticity, proving their potential to mimic biological synaptic weight update. In this work, we perform an extensive finite-element simulation on a typical Ge2Sb2Te5 (GST), a chalcogenide phase change material, mushroom cell memory device to imitate the plasticity of biological synapses. The model presented here provides significant insights to a better understanding of how these devices behave and the simulation results are benchmarked against empirical measurements.