Synaptic plasticity investigation in permalloy based channel material for neuromorphic computing

Abstract

Abstract Artificial synaptic devices capable of synchronized storing and processing of information are the critical building blocks of neuromorphic computing systems for the low-power implementation of artificial intelligence. Compared to the diverse synaptic device structures, the emerging electrolyte-gated synaptic transistors are promising for mimicking biological synapses owing to their analogous working mode. Despite the remarkable progress in electrolyte-gated synaptic transistors, the study of metallic channel-based synaptic devices remains vastly unexplored. Here, we report a three-terminal electrolyte-gated artificial synapse based on metallic permalloy as the active layer. Gating controlled, non-volatile, rewritable, and distinct multilevel conductance states have been achieved for analog computing. Representative synaptic behaviors such as excitatory postsynaptic conductance, paired-pulse facilitation, spike amplitude-dependent plasticity, spike duration-dependent plasticity, and long-term potentiation/depression have been successfully simulated in the synaptic device. Furthermore, switching from short-term to long-term memory regimes has been demonstrated through repeated training. Benefitting from the short-term facilitation, the synaptic device can also act as a high-pass temporal filter for selective communication. This research highlights the great potential of metallic channel-based synaptic devices for future neuromorphic systems and augments the diversity of synaptic devices.