Development of ‘Artificial’ Memristive Synapses Using Various sp2 C (Graphene-Like) and sp3 C (Diamond) Heterojunctions as Neuromorphic Devices

Inspired by human brain functionality and its low power consumption (10 W), memristors for neuromorphic computing have gained significance for implementing solid-state neurons and synapses due to their nanoscale footprint and reduced complexity. We report the fabrication of various carbon-based heterojunctions comprising graphene-like (sp²C)-diamond (sp³C) interfaces using microwave plasma-assisted chemical vapor deposition as “artificial” synapses, the key elements mimicking the characteristics of biological synapses and memory functions, that are game-changing energy saving devices. The resulting heterojunctions behave as memristors (i.e., resistors with tunable memory) having multiple resistance states and nonvolatile memory functions, a phenomenon that refers to the ability of synapses (neuronal links) to adapt in response to an increased or decreased activity, essential to human memory and learning. We performed I–V characteristics in response to photoirradiation at 300 nm, 532 nm, and 633 nm from laser emitting diodes and temperature (up to 250 ^oC) exhibiting linearity and symmetry when subjected to identical input pulses, essential for their role in online training of neural networks. Interestingly, high or low resistance states (equivalent short-term and long-term potentiation) can be controlled by combined bias voltage and irradiation, giving a resistive switching ratio of ~10³, observed in sparse materials and/or heterostructures. We attribute the observed behaviors to redox reaction at the sp²-sp³ interfaces and the role of hydrogen and oxygen movement by bias. Finally, heterostructure arrays could be usable as electrical and photo-controlled devices with potential switching, photo sensing (image sensors), and memory functions.