Impact of Low-Voltage Resistive Switching Devices on Neuromorphic Computing

The human brain is the most efficient machine around us in size, power efficiency, self-learning capability, decision-making, and simultaneous data storage and processing. Additionally, our brain performs all operations analogously by consuming energy of about 1 – 100 fJ per synaptic event. Even though the conventional computer works much faster than the brain, the von Neumann bottleneck and memory wall issues limit the performance and energy efficiency due to the physically separated storage and processing unit. After discovering the memristor (MR), a two-terminal device with multiple conducting states at a particular bias voltage, efforts are already underway toward developing machines similar to brain functionality. Despite massive progress in semiconductor technology, it is still challenging to mimic the functionality of synapses and neurons, the basic building blocks of our brain. Memtransistor (<i>mem</i>T), a gate-controlled memristor or memory transistor, is also coming up rapidly to the limelight for mimicking functionalities of synapses and neurons more controlled way as the building block of the artificial brain. Among various approaches, the redox-controlled MR and <i>mem</i>T are becoming very attractive to accomplishing the desired metrics for developing efficient neuromorphic computing tools. In this talk, I will introduce a few unconventional redox reaction-dependent molecular MR and organic <i>mem</i>T devices that are very efficient for data storage and mimic various synaptic and neural functions electronically.