pipette
ESEspañol

Multifunctional synaptic learning and neuromorphic computing using crystalline TiO/NiO heterojunction-based memory devices

Dip Manna, Koushik Mondal, Subhajit Karmakar, Pawan Kumar Ojha, Sourav Roy, Biswarup Satpati, Prasanta Karmakar, Supratic Chakraborty

PreprintBold claims, read criticallyReal-world use

In the authors' words

An attempt is made here to mimic different properties of biological synapses using Ag/TiO/NiO/p-Si memristor structure by studying its different transport properties under dc and pulsed bias. The presence of crystalline NiO with smaller gains is found to be helpful to get TiO deposited on its top with crystalline properties and larger grains. The heterostructure offers stable bipolar forming free non-volatile resistive switching characteristics under reverse biased condition with gradual set and reset features. These devices are also able to successfully implement the classical Pavlov's learning, study artificial nociceptor and Morse code detection. While incorporating synaptic weights derived from the measured conductance programming pulse relationship, an artificial neural network achieves nearly 95% accuracy in MNIST digit recognition. In brief, NiO/TiO heterojunction plays the pivotal role in getting such reproducible characteristics for mimicking different properties of biological synapses.

Main resultThe abstract does not state a limitation.

Appeared: Friday, October 9. arXiv. Preprint, not yet peer-reviewed.