Bipolar Resistive Switching in TiO₂ Artificial Synapse Mimicking Pavlov's Associative Learning

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Memristive devices are key for artificial synapses in neuromorphic computing (NC).
• They offer a potential alternative to traditional von Neumann architectures.
• Titanium dioxide (TiO2) based memristors are explored for their synaptic capabilities.

Purpose of the Study:

• To investigate Ag/TiO2/Pt memristors for NC applications.
• To demonstrate both digital and analog switching behaviors in TiO2 memristors.
• To emulate essential synaptic functions and complex neural behaviors.

Main Methods:

• Fabrication of Ag/TiO2/Pt memristive devices using pulsed laser deposition.
• Characterization of digital and analog resistive switching behaviors.
• Emulation of synaptic functions (pair-pulse facilitation, LTP, LTD) and learning paradigms (STDP, Pavlovian conditioning).

Main Results:

• TiO2 memristors exhibit robust digital switching with a memory window of ~10^3.
• Multiple analog conductance levels were achieved, supporting bio-inspired synapse development.
• Successful emulation of synaptic plasticity and associative learning, with 95.98% accuracy in MNIST pattern recognition.

Conclusions:

• The Ag/TiO2/Pt memristor demonstrates simultaneous digital and analog switching, crucial for advanced NC.
• The device successfully mimics biological synaptic functions, paving the way for efficient, low-power neuromorphic systems.
• This research contributes to the advancement of TiO2-based resistive random-access memory for practical NC implementations.