Biopolymer based artificial synapses enable linear conductance tuning and low-power for neuromorphic computing

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Neuromorphic computing offers a solution to the von Neumann bottleneck.
• Biomaterial-based artificial synapses are key for neuromorphic systems but face linearity and state limitations.

Purpose of the Study:

• To develop a linear, high-performance biomaterial-based memristor for neuromorphic computing.
• To enhance the functionality of natural biomaterial-based artificial synapses.

Main Methods:

• Fabrication of a silver nitrate (AgNO3) doped iota-carrageenan (ι-car) memristor.
• Characterization of memristor linearity, endurance, conduction states, and power consumption.
• Application of the memristor in deep learning for handwritten digit recognition.

Main Results:

• The AgNO3-doped ι-car memristor demonstrated linear conductance tuning.
• Achieved high endurance (~10^4), over 2000 conduction states, and low power consumption (~3.6 μW).
• Suppression of Ag filament formation by AgNO3 doping improved device stability and reduced Joule heating.

Conclusions:

• The proposed biomaterial-based memristor effectively resolves non-linearity issues in artificial synapses.
• The device shows significant potential for high-performance computational and wearable/implantable electronic systems.
• This work validates the use of biopolymers in advanced neuromorphic computing applications.