CMOS back-end compatible memristors for in situ digital and neuromorphic computing applications

Area of Science:

• Materials Science
• Computer Engineering
• Neuroscience

Background:

• Traditional von Neumann architecture faces limitations in data processing efficiency due to physically separated processing and storage units, leading to slow speeds and high power consumption.
• Digital and analog calculations in conventional systems have significant conversion limitations.
• Neuromorphic computing and in-memory logic calculations offer potential solutions to these limitations.

Purpose of the Study:

• To develop a flexible two-terminal memristor for next-generation computing architectures.
• To enable simultaneous digital memory and neuromorphic calculations.
• To simulate biological synapse and neuron dynamics with high fidelity.

Main Methods:

• Fabrication of a flexible memristor using a porous low-k SiCO:H material compatible with back-end CMOS processes.
• Verification of IMP-based and MAGIC-based digital logic calculations.
• Introduction of an Ag ion-based conductive filament for neuromorphic computing simulations.
• Modulation of conductance evolution and Ag diffusion dynamics using external electric fields.

Main Results:

• The SiCO:H memristor operates at a low voltage (200 mV) with a fast response speed (100 ns).
• It successfully performed both digital logic calculations and neuromorphic computing tasks.
• The device demonstrated a transition from short-term to long-term plasticity during enhancement and inhibition, mimicking biological synaptic behavior.
• Faithful simulation of biological synapses and neuron dynamics was achieved.

Conclusions:

• The developed SiCO:H memristor offers a novel solution for next-generation computer architectures by integrating digital and neuromorphic computing.
• Its high biological reality, compatibility, and scalability make it a promising candidate for advanced computing.
• The device overcomes the inherent limitations of the von Neumann architecture in terms of speed and power consumption.