Nonlinear Ion Dynamics Enable Spike Timing Dependent Plasticity of Electrochemical Ionic Synapses

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Spiking neural networks (SNNs) require programmable synaptic devices for energy-efficient operation.
• Electrochemical ionic synapses (EIS) offer low-energy, low-variability weight updates.
• Nonlinear kinetics in EIS are crucial for implementing complex synaptic plasticity.

Purpose of the Study:

• To leverage the nonlinear kinetics of EIS for implementing spike-timing-dependent plasticity (STDP).
• To demonstrate deterministic emulation of various STDP forms using proton-based EIS.
• To explore heterogeneous STDP and controllable timescales within EIS arrays.

Main Methods:

• Utilized proton-based electrochemical ionic synapses (EIS).
• Emulated STDP functions via linear superposition of pre- and post-synaptic signals.
• Investigated STDP behavior across a range of timescales (nanoseconds to milliseconds).

Main Results:

• Successfully implemented diverse STDP forms with deterministic prediction in EIS.
• Demonstrated heterogeneous STDP within an array for varied learning rules.
• Achieved lower variability in hardware STDP compared to other implementations.
• Controlled STDP timescales from nanoseconds to milliseconds.

Conclusions:

• EIS ion and charge transfer dynamics enable bio-plausible synapses for SNN hardware.
• EIS technology offers high energy efficiency, reliability, and throughput for neuromorphic computing.
• The deterministic nature of EIS leads to more uniform and reliable synaptic weight updates.