Ionotronic Halide Perovskite Drift-Diffusive Synapses for Low-Power Neuromorphic Computation.

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Developing efficient learning circuitry requires emulating brain-like signal processing.
• Few devices offer the tunable conductance range needed for mimicking biological synaptic spatiotemporal plasticity.
• Ionic semiconductors coupling electronic transitions with ionic kinetics can enable energy-efficient analog switching of conductance states.

Purpose of the Study:

• To utilize ionic-electronic coupling in halide perovskites for creating memristive synapses.
• To achieve dynamic, continuous transitions in conductance states for synaptic plasticity.
• To demonstrate the potential of these memristors for energy-efficient neuromorphic computation.

Main Methods:

• Ionic-electronic coupling in halide perovskite semiconductors was employed.
• Memristive synapses were fabricated with dynamic continuous conductance states.
• Optimized pulsing schemes were used to facilitate paired-pulse facilitation and spike-time-dependent plasticity.

Main Results:

• Halide perovskite synapses exhibited tunable conductance states and dynamic plasticity.
• Larger organic cations (ammonium, formamidinium) showed more notable plasticity than cesium.
• The memristive synapses demonstrated reconfigurability, learning, forgetting, and fault tolerance.
• Network simulations validated their utility for unsupervised learning of handwritten digits.

Conclusions:

• Halide perovskite memristors offer a pathway to energy-efficient neuromorphic computing.
• These devices enable emulation of synaptic plasticity crucial for learning and computation.
• The study paves the way for novel ionotronic neuromorphic architectures using halide perovskites.