Microscale optoelectronic synapses with switchable photocurrent from halide perovskite

TL;DR

This paper presents microscale halide perovskite artificial synapses capable of volatile, switchable photocurrent modulation via ion migration, enabling neuromorphic visual data processing with CMOS compatibility.

Contribution

The work introduces a novel perovskite-based optoelectronic synapse with switchable polarity and STDP learning, advancing neuromorphic device integration.

Findings

Photocurrent is modulated by bias voltage and illumination.

Device supports polarity switching between inhibitory and exhibitory modes.

Photocurrent changes are volatile, decaying over seconds.

Abstract

Efficient visual data processing by neuromorphic networks requires volatile artificial synapses that detect and process light inputs, ideally in the same device. Here, we demonstrate microscale back-contacted optoelectronic halide perovskite artificial synapses that leverage ion migration induced by a bias voltage to modulate their photocurrent. The photocurrent changes are due to the accumulation of mobile ions, which induces a transient electric field in the perovskite. The photocurrent changes are volatile, decaying on the order of seconds. The photocurrent changes can be controlled by both the applied voltage and illumination. The symmetric device supports changing of the photocurrent polarity, switching between inhibitory and exhibitory functioning. The photocurrent can be updated by spike-timing-dependent plasticity (STDP)-learning rules inspired by biology. We show with simulations how this could be exploited as an attention mechanism in a neuromorphic detector. Our fabrication procedure is compatible with high-density integration with CMOS and memristive neuromorphic networks for energy-efficient visual data processing inspired by the brain.