Energy-efficient, stable, and temperature-tolerant neuromorphic device based on single crystals of halide perovskites

As artificial intelligence emerges for substituting human labor, novel forms of computing are currently in demand. Synaptic memristors functioning as the basic element of the brain serve as a promising strategy to fundamentally approach brain-inspired computing. However, several great challenges must be addressed before practical memristor-based neuromorphic computing can be realized, including substandard power consumption, operational stability, and temperature tolerance. Here, we report on the capacity to confront all these challenges by exploiting the ability of single-crystalline halide perovskites to engage in artificial synapses with excellent synaptic weight modulation. Temperature-dependent electrical-cycling test and modeling unveil a unique synaptic weight updating mechanism where an ionic inductive effect drives resistive switching from a high-resistance state to a low-resistance state by symmetrically lowered Schottky barrier heights. A proof-of-concept demonstration for neuromorphic computing is performed by using the robust device comprising spiking neural networks, which are proven effective for recognizing handwritten digits, with an accuracy of over 90% in a wide temperature range.