Hybrid Perovskite‐Based Flexible and Stable Memristor by Complete Solution Process for Neuromorphic Computing

Abstract

Abstract The limits of transistor scaling and digital architectures are encouraging research into new electronic materials, devices, and systems to meet growing computing demands. In the realm of artificial intelligence, mimicking brain activity for neuromorphic computing is a promising approach. Herein, Ruddlesden–Popper (RP) perovskite‐based flexible and environmentally stable memristors are presented that achieve on‐demand resistive switching between several nonvolatile states by controlling the number of layers and compliance current (CC). The optimal flexible perovskite device based on n = 5 composition, fabricated by complete solution process and measured under ambient conditions without any encapsulation, shows excellent ON/OFF ratio ≈7 × 103, endurance performance (2500 cycles), and robustness to mechanical flexure up to 5 mm bending radii. The role of the physical/chemical reaction at the perovskite–electrode interface is investigated to reveal the origin of the resistive switching in these devices. The primary probing on synaptic characteristics shows stable learning (potentiation and depression) behavior measured up to 19 000 pulses. The invariant paired pulse facilitation index on flat and 5 mm bending radii demonstrates their feasibility for neuromorphic computing applications. The in‐depth analysis also validates the potential of RP‐based memristor devices for applications that require real‐time synaptic processing under extreme mechanical states such as electronic skins.

Keywords

• artificial synapses
• flexible memristors
• neuromorphic computation
• resistive switching
• Ruddlesden–Popper perovskites