Robust and Large‐Area Memristor Arrays Based on Encapsulated 2D van der Waals Heterostructures

ABSTRACT

The integration of memristors with 2D materials promises a comprehensive hardware revolution for computational power and energy efficiency in artificial intelligence through in‐memory and neuromorphic computing. Here, we report robust, large‐area HfO ₂ /WS _2‐x O _x /graphene memristor arrays fabricated via a scalable electrochemical deposition (ECD) and atomic layer deposition (ALD) process. The devices exhibit an ultralow areal energy density of 44 pJ/µm ² , switching speed below 300 ns, endurance exceeding 10 ⁹ cycles, and reliable operation from −100°C to 400°C. Systematic characterizations confirm the high crystallinity and uniform encapsulation of the 2D vertical heterostructures. Furthermore, the memristors emulate key synaptic functions (LTP, LTD, and STDP) with near‐linear conductance modulation, enabling hardware‐level convolutional neural networks that achieve 84% accuracy on CIFAR‐10 and effective sparse coding for feature extraction. These results establish encapsulated 2D van der Waals heterostructures (2D vdWHs) memristor arrays as a promising platform for high‐density, energy‐efficient, and harsh‐environment neuromorphic computing systems.