Anisotropic Optoelectronic Synapses in 2D NbGeTe for Direction-Programmable Neuromorphic Perception and Decision-Making.

Neuromorphic computing presents a promising solution for the von Neumann bottleneck, enabling energy-efficient and intelligent sensing platforms. Although 2D materials are ideal for bioinspired neuromorphic devices, achieving multifunctional synaptic operations with simple configurations and linear weight updates remains challenging. Inspired by biological axons, the in-plane anisotropy of 2D NbGeTe is exploited to develop dual electronic-optical synaptic devices. The device exhibits anisotropic hole mobilities (137.97 cm Vs along the a-axis and 78.29 cmVs along the b-axis) and a wavelength-dependent photoresponse. This enables directional synaptic plasticity under electrical-optical co-stimulation, achieving 98.3% accuracy along the a-axis and 88.3% along the b-axis in adaptive image processing. A machine vision system with 89.6% object recognition accuracy and an intelligent vehicle navigation platform with 90.2% decision-making accuracy is also demonstrated. The integration of anisotropic transport and spectrally tunable responses in a single material paves the way for compact neuromorphic hardware with multimodal sensing and parallel processing capabilities. This study advances 2D material-based neuroelectronics for edge computing, autonomous robotics, and adaptive artificial intelligence systems.