Flexible 2D Perovskite–hBN Optoelectronic Dendritic Neurons for Energy-Efficient Information Processing

Artificial dendrites, capable of nonlinear signal integration, are key building blocks for next-generation neuromorphic hardware, yet their realization in soft, solution-processed materials remains hindered by instability and poorly controlled ion dynamics. Here, we report an artificial dendrite element based on a vertically orchestrated ITO/Ag nanoparticle (NP)/hBN/(PEA)2MA4Pb5I16/Pt NPS/ITO heterostructure, where a monolayer hBN barrier governs metal-ion injection with atomic precision. The hBN-mediated field redistribution enables graded and time-dependent filament initiation within the oriented two-dimensional (2D) Ruddlesden-Popper perovskite, producing highly nonlinear current integration analogous to excitatory postsynaptic potentials. Under consecutively correlated stimuli, the device exhibits hallmark dendritic behaviors-temporal summation, threshold-dependent firing, refractory period, and wavelength-sensitive integration-arising from controlled Ag+ accumulation and reversible filament branching. The devices operate at ∼1 V, demonstrate stable and reproducible transient responses over >1011 cycles, and retain robust performance under mechanical deformation. Complementary physics-based numerical modeling and Density Functional Theory (DFT) calculations elucidate the influence of hBN on Ag+ diffusivity, as well as on filament formation. The dendritic neurons can efficiently encode velocity and direction, enabling compact feature maps for accurate motion classification and trajectory tracking. Our results introduce a promising material platform for high-density, energy-efficient neuromorphic processors capable of bio-inspired spatiotemporal learning.