Self‐Powered Flexible Amorphous‐Silicon Radial p‐i‐n Nanowire Device for High‐Speed Physical Reservoir Computing

ABSTRACT Self‐powered optoelectronic devices with fast transient dynamics are attractive hardware platforms for physical reservoir computing (RC), while realizing high‐speed temporal optical signal processing in such systems remains challenging. Here, we demonstrate a flexible amorphous‐silicon (a‐Si) p‐i‐n radial‐junction silicon nanowires (RJ SiNWs) device that operates under zero bias and exhibits nonlinear, short‐lived transient photovoltage dynamics with sub‐millisecond relaxation, providing the fading‐memory behavior required for physical RC. The radial p‐i‐n architecture maintains efficient photocarrier separation via the built‐in electric field while geometrically decoupling optical absorption from carrier collection distance. Benefiting from the rapid temporal response, the RJ SiNWs node reliably encodes a 4‐bit optical pulse stream within only 200 µs. Combined with a nonlinear MLP readout, the resulting RC framework achieves a test accuracy of 94.1% in handwritten‐digit recognition. Furthermore, the device exhibits excellent mechanical durability enabled by the vertically anchored SiNWs architecture, retaining ∼90% of the saturated photovoltage after 3000 bending cycles at a radius of 10 mm, highlighting its suitability for wearable and conformal neuromorphic vision systems. This work establishes flexible a‐Si RJ SiNWs as a promising class of fast, self‐powered optoelectronic nodes for low‐power, in‐sensor edge neuromorphic systems.