Integración monolítica de detección, computación y almacenamiento en un dispositivo neuromórfico basado en el efecto...

Conventional neuromorphic computing is hindered by the physical separation of the sensing, memory, and processing units. This work addresses the critical challenge of intrinsically integrating these functions within a monolithic device. We present a novel Ag/NiO/NiCo₂O₄ heterostructure that integrates magnetic sensing, memristor-based storage, and analog computation together. This device exploits the anomalous Hall effect in the NiCo₂O₄ layer, which exhibits perpendicular magnetic anisotropy, to concurrently enable magnetic field transduction (sensing) and intrinsic sigmoid activation function (computation). Concurrently, a nickel oxide layer serves as a nonvolatile memristor, achieving reconfigurable storage by further modulating the square anomalous Hall loop of the NiCo₂O₄ layer (storage). Crucially, the memory state of the nickel oxide layer electrically reconfigures the anomalous Hall effect response of the NiCo₂O₄ layer, which is achieved through interfacial modulation via oxygen vacancy migration. Experimentally, programming the nickel oxide memristor with low-voltage excitation (±2 V) controllably and nonvolatily modulates the sigmoidal activation threshold. Upon excitation, the device can be programmatically switched between high-sensitivity (3 V/(A·T)) and wide-dynamic-range (±200 mT) modes while maintaining native compatibility with analog computing architectures. This monolithic integration establishes a viable foundation for energy-efficient neuromorphic sensors and multifunctional autonomous systems.