SnSe/FeSe Memristors for Programmable Multi-Regime Switching and Neuromorphic Computing

Neuromorphic devices capable of emulating biological synapses have emerged as a promising alternative to conventional computing architectures. Here, solution-processed 2D SnSe/FeSe nanosheet heterostructures are demonstrated as versatile memristors exhibiting programmable multi-regime resistive switching. Percolating nanosheet networks with intrinsic defect dynamics enable the coexistence of digital and analog switching. Ag-electrode devices exhibit digital switching via filamentary electrochemical metallization (ECM), while Ag-free measurements confirm analog switching driven by vacancy motion and Schottky-barrier modulation. The ECM-based devices demonstrate robust endurance (~10⁴ cycles), lower switching voltage (< 0.5 V), and high retention as non-volatile memory elements, while the Ag-free devices display gradual potentiation and depression, short-term plasticity and paired-pulse facilitation, closely mimicking neuronal functions. Moreover, a memristive neural network is simulated using the hardware-software co-design and differential pair mapping mechanism, showing performance close to that of a pure software-based neural network. These findings highlight 2D nanosheet heterostructures as low-cost, scalable platforms for adaptive, brain-inspired computing, bridging the gap between memory storage and neuromorphic processing.