Experimental Investigation and Performance Analysis of V2O5-Based Memristive Devices for Brain-Inspired Computing

Neuromorphic computing, drawing its inspiration from the human brain, empowers the creation of neural activity that excels in both energy efficiency and fast processing. Replicating the human brain involves addressing essential challenges, such as emulating synapse potentiation/depression and achieving robust connectivity in biological neurons. Memristive device-based artificial neurons have garnered significant attention for their simple structure, high data density, and remarkable scalability, rendering them effective in mimicking biological neurons for computing applications. However, the unreliability concerns associated with memristors have posed a primary hindrance to the advancement of memristor-based artificial neurons and neuromorphic computing. In our study, we have fabricated
$4\times 4$
cross-cell memristive devices capable of low-power neural activity and synaptic functionality, essential for brain-inspired computing. Our fabricated device exhibits analog resistive switching (ARS) and bipolar resistive switching (BRS) behaviors, suitable for neuromorphic computing and memory applications. To optimize its electrical performance, we used a Keithley-4200A semiconductor parameter analyzer with a triangular dc sweep voltage (−0.8 V/+0.8 V) at different temperatures. Finally, we optimized the memristor’s performance by assessing its excitatory postsynaptic current, data storage capabilities, excellent linearity for energy-efficient edge computing devices, and synaptic responses at different read voltages (RVs) (0.1–0.5 V) and pulsewidths (PWs) (10, 20, 30, 40, and
$50~\mu $
s).