Interface-driven performance boost in NbOx/V2O5 bilayer memristors for next-generation neuromorphic systems

Abstract

The advancement of bilayer memristor devices is crucial for enhancing performance and enabling multifunctionality in next-generation memory technologies for neuromorphic applications. Here, we present an optimized fabrication and characterization of Resistive Random-Access Memory (RRAM) devices composed of a NbOx/V2O5 bilayer structure. By systematically varying the thickness of the V2O5 layer while maintaining a constant NbOx thickness of less than 14 nm, we achieved a significant enhancement in device performance. Our optimized bilayer device with a 60 nm thick V2O5 layer exhibited an on/off current ratio of 106, achieving a four-order-of-magnitude improvement over single-layer NbOx devices while maintaining the same set voltage. Additionally, this bilayer structure demonstrated endurance over 103 DC cycles and retention exceeding 104 s. Comprehensive material analysis using HRTEM, EDX, and XPS with depth profile confirmed the precise composition and structural integrity of the devices. Furthermore, neuromorphic pulse measurements revealed synaptic-like behavior, underscoring the potential of the NbOx/V2O5 bilayer structure for integration into neuromorphic computing systems. This behavior was further applied for MNIST dataset based digit recognition to obtain an average recognition accuracy of 97.69 %. These findings not only advance the state-of-the-art in RRAM technology but also open new avenues for its application in next-generation computing architectures.