Intrinsic Self-Rectification and Analogue Resistive Switching in Single-Component Polymeric Graphitic Carbon Nitride for Neuromorphic Applications

Abstract

The increasing computational demands of artificial intelligence and data-centric technologies expose fundamental limitations of conventional von Neumann architectures, particularly due to latency and high-power consumption caused by excessive data movement. Memristive devices that integrate memory and computation within a single structure offer a promising route toward energy-efficient in-memory and neuromorphic systems. A simple, solution-processed, single-component Ag/g-C₃N₄/ITO resistive switching (RS) device based on graphitic carbon nitride (g-C₃N₄) nanosheets is demonstrated. The device exhibits stable analogue resistive switching with intrinsic self-rectifying behavior. Unlike conventional binary devices, it shows gradual and controllable conductance modulation under consecutive voltage sweeps and pulse stimulation, enabling reliable synaptic weight updates. A rectification ratio of ∼5 provides inherent nonlinearity, supporting sneak-path suppression without additional selector elements. The device maintains stable endurance over 100 cycles and retention up to 10⁴ s, and demonstrates clear potentiation-depression characteristics under pulsed operation. Conduction analysis indicates a transition from quasi-Ohmic behavior to Schottky emission and trap-controlled space-charge-limited conduction (SCLC), suggesting that switching is governed by interface barrier modulation and trap-assisted transport. The structural simplicity, analogue switching capability, intrinsic rectification, and low-power operation highlight the potential of g-C₃N₄-based devices for artificial synapses and neuromorphic computing applications.