Atomic-layer-deposited TiN interlayer suppressing oxygen migration in HfO2 RRAM for neuromorphic computing

Abstract

With the rapid advancement of in-memory and neuromorphic computing, resistive random-access memory (RRAM) has emerged as a key candidate owing to its high scalability, analog tunability, and low-power operation. However, achieving stable and uniform resistive switching remains a major challenge, particularly in hafnium oxide (HfO<inf>2</inf>)-based devices, where oxygen scavenging by Ti bottom electrodes often leads to performance degradation. In this study, we propose a TiN/Hf/HfO<inf>2</inf>/TiN/Ti RRAM device incorporating a 10 nm atomic-layer-deposited (ALD) TiN anti-scavenging layer (ASL) to suppress oxygen migration at the interface. The ALD-grown TiN ASL effectively enhances interfacial stability, confines conductive filament formation, and improves cell-to-cell switching uniformity. The device exhibits reliable bipolar switching with a SET voltage of + 5 V and a RESET voltage of −0.5 V, maintaining a clear conductance window and < 10 % variation between resistance states. Volatile retention and pulse-based measurements further confirm short-term memory (STM) characteristics and dynamic synaptic modulation. Moreover, the device demonstrates biologically inspired plasticity behaviors, including spike-amplitude-, spike-rate-, spike-width-, and spike-number-dependent plasticity (SADP, SRDP, SWDP, and SNDP). These results highlight the critical role of ALD-engineered TiN interlayers in stabilizing resistive switching and enabling reliable, real-time neuromorphic computing applications. © 2025 Elsevier B.V.