Artificial Neural Network Classification Using Al-Doped HfO<sub><i>x</i></sub>-Based Ferroelectric Tunneling Junction with Self-Rectifying Behaviors
Eunjin Lim, Dongyeol Ju, Jung Woo Lee, Yongjin Park, Min-Hwi Kim, Sungjun Kim
Abstract
In this study, we meticulously engineered an Al-doped hafnia-based ferroelectric tunneling junction (FTJ) with a metal–ferroelectric–silicon (MFS) structure. We conducted a thorough analysis of its memory characteristics, revealing a substantial remnant polarization of 24.17 μC/cm 2, a noteworthy tunneling electroresistance value of 265, exceptional endurance with 10 6 operational cycles, and robust retention (>10 4 s), thereby demonstrating the viability of the FTJ as a nonvolatile memory device. Additionally, through rectification of this MFS FTJ, an effective array scale of approximately 1349 with a modified read scheme was ensured. Expanding our study of neuromorphic applications, we explored phenomena such as potentiation/depression, paired-pulse facilitation (PPF), excitatory postsynaptic currents (EPSC), and spike-rate-dependent plasticity (SRDP). Notably, this memristor has outstanding potential for visual memory processing. In conclusion, our findings unequivocally underscore the immense potential of the hafnia-based FTJ for applications in neural networks, emphasizing its significance in advancing neuromorphic computing.