当前位置: X-MOL 学术Front Phys. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Self-aligned TiOx-based 3D vertical memristor for a high-density synaptic array
Frontiers of Physics ( IF 6.5 ) Pub Date : 2024-07-03 , DOI: 10.1007/s11467-024-1419-2
Subaek Lee , Juri Kim , Sungjun Kim

The emerging nonvolatile memory, three-dimensional vertical resistive random-access memory (VRRAM), inspired by the vertical NAND structure, has been proposed to replace NAND flash memory which has reached its integration limit. To improve the vertical ionic diffusion occurring in the conventional VRRAM structure, we propose a Pt/HfO2/TiO2/Ti self-aligned VRRAM with physically confined switching cells through sidewall thermal oxidation. We achieved stable bipolar switching, endurance (>104 cycles), and retention (>104 s) responses, and improved the interlayer leakage current issue through a distinctive self-aligned structure. Additionally, we elucidated the switching mechanism by analyzing current levels concerning ambient temperature. To utilize VRRAM for neuromorphic computing, the biological synaptic functions are emulated by applying pulse stimulation to the synaptic cell. The weight modulation of biological synapses is demonstrated based on potentiation, depression, spike-rate-dependent plasticity, and spike-timing-dependent plasticity. Additionally, we improve the pattern recognition rate by creating a linear conductance modulation with an incremental pulse train in pattern recognition simulations. The stable electrical characteristics and implementation of various synaptic functions demonstrate that self-aligned VRRAM is suitable for neuromorphic systems as a high-density synaptic device.



中文翻译:


用于高密度突触阵列的自对准 TiOx 基 3D 垂直忆阻器



受垂直NAND结构启发,新兴的非易失性存储器——三维垂直电阻式随机存取存储器(VRRAM)被提出来取代已达到集成度极限的NAND闪存。为了改善传统 VRRAM 结构中发生的垂直离子扩散,我们提出了一种通过侧壁热氧化具有物理限制开关单元的 Pt/HfO 2 /TiO 2 /Ti 自对准 VRRAM 。我们实现了稳定的双极开关、耐久性(>10 4 周期)和保持(>10 4 s)响应,并通过独特的自对准结构改善了层间漏电流问题。此外,我们通过分析与环境温度相关的电流水平来阐明切换机制。为了利用 VRRAM 进行神经形态计算,通过对突触细胞施加脉冲刺激来模拟生物突触功能。生物突触的权重调节是基于增强、抑制、尖峰速率依赖性可塑性和尖峰时间依赖性可塑性来证明的。此外,我们通过在模式识别模拟中使用增量脉冲序列创建线性电导调制来提高模式识别率。稳定的电气特性和各种突触功能的实现表明,自对准 VRRAM 适合作为高密度突触设备用于神经形态系统。

更新日期:2024-07-03
down
wechat
bug