神经形态工程学
电阻随机存取存储器
重置(财务)
材料科学
偏压
线性
多物理
光电子学
模拟电子学
电压
电子线路
电子工程
电气工程
计算机科学
物理
工程类
有限元法
人工智能
经济
金融经济学
热力学
人工神经网络
作者
Matthew P. West,Fabia F. Athena,Samuel Graham,Eric M. Vogel
摘要
Resistive random-access memory (RRAM) devices have been widely studied for neuromorphic, in-memory computing. One of the most studied RRAM structures consists of a titanium capping layer and a HfOx adaptive oxide. Although these devices show promise in improving neuromorphic circuits, high variability, non-linearity, and asymmetric resistance changes limit their usefulness. Many studies have improved linearity by changing materials in or around the device, the circuitry, or the analog bias conditions. However, the impact of prior biasing conditions on the observed analog resistance change is not well understood. Experimental results in this study demonstrate that prior higher reset voltages used after forming cause a greater resistance change during subsequent identical analog pulsing. A multiphysics finite element model suggests that this greater analog resistance change is due to a higher concentration of oxygen ions stored in the titanium capping layer with increasing magnitude of the reset voltage. This work suggests that local ion concentration variations in the titanium capping layer of just tens of atoms cause significant resistance variation during analog operation.
科研通智能强力驱动
Strongly Powered by AbleSci AI