29 August 2011 in Applied Physics Letters
Resistive switching in transition metal oxides could form the basis for next-generation non-volatile memory (NVM). It has been reported that the current in the high-conductivity state of several technologically relevant oxide materials flows through localized filaments, but these filaments have been characterized only individually, limiting our understanding of the possibility of multiple conductive filaments nucleation and rupture and the correlation kinetics of their evolution. In this study, direct visualization of uncorrelated multiple conductive filaments in ultra-thin HfO2-based high-κ dielectric resistive random access memory (RRAM) device has been achieved by high-resolution transmission electron microscopy (HRTEM), along with electron energy loss spectroscopy (EELS), for nanoscale chemical analysis. The locations of these multiple filaments are found to be spatially uncorrelated. The evolution of these microstructural changes and chemical properties of these filaments will provide a fundamental understanding of the switching mechanism for RRAM in thin oxide films and pave way for the investigation into improving the stability and scalability of switching memory devices.
20 December 2019 in IEEE Access
Neuromorphic computing is a promising candidate for breaking the von Neumann bottleneck and developing high-efficient computing systems. Here we present a W/TaOx/Pt high-precision electronic synapse with excellent analog properties for neuromorphic computing. The device exhibits the potential of 10-bit weight precision, which is state of the art in conductance levels. Furthermore, the device shows linear weight update behavior in a specific conductance range, linear I-V curves in low voltage regime, long time retention, and precise modulation of weight. These characteristics are very helpful for improving the accuracy of neuromorphic networks. Finally, a 400×60×10 three-layer perceptron was constructed with W/TaOx/Pt synapses for MNIST classification and ∼92% accuracy was achieved.
1 August 2020 in AIP Advances
The activation function is a crucial part for memristive neural networks. For the first time, we propose a memristor-based activation function by using the natural non-linear characteristics of the memristor itself. Compared to the virtual ground circuit in traditional memristive neural networks, the feedback resistance was replaced by the W/TaOx/Ru memristor with no additional expense. Simulation results demonstrate that the proposed memristor-based activation function can achieve a performance similar to that of traditional activation functions on the Mixed National Institute of Standards and Technology database. In addition, an improvement in the recognition rate of up to 2% can be obtained in different neuromorphic networks by modulating the non-linearity of the memristor. Furthermore, the memristor-based activation function can also receive a 94% recognition rate even considering the non-ideal factors of the device.