Multi-ReRAM synapses for artificial neural network training
Irem Boybat, Cecilia Giovinazzo, et al.
ISCAS 2019
Phase-change memory devices are expected to play a key role in future computing systems as both memory and computing elements. A key challenge in this respect is the temporal evolution of the resistance levels commonly referred to as “resistance drift.” In this paper, a comprehensive description of resistance drift as a result of spontaneous structural relaxation of the amorphous phase-change material toward an energetically more favorable ideal glass state is presented. Molecular dynamics simulations provide insights into the microscopic origin of the structural relaxation. Based on those insights, a collective relaxation model is proposed to capture the kinetics of structural relaxation. By linking the physical material parameters governing electrical transport to such a description of structural relaxation, an integrated drift model that is able to predict the current–voltage characteristics at any instance in time even during nontrivial temperature treatments is obtained. Accurate quantitative matching with experimental drift measurements over a wide range of time (10 decades) and temperature (160–420 K) is demonstrated.
Irem Boybat, Cecilia Giovinazzo, et al.
ISCAS 2019
Christophe Piveteau, Manuel Le Gallo, et al.
IMW 2019
Abu Sebastian, Tomas Tuma, et al.
Nature Communications
Malte J. Rasch, Diego Moreda, et al.
AICAS 2021