Simulation of the crystallization kinetics of Ge${}_2$Sb${}_2$Te${}_5$ nanoconfined in superlattice geometries for phase change memories

Abstract

Phase change materials are the most promising candidates for the realization of artificial synapsis for neuromorphic computing. Different resistance levels corresponding to analogic values of the synapsis conductance can be achieved by modulating the size of an amorphous region embedded in its crystalline matrix. Recently, it has been proposed that a superlattice made of alternating layers of the phase change compound Sb${}_2$Te${}_3$ and of the TiTe${}_2$ confining material allows for a better control of multiple intermediate resistance states and for a lower drift with time of the electrical resistance of the amorphous phase. In this work, we consider to substitute Sb${}_2$Te${}_3$ with the Ge${}_2$Sb${}_2$Te${}_5$ prototypical phase change compound that should feature better data retention. By exploiting molecular dynamics simulations with a machine learning interatomic potential, we have investigated the crystallization kinetics of Ge${}_2$Sb${}_2$Te${}_5$ nanoconfined in geometries mimicking Ge${}_2$Sb${}_2$Te${}_5$/TiTe${}_2$ superlattices. It turns out that nanoconfinement induces a slight reduction in the crystal growth velocities with respect to the bulk, but also an enhancement of the nucleation rate due to heterogeneous nucleation. The results support the idea of investigating Ge${}_2$Sb${}_2$Te${}_5$/TiTe${}_2$ superlattices for applications in neuromorphic devices with improved data retention. The effect on the crystallization kinetics of the addition of van der Waals interaction to the interatomic potential is also discussed.