Atomistic Simulations of the Crystallization Kinetics of Ge₂Sb₂Te₅ and GeTe in Confined Geometries

Phase change alloys are among the most promising materials for the realization of artificial neurons
and synapses for neuromorphic computing. In these applications, one exploits the different resistive
levels that can be realized by full or partial crystallization of the amorphous phase upon application
of current pulses. In a recent work [1], it was proposed that a superlattice (SL) geometry made of
alternating layers of the phase change material Sb₂Te₃ and more thermally stable confining layers of
TiTe₂ woud exhibit superior properties for neuromorphic computing. However, Sb₂Te₃ suffers from
insufficient data retention due to its low crystallization temperature T_x . Substituting Sb₂Te₃ with a
phase change compound with a higher T_x, such as GeTe or Ge₂Sb₂Te₅ (GST), seems an interesting
option in this respect. Nanoconfinement might, however, alters the crystallization kinetics with
respect to the bulk. In this contribution, we will discuss the results of molecular dynamics
simulations of the crystallization process of Ge₂Sb₂Te₅ and GeTe [2] nanoconfined in geometries
mimicking GST/TiTe₂ or GeTe/TiTe₂ superlattices. To this aim, we performed large scale
simulations with machine learning interatomic potentials [3,4]. The simulations reveal that
nanoconfinement induces a mild reduction in the crystal growth velocities which would not hinder
the application of GST/TiTe₂ or GeTe/TiTe₂ heterostructures in neuromorphic devices with superior
data retention.
[1] K. Ding et al, Science 366, 210 (2019)
[2] D. Acharya, O. Abou El Kheir, D. Campi, and M. Bernasconi , Sci. Rep. 14, 3224 (2024)
[3] O. Abou El Kheir, L. Bonati, M. Parrinello, and M. Bernasconi, npj Comp. Mater. 10, 33 (2024)
[4] S. Gabardi et al, J. Phys. Chem. C 121, 23827 (2017)