A machine-learning interatomic potential for the Ge2Sb2Te5 phase change compound

In the last fifteen years atomistic simulations based on density functional theory (DFT) have provided useful insights on the structural and functional properties of phase change materials [1]. However, several key issues such as the effect of confinement and nanostructuring on the crystallization kinetics, just to name a few, are presently beyond the reach of DFT simulations. A route to overcome the limitations in system size and time scale and enlarge the scope of DFT methods is the exploitation of machine learning techniques trained on a DFT database to generate interatomic potentials for large scale molecular dynamics simulations. The first example of the application of such an approach to the study of phase change compounds dates back to 2012 when an interatomic potential for GeTe [2] was devised within the neural network (NN) framework proposed by Behler and Parrinello [3]. The same scheme was also applied to elemental Sb [4]. The NN potentials were then used to address several issues such as the crystallization in ultrathin films [4] and nanowires, and the thermal conductivity and aging of the amorphous phase [5]. More recently, a different machine learning technique within the Gaussian approximation potential (GAP) framework was exploited to generate an interatomic potential for Ge2Sb2Te5 [6]. In this talk, we report on the generation of an interatomic potential for the Ge2Sb2Te5 compound within the neural network framework implemented in the DeePMD-kit package [7]. The interatomic potential allows simulating several tens of thousands of atoms for tens of ns at a modest computational cost. The validation of the potential and its application to the study of the crystallization kinetics in the bulk phase will be discussed [8]. REFERENCES [1] W. Zhang, R. Mazzarello, M. Wuttig, and E. Ma, Nat. Rev. Mater., 4, 150 (2019). [2] G. C. Sosso, G. Miceli, S. Caravati, J. Behler, and M. Bernasconi, Phys. Rev. B 85, 174103 (2012). [3] J. Behler and M. Parrinello, Phys. Rev. Lett. 98, 146401 (2007). [4] D. Dragoni, J. Behler, and M. Bernasconi, Nanoscale 13, 16146 (2021). [5] G. C. Sosso and M. Bernasconi, MRS Bulletin 44, 705 (2019). [6] F. C Mocanu, K. Konstantinou, T. H. Lee, N. Bernstein, V. L. Deringer, G. Csányi, and S. R. Elliott, J. Phys, Chem B 122, 8998 (2018). [7] H. Wang, L. Zhang, J. Han, and W. E, Comp. Phys. Commun. 228, 178 (2018); L. Zhang, J. Han, H. Wang, R. Car, W. E, Phys. Rev. Lett. Phys. Rev. Lett. 120, 143001 (2018). [8] O. Abou El Kheir, L. Bonati, M. Parrinello, and M. Bernasconi, arXiv:2304.03109 (2023).