Séminaire ICube MSII

Nous avons le plaisir d'accueillir Marco Bernasconi, de l'University of Milano-Bicocca, Milan, Italy pour un séminaire ICube MSII.

Date et lieu : mardi 5 mai à 10h, en salle A302 au pôle API de Telecom Physique STrasbourg

La présentation aura lieu en anglais.

Titre : Atomic simulations of phase change materials for electronic memories with machine learning interatomic potentials

Abstract:

Phase change materials (PCM) such as the flagship Ge₂Sb₂Te₅ (GST225) compound are exploited in key enabling technologies such as non-volatile electronic memories, in-memory and neuromorphic computing, and embedded memories in microcontrollers [1]. The two digital states of the memories are encoded in the amorphous and crystalline phases of PCMs that feature a large difference in the electrical resistivity. Readout of the memory consists of the measurement of the resistance at low bias while the set/reset processes consist of a fast and reversible transformation between the two phases induced by heating via current pulses. Partial crystallization during set or modulation of the amorphous region during reset allow encoding different analogic resistance levels for neuromorphic and in-memory computing [1].

Atomistic simulations based on density functional theory (DFT) have provided useful insights on the structural and functional properties of PCMs over the years [1]. However, several key issues for the operation of the devices 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 generation of machine learning interatomic potential (MLIP) trained on a DFT database for large scale molecular dynamics (MD) simulations. The first application of this approach to PCMs dates to 2012, when we devised a MLIP for GeTe [2] to study the crystallization process and the thermal conductivity and aging in bulk and nanowires [3]. 

In this talk, I will report on the generation of a MLIP for the flagship GST225 ternary alloy [4], within the neural network framework implemented in the DeePMD-kit package. The MLIP allowed us to perform MD simulations of million-atom models at the length and time scales of the real device to uncover the kinetics of the set process (crystallization of the amorphous phase) at the operation conditions of the memory [5]. Extension to non-stoichiometric Ge-rich Ge_xTe [6] and GeSbTe alloys that feature a higher crystallization temperature, as required by memories embedded in microcontrollers for the automotive sector, will also be presented. Insights on the relaxation times and on heterogeneous dynamics in the supercooled liquid of relevance for the crystallization kinetics will be discussed as well [7].

• [1] W. Zhang, R. Mazzarello, M. Wuttig, E. Ma, Nat. Rev. Mater. 4, 150 (2019)
• [2] G. C. Sosso, G. Miceli, S. Caravati, J. Behler, M. Bernasconi, Phys. Rev. B 85, 174103 (2012)
• [3] G. C. Sosso and M. Bernasconi, MRS Bulletin 44, 705 (2019)
• [4] O. Abou El Kheir, L. Bonati, M. Parrinello, M. Bernasconi, npj Comp. Mater. 10, 33 (2024)
• [5] O. Abou El Kheir, M. Bernasconi, Adv. Electr. Mater. 11, e2500110 (2025)
• [6] D. Baratella, O. Abou El Kheir, and M. Bernasconi, Acta Materialia 284, 120608 (2025)
• [7] S. Marcorini, R. Pomodoro, O. Abou El Kheir, M. Bernasconi, J. Chem. Phys. 163, 154501 (2025)

Short Biography :

Marco Bernasconi is full professor of Theoretical Condensed Matter Physics at the University of Milano-Bicocca, Milano, Italy since 2016. He holds a Phd and a Master degree in Theoretical Condensed Matter Physics from SISSA-Trieste and an undergraduate degree in Physics from the University of Milano.  He spent a few years as a postdoctoral fellow at Sissa and at the Max-Planck-Institut FKF in Stuttgart (D) before moving to the University of Milano-Bicocca as an assistant and later associate professor.  He was the dean of the Phd program in Materials Science and Nanotechnology at the University of Milano-Bicocca in 2017-2022. His research activity is mostly dedicated to electronic structure calculations and atomistic simulations of materials for microelectronics and photonics. He published over 200 articles that received over 9000 citations (Scopus). In the last ten years his activity has been mostly focused on phase change materials for electronic memories.