Information storage based on phase-change materials (PCM) is widely considered a promising alternative to flash memories for the non-volatile memory technologies of the next decade.
Non-volatile memories based on PCMs are made by a thin film of chalcogenides, which is highly conductive in the crystalline phase and insulating in the amorphous phase.
The most known and used PCM is the so-called GST (Ge2Sb2Te5) which is employed in data-storage applications. The GST liquid, crystalline and amorphous phases have already been characterized using ab-initio and machine learning simulations1,2 and, consequently, their electronic structure. The enrichment of GST with Ge showed a marked increase in the crystallization temperature, but the segregation of Ge atoms has been observed, raising problems of atomic diffusion and thermodynamical properties.
The project objective is to achieve an accurate description of Ge-rich GST amorphous materials cannot be reached using only ab-initio simulations, that are limited to short timescales (few tenth of ps) and to few hundreds of atoms.
For this reason, classical molecular dynamics (MD) simulations, able to reach several ns for thousands of atoms, must be employed. For MD simulations, the force field of Ge-rich GST will be derived from a machine learning approach using as input the structures and energies coming from ab-initio simulations, obtaining a high structural and thermodynamical accuracy extended over long timescales and for large systems.
Innovation potential: the proposed project aims at obtaining two high-impact objectives: 1) a structural, electronical, mechanical and thermodynamical description of Ge-rich GST materials and 2) a freely-available database of Ge-rich GST structures that will be employed by other scientists to study these systems without repeating same simulations.
The goal is to identify a direct relationship between the microscopic properties of materials and the performances of storage-class memories.