Machine learned potentials for molecular dynamics simulations of halide perovskites

• Grant number: 24/01461-6
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: June 01, 2024
• End date: May 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Physics - Condensed Matter Physics
• Principal Investigator: Gustavo Martini Dalpian
• Grantee: Thiago Puccinelli Orlandi Nogueira
• Host Institution: Instituto de Física (IF). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Hybrid perovskites have been playing a crucial role in materials chemistry due to their diverse compositions and structural forms. When used in photovoltaic applications their efficiencies can exceed 25%. However, stability issues hinder their widespread use due to complex temperature dependent phase transitions and soft nature of the dynamics of structural motifs in the perovskites crystalline structure. In order to tackle those phenomena, machine learning techniques, especially effective potentials like Machine Learned Potentials, offer a cost-effective solution for accelerated simulations. Thus, this postdoctoral project proposes compiling a simulation dataset, developing these effective potentials from machine learning techniques for a comprehensive study of halide perovskites' free energy, and testing transferability across compositions. Success could significantly contribute to the accelerated discovery and optimization of materials, particularly in renewable energy applications, overcoming computational cost limitations in current methods such as Density Functional Theory and ab-initio Molecular dynamics computations. (AU)