Da
21
Mag
Seminari e Convegni
Atomistic Simulations for Energy Related Materials with Machine Learning-Based Interatomic Potentials
The course titled "Atomistic Simulations for Energy Related Materials with Machine Learning-Based Interatomic Potentials" will be taught by Umberto Raucci and Francesco Mambretti of Istituto Italiano di Tecnologia.
Description
The need for efficient energy storage and clean energy sources like hydrogen is increasing, driving the importance of characterization of materials at the atomistic scale, which is crucial to enhancing performance, particularly under operando conditions. Experimental characterization at atomic resolution is challenging yet crucial, while ab initio atomistic simulations offer a theoretical alternative to studying atomic-scale dynamics. However, the high cost of ab initio molecular dynamics restricts its use in realistic systems. recently, machine learning (ML)-based interatomic potentials have provided a practical way to balance ab initio accuracy with classical force field efficiency. These ML potentials, trained on extensive quantum mechanical data, allow for simulations of larger systems over longer times. However, despite its power, constructing these potentials represents a complex endeavor.
This PhD course for engineering students, even those with minimal molecular simulations knowledge, offers a comprehensive overview of the current state-of-the-art methodologies in the field. We will discuss training ML potentials, collecting necessary data, and the pros and cons of each method. Additionally, the course will feature illustrative applications to deepen participants’ understanding of structural dynamics, chemical kinetics, and transport phenomena in energy systems, focusing on specific examples like temperature effects on cathode materials and catalyst-involved chemical reactions. The hands-on section will cover the complete training of a ML potential through case studies using Quantum Espresso (QE), DeepMD-Kit, MACE, LAMMPS, and Python.
Lecturers
Umberto Raucci obtained his Ph.D in Chemical Science at the University Federico II of Naples. He is now a Postdoctoral Researcher in the Atomistic Simulations group led by Prof. Parrinello at the Italian Institute of Technology, where he moved after a postdoctoral experience in the Martínez group at the Stanford University. Umberto develops and applies theoretical methods to discover new chemical reactions in complex environments using enhanced sampling techniques and machine learning based interatomic potentials.
Francesco Mambretti is a physicist and works as Postdoctoral Researcher in the Atomistic Simulations group, at the Italian Institute of Technology, in Genova. He is expert in Molecular Dynamics, Machine Learning and ab-initio simulations. His current research focuses on the study of catalytic reactions for hydrogen production and for graphene growth.
Classes will be held from May 21st to May 23rd, 2024, from 10 am to 1 pm.
To enroll please contact: matteo.fasano@polito.it
Description
The need for efficient energy storage and clean energy sources like hydrogen is increasing, driving the importance of characterization of materials at the atomistic scale, which is crucial to enhancing performance, particularly under operando conditions. Experimental characterization at atomic resolution is challenging yet crucial, while ab initio atomistic simulations offer a theoretical alternative to studying atomic-scale dynamics. However, the high cost of ab initio molecular dynamics restricts its use in realistic systems. recently, machine learning (ML)-based interatomic potentials have provided a practical way to balance ab initio accuracy with classical force field efficiency. These ML potentials, trained on extensive quantum mechanical data, allow for simulations of larger systems over longer times. However, despite its power, constructing these potentials represents a complex endeavor.
This PhD course for engineering students, even those with minimal molecular simulations knowledge, offers a comprehensive overview of the current state-of-the-art methodologies in the field. We will discuss training ML potentials, collecting necessary data, and the pros and cons of each method. Additionally, the course will feature illustrative applications to deepen participants’ understanding of structural dynamics, chemical kinetics, and transport phenomena in energy systems, focusing on specific examples like temperature effects on cathode materials and catalyst-involved chemical reactions. The hands-on section will cover the complete training of a ML potential through case studies using Quantum Espresso (QE), DeepMD-Kit, MACE, LAMMPS, and Python.
Lecturers
Umberto Raucci obtained his Ph.D in Chemical Science at the University Federico II of Naples. He is now a Postdoctoral Researcher in the Atomistic Simulations group led by Prof. Parrinello at the Italian Institute of Technology, where he moved after a postdoctoral experience in the Martínez group at the Stanford University. Umberto develops and applies theoretical methods to discover new chemical reactions in complex environments using enhanced sampling techniques and machine learning based interatomic potentials.
Francesco Mambretti is a physicist and works as Postdoctoral Researcher in the Atomistic Simulations group, at the Italian Institute of Technology, in Genova. He is expert in Molecular Dynamics, Machine Learning and ab-initio simulations. His current research focuses on the study of catalytic reactions for hydrogen production and for graphene growth.
Classes will be held from May 21st to May 23rd, 2024, from 10 am to 1 pm.
To enroll please contact: matteo.fasano@polito.it