Special MechE Colloquium: Physics based and data-driven multiscale materials modelling
Event details
| Date | 30.04.2020 |
| Hour | 16:00 › 17:00 |
| Speaker | Prof. Michele Ceriotti, Laboratory of Computational Science and Modelling, School of Engineering, Institute of Materials, EPFL |
| Location | |
| Category | Conferences - Seminars |
Abstract:
Machine learning models are proving to be extremely effective in predicting the properties of atomistic configurations of matter, circumventing the need for time consuming electronic structure calculations when modeling materials at the atomic scale. The most successful schemes achieve transferability by means of a local representation of structures, in which the problem of predicting a property is broken down into the prediction of local, atom-centered contributions. I will presented an overview of these approaches, including examples of applications to different classes of materials. Locality, however, breaks down when describing long-range inter-atomic forces, such as those arising due to electrostatic interactions. I will present a possible solution to this conundrum based on the long-distance equivariant (LODE) framework, that combines a local description of matter with the appropriate, long-range asymptotic behavior of interactions.
Bio:
Michele Ceriotti is a Tenure-Track Assistant Professor in the department of Materials Science at EPFL where he has established the Laboratory of Computational Science and Modeling (COSMO). His research spans different classes of compounds, including hydrogen-bonded compounds, metals and materials for energy applications, with the goal of increasing both the predictive and interpretative power of computer simulations when it comes to understanding the relationships between structure and properties of materials. Dr. Ceriotti obtained his PhD in Physics from ETH Zurich in 2010, working in the group of Michele Parrinello to develop algorithms to improve several aspects of molecular dynamics simulations. These included linear-scaling electronic structure methods to simulate larger systems, a novel framework to use correlated-noise Langevin dynamics to manipulate with exquisite precision the sampling properties of molecular dynamics, and a non-linear dimensionality reduction method to describe in a coarse-grained manner the configuration space of structurally complex materials. After graduating, he moved to Oxford. After a brief collaboration with Andrea Cavalleri and Nicola Marzari, he joined the group of David Manolopoulos in the department of Theoretical Chemistry. He combined path integral molecular dynamics and correlated-noise generalized Langevin equations to dramatically reduce the computational burden associated with the modeling of the quantum properties of light nuclei. These enhanced methods made it possible to understand some features of the behavior of water, including quantum fluctuations of the hydrogen bond, isotope effects on the melting of water,and isotope fractionation at the water-vapor interface. He joined EPFL in Fall 2013.
Machine learning models are proving to be extremely effective in predicting the properties of atomistic configurations of matter, circumventing the need for time consuming electronic structure calculations when modeling materials at the atomic scale. The most successful schemes achieve transferability by means of a local representation of structures, in which the problem of predicting a property is broken down into the prediction of local, atom-centered contributions. I will presented an overview of these approaches, including examples of applications to different classes of materials. Locality, however, breaks down when describing long-range inter-atomic forces, such as those arising due to electrostatic interactions. I will present a possible solution to this conundrum based on the long-distance equivariant (LODE) framework, that combines a local description of matter with the appropriate, long-range asymptotic behavior of interactions.
Bio:
Michele Ceriotti is a Tenure-Track Assistant Professor in the department of Materials Science at EPFL where he has established the Laboratory of Computational Science and Modeling (COSMO). His research spans different classes of compounds, including hydrogen-bonded compounds, metals and materials for energy applications, with the goal of increasing both the predictive and interpretative power of computer simulations when it comes to understanding the relationships between structure and properties of materials. Dr. Ceriotti obtained his PhD in Physics from ETH Zurich in 2010, working in the group of Michele Parrinello to develop algorithms to improve several aspects of molecular dynamics simulations. These included linear-scaling electronic structure methods to simulate larger systems, a novel framework to use correlated-noise Langevin dynamics to manipulate with exquisite precision the sampling properties of molecular dynamics, and a non-linear dimensionality reduction method to describe in a coarse-grained manner the configuration space of structurally complex materials. After graduating, he moved to Oxford. After a brief collaboration with Andrea Cavalleri and Nicola Marzari, he joined the group of David Manolopoulos in the department of Theoretical Chemistry. He combined path integral molecular dynamics and correlated-noise generalized Langevin equations to dramatically reduce the computational burden associated with the modeling of the quantum properties of light nuclei. These enhanced methods made it possible to understand some features of the behavior of water, including quantum fluctuations of the hydrogen bond, isotope effects on the melting of water,and isotope fractionation at the water-vapor interface. He joined EPFL in Fall 2013.
Practical information
- General public
- Free