סמינר מחלקה אלקטרוניקה פיזיקאלית: Natalia Kuritz

You are invited to attend a department seminar on

Classical and Statistical Learning based methods for materials modeling for energy applications.
By:
Natalia Kuritz
PhD student under the supervision of Dr. Amir Natan
Abstract
Atomistic level simulations are important for the understanding of complex materials and systems. Energy applications as well as the development of new optical-active materials require the understanding of both quantum level, single particle phenomena, and a collective, statistical behavior of an ensemble of molecules/atoms. A computationally affordable approach to describe a material behavior at an atomistic level is the use of classical Molecular Dynamics (MD) simulations. To describe the electronic structure, more expensive quantum calculations, such as Density Functional Theory (DFT), are needed. In this PhD work I use both classical MD and DFT to provide theoretical trends and understanding in the fields of metal-air batteries [1,2] and bio-inspired materials [3.4]. In the second part of this work I present a local deep learning model which I developed to predict atomic forces with DFT accuracy but at a much lower computational cost [5]. Such a model can combine the advantages of classical MD speed and scaling with the accuracy of DFT calculations and hence allow the calculation of larger interfaces and surfaces.

References: [1] J. Phys. Chem. B 120 (13), pp 3370–3377 (2016), [2] J. Electrochem. Soc. (JES) 165 (13) A3095-A3099 (2018), [3] Langmuir 32, 2847 (2016), [4] J. Phys. Chem. A, 123 (9), 1758-1765 (2019), [5] Phys. Rev. B 98, 094109 (2018)

On Sunday, Dec 29th, 2019, 16:30
Room 011, EE-Class Building