BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:MARVEL Distinguished Lecture - Georg Kresse
DTSTART:20210518T150000
DTEND:20210518T161500
DTSTAMP:20260509T054413Z
UID:9eb1a94d2261fb5abe2454ab89194af7ef11b4d237d59709fca08314
CATEGORIES:Conferences - Seminars
DESCRIPTION:Georg Kresse\, University of Vienna\nhttps://epfl.zoom.us/j/83
 610384294\nPasscode: 4258\n\nProf. Georg Kresse\nProfessor of Computationa
 l Materials Physics at the University of Vienna\n\nFinite temperature prop
 erties with first principles accuracy\, is machine learning the way to go
 ?\nAccurate predictions of phase transition temperatures have always been 
 a dream of materials physicists. Using first principles methods calculatio
 ns are usually extremely time-consuming and challenging\, whereas force fi
 elds without extensive and careful tuning tend to provide inaccurate answe
 rs. Machine-learned force fields are an obvious solution to this dilemma b
 ut training them can be a time-consuming and laborious process.\nIn this t
 alk\, I demonstrate that training on the fly yields highly accurate machin
 e-learned force fields (MLFF) that meet the challenges of predicting finit
 e temperature properties with an accuracy close to the original first-prin
 ciples method. Our machine learning approach is based on Bayesian regressi
 on and uses a combination of radial and angular features computed locally 
 for each atom. The Bayesian regression not only provides predictions for t
 he energies\, forces\, and stress tensor\, but also predicts the uncertain
 ty of these predictions. If the uncertainties exceed a certain threshold\,
  first principles calculations are performed “on the fly”\, the struct
 ure is added to the training data set\, and the MLFF is refined "on the fl
 y". Training is performed simply by heating (or cooling) all phases of int
 erest. Typically\, an accurate force field can be obtained in few days and
  the training requires no special intervention or expertise from the user.
 \nThe accuracy of the approach is demonstrated for several materials. For
  metallic zirconium\, our simulations successfully reproduce the first or
 der displacive martensitic phase transition from hcp to bcc Zr [1]. For Z
 r\, we also show that the MLFF reproduces phonon dispersions and elastic 
 properties with excellent precision. Zirconia (ZrO2) constitutes a more ch
 allenging test\, with two phase transitions from monoclinic to tetragonal 
 to cubic. Again\, the MLFF yields excellent predictions for both transitio
 n temperatures [2]. Moreover\, we are able to predict the thermal conducti
 vity in very good agreement with experiment. Melting temperatures of Al\, 
 Si\, Ge\, Sn and MgO are predicted using ML-FF trained using various densi
 ty functionals [3]. In this case\, we show that the differences between di
 fferent density functionals are far larger than the errors introduced by M
 L. Finally\, we address the phase transitions in hybrid perovskites – a 
 class of materials promising for thin film solar cells. Specifically\, we 
 calculate the phase transition temperatures of MAPbO3 and several other o
 rganic perovskites and find again very good agreement with experiment [4].
 \nApart of demonstrating that on the fly MLFFs provide excellent predictio
 ns on par with the original density functional\, we also show that diverse
  methods are required to calculated phase transition temperatures: these i
 nclude slow heating and cooling (Zr)\, thermodynamic integration (ZrO2)\, 
 interface pinning (melting temperatures) as well as umbrella sampling (MAP
 bI3) and free energy perturbation theory.\n[1] P. Liu\, C. Verdi\, F. Kars
 ai\, and G. Kresse\, submitted\n[2] C. Verdi\, F. Karsai\, P. Liu\, R. Jin
 nouchi\, and G. Kresse\, submitted\n[3] R. Jinnouchi\, F. Karsai\, G. Kres
 se\, Phys. Rev. B 100\, 014105 (2019).\n[4] R. Jinnouchi\, J. Lahnsteiner\
 , F. Karsai\, G. Kresse\, and M. Bokdam\, \nPhys. Rev. Lett. 122\, 225701
  (2019).\n\nAbout the speaker\nProfessor Georg Kresse received his docto
 ral degree from the Vienna  University of Technology in 1993. After his h
 abilitation at the Vienna University of Technology in 2001\, he was offere
 d a full professorship by both the University of Oxford and the University
  of Vienna in 2006.  In 2007 he accepted the chair for Computational Quan
 tum Mechanics in Vienna. Since 2011 Kresse is a full member of the Austr
 ian Academy of Sciences and since 2012 of the International Academy of Qua
 ntum Molecular Sciences. He is the recipient of several awards\, including
  the 2003 "START Grant" of the Austrian Science Fund (FWF)\, the "Hellmann
  Preis" of the Internationale Working group for  Theoretical Chemistry\, 
 and the Kardinal-Innitzer-Preis in 2016.\nHis main research interests are 
 Theoretical Solid State Physics\, Surface Sciences and Computational Mater
 ials Physics. His work on ab initio density functional theory has shaped t
 he application of density functional theory in materials sciences worldwid
 e. Georg Kresse is the main author and developer of the computer program
  "VASP" (Vienna ab initio simulation package)\, which is  the most widely
  used program for quantum mechanical simulations of solids and their surfa
 ces. The three publications on the algorithms implemented in VASP have bee
 n cited between 40.000 and 65.000 times each and belong to the 100 most ci
 ted research articles ever published.\nHis current work focuses on the pre
 cise description of electron interactions in solids and real materials and
  encompasses modern perturbative many-body theory\, quantum Monte Carlo me
 thods\, and machine learning. Georg Kresse is the author of more than 40
 0 research articles. With a Web of Sciences h-index of over 105 he is amon
 g the most cited physicists worldwide.\n\n\nDid you miss previous MARVEL 
 Distinguished Lectures? You can watch them on the Materials Cloud dedic
 ated page.\n\n 
LOCATION:https://epfl.zoom.us/j/83610384294?pwd=WktRTUU2UXFFSXBhb3JmTGNRUH
 BUdz09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR