BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Four Generations of High-Dimensional Neural Network Potentials - P
 rof. Jörg Behler (Ruhr-Universität Bochum)
DTSTART:20250619T130000Z
DTEND:20250619T143000Z
UID:TALK227407@talks.cam.ac.uk
CONTACT:Bo Peng
DESCRIPTION:Machine learning potentials (MLPs) have become an important to
 ol for atomistic simulations in many fields\, from chemistry to materials 
 science. The reason for the popularity of MLPs is their ability to provide
  very accurate energies and forces\, which are essentially indistinguishab
 le from the underlying reference electronic structure calculations. Still\
 , the computational costs are much reduced enabling large-scale simulation
 s of complex systems. Almost two decades ago\, in 2007\, the introduction 
 of high-dimensional neural network potentials (HDNNP) by Behler and Parrin
 ello paved the way for the application of MLPs to condensed systems contai
 ning a large number of atoms. Still\, the original second-generation HDNNP
 s\, like most current MLPs\, are based on a locality approximation of the 
 atomic interactions that are truncated at some finite distance. Third-gene
 ration MLPs contain long-range electrostatic interactions up to infinite d
 istance and overcome this restriction to short-range energies. Still\, the
 re are surprisingly many systems in which long-range electrostatic interac
 tions are insufficient for a physically correct description\, since non-lo
 cal phenomena like long-range charge transfer are essential. Such global e
 ffects can be considered in fourth-generation HDNNPs. In this talk the evo
 lution of HDNNPs will be discussed along with some key systems illustratin
 g their applicability.
LOCATION:Seminar Room 3\, RDC
END:VEVENT
END:VCALENDAR