Back to overview

Development of a Neural Network Potential with Accurate Electrostatic Interactions

Applicant Goedecker Stefan
Number 182877
Funding scheme Project funding (Div. I-III)
Research institution Departement Physik Universität Basel
Institution of higher education University of Basel - BS
Main discipline Material Sciences
Start/End 01.04.2019 - 31.03.2022
Approved amount 166'005.00
Show all

All Disciplines (2)

Material Sciences
Information Technology

Keywords (4)

Machine learning; force fields; charge equilibration; neural networks

Lay Summary (German)

Wir werden Methoden entwickeln, die es erlauben eine groessere Anzahl von Materialien mit sowohl ionischen als auch kovalenten Bindungen mit Machine Leraning Techniken zu simulieren.
Lay summary
Atomistische Simulationen benoetigen einen sehr grossen Rechenaudwand, wenn die Kraefte, die auf die Atome wirken, quantenmechanisch berechnet werden. Neu entwickelte Machine Lerning Methoden koennen die quantenmecha Ergebnisse oft viel effizienter reproduzieren, es koennen jedoch auch
unphysikalische Vorhersagen gemacht werden. Um solche unphysikalischen Ergebnisse zu vermeiden, ist es wichtig, dass das Maschine Learning Schema auf einem soliden physikalischen Formalismus beruht.
In diesem Projekt werden wir ein Maschine Learning Schema entwickeln, das sowohl elektronischen
Ladungstransfer und ionische Bindungen als auch kovalente Bindungen beschreiben kann. Es wird auf einem
Charge Equilibarion Formalismus beruhen wo alle relvanten Groessen mittels eines neuronalen Netzwerkes bestimmt werden. Die neuen Methoden werden wichtig sein fuer die Entwicklung neuer Materialien, Arzneimittel und Chemikalien.
Direct link to Lay Summary Last update: 16.02.2019

Responsible applicant and co-applicants


Name Institute

Associated projects

Number Title Start Funding scheme
165554 Structure and dynamics of materials based on advanced electronic structure calculations 01.05.2016 Project funding (Div. I-III)
154474 Impact of composition and nanometer scale DISorder in transparent Conductive Oxides: a new route to design materials with enhanced transport properties (DisCO) 01.01.2015 Sinergia
141828 NCCR MARVEL: Materials’ Revolution: Computational Design and Discovery of Novel Materials (phase I) 01.05.2014 National Centres of Competence in Research (NCCRs)


In recent years, a new generation of interatomic potentials based on machine learning techniques has been introduced. These potentials, which provide a direct functional relation between the atomic positions and the potential-energy, combine the accuracy of electronic structure methods with the efficiency of simple empirical potentials. Because of the absence of system-specific terms they allow to perform extended simulations of a large variety of systems. Most of these potentials rely on atomic properties like energies and charges depending only on the local chemical environments of the atoms. Such local charges are, however, unable to capture long-range charge transfer. This prevents the accurate description of systems in which distant structural features have global effects on the charge distribution in the system. Examples for such systems are semiconductors including defects, polar surfaces of oxides and metal-organic molecules with different possible metal oxidation states. In order to overcome these intrinsic limitations of current machine learning potentials, we propose to combine high-dimensional neural networks with the charge equilibration neural network technique. The resulting new method will be generally applicable to all types of systems, which we will demonstrate by analyzing the potential-energy surfaces of different model systems covering all types of bonding using the minima hopping method.