Neural Network Potentials in Materials Modeling
- 1.5k Downloads
The availability of reliable interatomic potentials is necessary for carrying out computer simulations of complex materials. While electronic structure methods like density functional theory have been applied with great success to many systems, the high computational costs of these methods severely restrict the scientific problems that can be studied. Consequently, in recent years a lot of effort has been spent on the development of more efficient potentials enabling large-scale simulations. In particular, machine learning potentials have received considerable attention, because they promise to combine the accuracy of first-principles methods with the efficiency of force fields. In this chapter an important class of machine learning potentials employing artificial neural networks will be reviewed and discussed.