Applications of Neural Network Potentials
Jörg Behler
Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum
On this page we plan to collect some results, which have been obtained using Neural Network potentials.
The following examples are currently available:
Chemical Processes at Surfaces
Oxygen Dissociation at the Al(111) Surface
|
|
Using constrained DFT we could show that the
low sticking probability of thermal oxygen molecules
at the Al(111) surface is governed by spin-selection rules.
A large number of MD trajectories has been calculated
employing a Neural Network potential-energy surface.
|
Related publications:
J. Behler, B. Delley, S. Lorenz, K. Reuter and M. Scheffler, Phys. Rev. Lett. 94, 36104 (2005).
J. Behler, B. Delley, K. Reuter and M. Scheffler, Phys. Rev. B 75, 115409 (2007).
J. Behler, S. Lorenz, and K. Reuter, J. Chem. Phys. 127, 014705 (2007).
J. Behler, K. Reuter, and M. Scheffler, Phys. Rev. B 77, 115421 (2008).
C. Carbogno, J. Behler, A. Gross, and K. Reuter, Phys. Rev. Lett. 101, 096104 (2008).
C. Carbogno, J. Behler, K. Reuter, and A. Gross, Phys. Rev. B 81, 035410 (2010).
Molecular Systems
Neural Network Potential-Energy Surfaces for Organic Molecules: R,R-Tartaric Acid
|
|
A Neural Network potential for R,R-tartaric acid is used as a benchmark system to test the applicability to molecular systems.
|
Neural Network Potential-Energy Surfaces for Water
|
|
A Neural Network potential for flexible water molecules is currently being developed, which allows also for a dissociation of water molecules.
|
Related publications:
T. Morawietz, V. Sharma, and J. Behler, J. Chem. Phys., 136, 064103 (2012).
Materials Science
High-Pressure Phases of Silicon
|
|
In a sequence of metadynamics simulations
based on a Neural Network potential the full sequence
of pressure-induced phase transitions in silicon could be obtained.
|
Related publications:
J. Behler, and M. Parrinello, Phys. Rev. Lett. 98, 146401 (2007).
J. Behler, R. Martonak, D. Donadio, and M. Parrinello, Phys. Rev. Lett. 100, 185501 (2008).
J. Behler, R. Martonak, D. Donadio, and M. Parrinello, phys. stat. sol. (b) 245, 2618 (2008).
A Neural Network Potential for Sodium
|
|
An interatomic potential for the high-pressure and
high-temperature crystalline and liquid phases
of sodium is created using a Neural Network
representation of the DFT potential-energy surface.
|
Related publications:
H. Eshet, R.Z. Khaliullin, T.D. Kühne, J. Behler, and M. Parrinello, Phys. Rev. B 81, 184107 (2010).
H. Eshet, R.Z. Khaliullin, T.D. Kühne, J. Behler, and M. Parrinello, Phys. Rev. Lett., accepted (2012).
Study of the Graphite-Diamond Phase Coexistence Employing a Neural Network Potential-Energy Surface
|
|
Based on a Neural Network potential the
thermodynamics of the graphite-diamond coexistence has been studied by molecular dynamics.
|
Related publications:
R.Z. Khaliullin, H.Eshet, T.D. Kühne, J. Behler, and M. Parrinello, Phys. Rev. B 81, 100103 (2010).
R.Z. Khaliullin, H.Eshet, T.D. Kühne, J. Behler, and M. Parrinello, Nature Materials 10, 693 (2011).
A Neural Network Potential-Energy Surface for Copper
|
|
A Neural Network potential for studying real copper surfaces including steps, vacancies and adatoms.
|
Related publications:
N. Artrith, and J. Behler, Phys. Rev. B 85, 045439 (2012).
A Neural Network Potential-Energy Surface for Zinc Oxide
|
|
We have constructed a neural network potential for zinc oxide crystals, surfaces and clusters. Electrostatic interactions are taken into account explicitly.
|
Related publications:
N. Artrith, T. Morawietz, and J. Behler, Phys. Rev. B 83, 153101 (2011).
A Neural Network Potential-Energy Surface for phase change materials
|
|
Phase change materials are important
materials for optical storage devices.
The goal of our studies is to understand
the basic principles of the transitions
between the crystalline and amorphous
phases of GeTe.
|
Related publications:
G.C. Sosso, D. Donadio, S. Caravati, J. Behler, and M. Bernasconi, submitted (2012).
G.C. Sosso, G. Miceli, S. Caravati, J. Behler, and M. Bernasconi, submitted (2012).
We are currently working on several other systems. The results will be added to this page as soon as possible.
Many of the results have been obtained with our neural network code RuNNer. Further information about this code can be found here.