Related Experiment Video
Updated: Nov 22, 2025
10:52
Multiscale Sampling of a Heterogeneous Water/Metal Catalyst Interface using Density Functional Theory and Force-Field Molecular Dynamics
Published on: April 12, 2019
13.1K
Machine Learning in Computational Surface Science and Catalysis: Case Studies on Water and Metal-Oxide Interfaces
Xiaoke Li1, Wolfgang Paier2, Joachim Paier1
1Institut für Chemie, Humboldt-Universität zu Berlin, Berlin, Germany.
Frontiers in Chemistry
|January 11, 2021
Summary
Machine learning bridges atomistic and macroscopic scales in materials simulations. This study uses on-the-fly machine-learned force fields to accurately model oxide surfaces and water adsorption.
Area of Science:
- Computational Physics
- Materials Science
- Chemistry
Background:
- Bridging atomistic and macroscopic length/time scales in simulations is a key challenge.
- Machine learning (ML) offers promising solutions for multiscale modeling.
Purpose of the Study:
- To apply on-the-fly ML force fields for simulating oxide materials.
- To investigate the generalizability and accuracy of ML force fields for surface properties and water adsorption.
Main Methods:
- Utilized Gaussian approximation potentials with Bayesian regression and active learning.
- Employed the Vienna ab initio simulation package (VASP) for molecular dynamics simulations.
- Tested ML force fields on simple (MgO) and complex (iron oxide) metal oxides.
Main Results:
- Successfully modeled surface properties of pristine and reconstructed MgO and Fe3O4 surfaces.
- Demonstrated the application of ML force fields to water adsorption on metal oxides.
- Identified challenges in accurately describing water-oxide interfaces, particularly for iron oxides.
Conclusions:
- On-the-fly ML force fields show potential for multiscale simulations of materials.
- Accurate modeling of water-oxide interfaces requires further research and development.
- ML-driven simulations offer new avenues for understanding complex material behaviors.
