Just another WordPress site - Ruhr-Universität Bochum

Department Artificial Intelligence for Integrated Material Science
Research Group

Discovery and Characterization of Inorganic Stoichiometric Compounds

Our research group combines high-throughput first-principles (ab initio) calculations and machine learning techniques to accelerate materials discovery.

Dr. Haichen Wang

Postdoctoral Researcher

Room: 00-105
Tel.:
E-Mail: haichen.wang@rub.de

Research

We focus on large-scale screening of candidate materials throughout the entire chemical space and across diverse structural motifs. Our work addresses fundamental condensed matter phenomena including superconductivity, phonon dynamics, optical absorption, and magnetic ordering.

Examples of 2D prototypes in Alexandria database.

ICAMS, RUB

Discovery of novel materials
We employ high-throughput screening strategies to identify compounds with advanced functionalities, targeting applications in photovoltaics, UV absorption, superconductivity, and altermagnetism. Our work emphasizes complex and low-symmetry systems, particularly those exhibiting chiral structures, reduced dimensionality, or solid-solution behavior. We characterize exotic electronic properties of solids, including topological flat bands, strong electron–phonon coupling, frustrated magnetism, and related phenomena.

Development of machine-learning interatomic potentials
We curate large, universal datasets for training machine-learning interatomic potentials that accurately predict material properties and response functions. We are particularly interested in response properties with respect to perturbations, including but not limited to atomic displacements, strain, and external electric or magnetic fields.

Competences

Density functional theory (DFT) simulations
Machine-learning interatomic potentials
High-throughput computational materials screening

Members

Recent Publications

Research Examples

Predicting stable crystalline compounds using chemical similarity

We propose an efficient high-throughput scheme for the discovery of stable crystalline phases based on the substitution of atoms in the crystal structure with chemically similar ones. The concept of similarity is defined quantitatively using data-mining experimental databases.

Training machine learning interatomic potentials for accurate phonon properties

We develop an extensive dataset of phonon calculations using density-functional perturbation theory (DFPT). We then show how this dataset can be used to train neural-network force fields, by implementing the training and the prediction of force constants in periodic crystals.