Mapping Liquid Structure: Experimental Data for Machine-Learning Potentials

Institut

Professur für Multiscale Modeling of Fluid Materials (TUM-ED)

Typ

Inhalt

 

Beschreibung

In recent years, machine-learning interatomic potentials (MLIPs) have become powerful tools for simulating molecular

liquids. Their reliability, however, depends on high-quality experimental data for systematic validation and benchmark-

ing. Scattering experiments provide detailed structural information: neutron and X-ray diffraction probe collective

correlations, radial distribution functions (RDFs) describe local atomic organization, and bulk densities provide a

complementary benchmark. Yet these data are scattered across publications and repositories and vary in units, nor-

malization, experimental conditions, and reported uncertainties. A reusable, well-documented collection is therefore

still missing.
The project will establish a curated, provenance-aware dataset of experimental scattering and density data for multiple liquid solvents, enabling consistent validation of accurate and transferable MLIPs. 

The student will: 

• learn how scattering and molecular interactions reveal liquid structure and relate to RDFs and density;

• collect, standardize, and document experimental diffraction, RDF, and density data;

• create basic tools to check data quality, consistency, and provenance.

Voraussetzungen

• experience with Python programming and numerical data analysis;

• basic knowledge of molecular simulation, statistical mechanics, or materials characterization;

• an interest in machine learning, MLIPs, molecular simulation, or scattering experiments;

• careful and systematic working habits.

Möglicher Beginn

sofort

Kontakt

Claudio Colturi
claudio.colturitum.de

Ausschreibung