Applicant
Prof. Dr. Jens Harting
Department of Chemical and Biological Engineering and Department of Physics, FAU
Helmholtz Institute Erlangen-Nürnberg for Renewable Energy
Project Overview
Complex fluids are systems composed of one or more fluid phases (liquids, gas) within which particles on a molecular and colloidal scale are suspended in. The multiphase, multicomponent and heterogeneous nature of these systems leads to very complex rheological behaviour. This includes dynamical and rheological phase transitions akin to their composition (multicomponent and multiphase fluids) and structural complexity (heterogeneity and clustering) involving many-body interactions and multiscale problems. Understanding, predicting and controlling rheology is not just an academic challenge but also an industrial one. In both respects, computer simulations allow the probing of regimes where theoretical approaches lose predictive power, at the same time with a level of control over the parameters that are usually not possible in experiments.
Our group is maintaining, developing and using the software package 3nskog (former LB3D). Apart from our group, it is used by several scientific groups and projects internationally. The code is written for use on multi-CPU architectures and utilizes the Message Passing Interface (MPI) for synchronization and communication between processors.
The versatility of the LB3D package offers a wide range of applications to the users, as a result of the embedding of the project team into a scientific research institute. Performance considerations have traditionally not been prioritized over accomplishment of scientific objectives. However, previous investments in the computational throughput (in particular, single-component LBM), have proven to be greatly beneficial for the code performance, and the research group has been profiting from this progress ever since.
It is our goal to re-calibrate our optimization patterns to the current state of HPC and enable users to design and conduct simulations responsibly.