CHE Seminars  

Speaker Dr. Martin Horsch
Computational Molecular Engineering, Department of Mechanical and Process Engineering, University of Kaiserslautern, Germany
Topic Quantitatively reliable massively-parallel molecular modelling and simulation of vapour-liquid interfaces
Date 30,December 2015 (Wednesday)
Place L-5 Lecture Hall Complex (LHC)
4 pm. - 5 pm.

To reach the quantitative precision required in engineering applications, molecular force field models of fluids need to be adjusted to experimental data for diversephysical properties. This implies that there are multiple conflicting objectivesduring model design, a conflict which is addressed here by computing the Pareto set,i.e. the set of all rational compromises between multiple objectives. Two local optimization algorithms,sandwiching and hyperboxing, are combined with a global exploration of the parameter space. For this purpose, a highly scalable molecular dynamics code is used: ls1 mardyn (large systems 1: molecular dynamics).

For a variety of low-molecular fluids, two-centre Lennard-Jones plus point-quadrupole (2CLJQ) and two-centre Lennard-Jones plus point-dipole (2CLJD) models are applied to describe bulk and interfacial properties of pure-componentvapour-liquid equilibria. The surface tension is computed at a high accuracy bymolecular dynamics simulation with a numerically efficient long-range correction for planar interfaces. Literature models, which were adjusted to bulk properties, but not to interfacial properties, are validated against the surface tension of real fluids. These models are on average found to overestimate the surface tension by 15 to 20%. By analysing the Pareto sets, it is discussed how the agreement with the surface tension can be improved while retaining a satisfactory accuracy for the bulk properties.

It is found that while it is usually possible to model two properties simultaneously with optimal accuracy (i.e. close to the accumulated error from experiment and simulation), it is impossible to reconcile the surface tension and multiple bulk properties to that level of accuracy. However, acceptable compromises can be reached which are visualized here by self-organizing patch plots. This approach enables users to design molecular models with very little effort, so that industrial users can experience more flexibility in employing molecular simulation in practice.

Dr Martin Horsch is a Junior Professor for Computational Molecular Engineering at the Department of Mechanical and Process Engineering of the University of Kaiserslautern, Germany (since 2011). Previously, he was a DAAD postdoctoral research fellow at the Department of Chemical Engineering of Imperial College London (2010/11). He obtained his doctorate in Mechanical Engineering at the University of Paderborn, Germany (2010), and his diploma (Master degree) in Computer Science at the University of Stuttgart, Germany (2006). He works on massively-parallel molecular simulation of soft matter with a focus on interfacial phenomena.

List of Seminars