Transferability Issues in Multiscale Modeling of Hierarchical Phenomena
This IRTG Minisymposium was organized by A. Chaimovich, A.C. Fogarty, V. Guzman, and J.F. Rudzinski in Mainz in December 7, 2015.
Along with computational and theoretical advances in multiscale modeling, the past two decades have witnessed a tremendous increase in powerful methodologies which provide systematic routines for building lower resolution, or coarse-grained (CG), models that optimally reproduce known properties of particular atomically-resolved systems. The effective potential energy functions employed in CG models must accurately account for the missing degrees of freedom with respect to the underlying, atomically-resolved, system. The dominant effects of these missing features on the remaining CG degrees of freedom may be dependent on the chemical make-up of the system, the thermodynamic state point of interest, or even the local environment of a single component within the system. Consequently, a CG model which is parameterized for a particular system or thermodynamic state point may not have sufficient accuracy for systems which differ from the parameterization data. In this case, the model is said to lack “transferability”. The problem of transferability is a fundamental issue in multiscale modeling, and its origin and consequences have been discussed extensively. However, there is no general theory or methodology for dealing with the inherent lack of transferability of systematically derived CG models. Moreover, developing transferable models is especially challenging for systems which exhibit hierarchical phenomena, i.e., systems whose structure or dynamics on large length and time scales are intimately linked to fine details of the microscopic interactions present on much smaller scales.
Purpose of the workshop
The workshop brought together a group of experts to discuss how to identify and address problems of transferability that arise when employing advanced multiscale modeling techniques. The purpose was highlight a few key questions: 1. What methods are available for systematically parameterizing a transferable model? 2. How can one anticipate or, eventually, predict errors in a given model away from the state point of parameterization? 3. How can one confidently attain predictive insight into complex systems using multiscale schemes?
The program can be found on https://trr146.de/download_file/TIMMHP_program.pdf
- Wissenschaftlicher Koordinator des TRR 146
- Dr. Giovanni Settanni
- Staudingerweg 9
- D-55128 Mainz