Dresden 2020 – wissenschaftliches Programm

Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

DY: Fachverband Dynamik und Statistische Physik

DY 35: Modelling and Simulation of Soft Matter I (joint session CPP/DY)

DY 35.10: Vortrag

Mittwoch, 18. März 2020, 12:15–12:30, ZEU 255

Chemically-transferable structure-based coarse-grained models — •Kiran H. Kanekal and Tristan Bereau — Max Planck Institute for Polymer Physics, Mainz, Germany

An attractive feature of the popular Martini [1] coarse-grained force field is its chemical transferability. Multiple chemical fragments can be assigned to the same Martini representation based on their similar hydrophobicity, maintaining thermodynamic accuracy in the resulting simulations [2]. However, since the Martini force field was optimized in order to reproduce certain thermodynamic properties of various condensed phase systems in a top-down fashion, it does not accurately portray the internal structures of these systems when compared to results from corresponding simulations at atomistic resolutions. On the other hand, implementing bottom-up coarse-graining approaches results in highly accurate structural properties by construction, yet these methods are traditionally performed on a small set of compounds, limiting their chemical transferability. In this work, we establish a new technique for building chemically-transferable coarse-grained models with high structural accuracy by coupling bottom-up coarse-graining methods with unsupervised machine learning. We validate our results by comparing our new model to Martini and show how changing the resolution of our coarse-grained model affects these results. 1. Marrink, S. J. & Tieleman, D. P. Chem. Soc. Rev. 42, (2013). 2. Menichetti, R., Kanekal, K. H. & Bereau, T. ACS Cent. Sci. 5, (2019).