Efficient parallel HPC implementation of pH-dependence in molecular dynamics simulation using LAMMPS

Antragssteller

Prof. Dr. Dirk Zahn
ComputerChemieCentrum – LS Theoretische Chemie
Friedrich-Alexander-Universität Erlangen-Nürnberg

Projektübersicht

Molecular simulations have made significant contributions to our current understanding of solidsolvent interfaces, crystal nucleation and growth as well as polymorphic transitions in the precipitate. In a similar manner, the in-depth characterization of crystal dissociation could benefit from molecular simulation considerably. Acid-induced dissociation of solids is a common route to dissolving crystals with abundant application ranging from the laboratory to household cleaning and pharmacokinetics. The efficient modelling of pH-dependence however still imposes an unresolved challenge to high-performance computing. While fundamental theory and selected application studies are subject to ongoing research in our group, the purpose of the present proposal is to drastically improve the software implementation.

In particular, we aim at extending the parallel replica module of LAMMPS to enable the addressing of thousands of possible protonation steps in parallel. Typical simulation systems comprise 105 – 106 atoms to model 10 nm sized crystallites embedded in solution, and as stand-alones, would call for mid-sized computing resources. This, however only applies if possible protonation reactions are considered one after the other. While this sequential approach actually reflects the current state-of-the-art in modelling acid-induced crystal dissociation, parallel-replica techniques offer practically ideal parallelization by characterizing different protonation sites simultaneously.

The envisaged code implementation to LAMMPS (open source) shall extend the already wellestablished parallel-replica scheme based on different temperature and/or different interaction potentials. Project success shall enable the scientific community (applications range from chemistry, physics and biology to particle engineering and pharmaceutical formulation) to userfriendly parallel-Hamiltonian molecular dynamics simulations of pH-dependent model systems. The developed code will be open source basis and available to academic use free of charge.