Correlation function analysis from large scale molecular dynamics simulations

NAISS 2023/6-277


NAISS Medium Storage

Principal Investigator:

Paul Erhart


Chalmers tekniska högskola

Start Date:


End Date:


Primary Classification:

10304: Condensed Matter Physics



In several ongoing projects we need to analyze dynamical correlation functions on timescales from 100 ns to 1 µs via large-scale molecular dynamics (MD) simulations containing millions of atoms. These simulations generate very large trajectories that can easily reach many 100 GB in size. As a result, we require disk space considerably beyond the resources available to us via Small Storage projects, not only to run these simulations but to analyze them. Our workflow includes (1) generating trajectories and data, (2) analyzing these trajectories using in particular the dynasor package developed by us, and (3) removing the raw data. In other words, the purpose of this project is not to store data for longer time periods but rather to be able to turn over data on relatively short time scales. ## 2D halide perovskites Halide perovskites are promising for optoelectronic applications and we are currently running several projects concerning these materials focusing on phase transitions and dynamics. In 2D halide perovskites one or more inorganic perovskite units are linked via organic spacer molecules. The dynamics of the latter are particular complex. They introduce vibrational and rotational degrees of freedom with time scales that range from pico to the many nanoseconds. In order to understand these processes and compare with experiments, we aim to run large-scale molecular dynamics (MD) simulations using systems with millions of atoms over many nanoseconds, which requires a large amount of storage. ## Oxide perovskites This storage allocation would also enable us to analyze phase transitions as well as vibrational and Raman spectra of oxide perovskites. In these systems as one approaches continuous-order phase transitions, the length and time-scales typically diverge. This means that very large systems and long trajectories are required to be able to capture the phenomena observed experimentally. Here, we have already started with the analysis of the spectra of barium zirconate, which (again) proves challenging since it requires large amount of MD sampling to converge the vibrational spectra. Another project we have just started is concerning the ultrafast non-equilibrium dynamics of strontium titanate. Here, atomistic simulations provides unique opportunities to further the understanding of the non-equilibrium dynamics. In order to achieve statistics over these simulations, however, they must be repeated thousands or tens of thousands of times which yields large amount of data to be analyzed. ## Organic liquids Finally, we are also studying the structural and dynamics of liquid chromophores. These are liquids comprised of organic molecules such as benzene. The individual molecules remain intact and exhibit regular motion on the timescale of femtoseconds (typical C-H vibration). At low temperature the molecules will adopt long-range order, but at high temperatures the molecules start to diffuse and rotate similar to regular liquids. However, this rotational and diffusive motion happens on a nanosecond timescale and is thus much slower than the inter-molecular motion. This becomes a computational challenge, as simulating and analyzing dynamics on both the femto and nanosecond timescales requires very long simulations and large trajectories.