Simulation and Analyses for "Defect-Driven Anomalous Transport in Fast-Ion Conducting Solid Electrolytes"
This reponsitory contains the code scripts and sample data for the manuscript "Defect-Driven Anomalous Transport in Fast-Ion Conducting Solid Electrolytes", which is undergoing peer review as of May 13, 2021.
The large-scale classical molecular dynamics simulations are run via LAMMPS. Full datasets are not included.
Disclaimer: we do not claim this is good code. This work has been a learning experience more than anything else.
sample_data: two sample datasets, one for Na beta"-alumina, and one for Ag beta-alumina.sherlock_scripts: shell, python, and LAMMPS scripts for running simulations and extracting hopping events and statistical descriptors of transport.structures: sample crystal structures with quasi-randomly generated distributions of defect atoms.templates: template files for generating crystal structures.utils: supporting python functions.- python scripts: generating crystal structures for simulation, and making figures in the manuscript.
Python module dependencies are:
Python dependencies can be installed via requirements.txt :
pip install -r requirements.txt
Parallelization of several analyses is done with parallel.
We have used the scientific color map batlow (Crameri 2018) to prevent visual distortion of the data and exclusion of readers with color vision deficiencies (Crameri et al., 2020).
Python scripts are organized similar to jupyter notebooks, and are best run in blocks/cells using the spyder IDE.
Python scripts using networkx and freud sample random configurations of defect placements, starting with DL_POLY CONFIG template files. Then LAMMPS input .lmp files are created optimizing for most-representative structures. The two sets of scripts are make_repeating_structures_[phase].py and make_structure_[phase].py, where phase is beta, beta'', and Mg-doped beta. The outputs are under structures.
Sample LAMMPS input files are included. These are easy to modify for multiple mobile ions and temperatures using only variable declarations at the beginnings of the files. Because of sherlock job duration limitations, restarting was used for 100-nanosecond-long simulations. Shell scripts for slurm are included.
LAMMPS trajectories for mobile ions and their center of mass are parsed to individual atom trajectories. Then, hopping events within each two-dimensional conduction plane are identified as boundary crossings using the input LAMMPS crystal structures; this relies on the two-dimensional nature of conduction and ignores small motions of the host atoms. Then the hopping events from individual conduction planes within each simulation are combined.
Statistical descriptors of transport are computed from the trajectories of mobile ions: mean-square displacements, fourth moment of displacements, and non-Gaussian parameter in two dimensions, distribution of displacements along [100], self part of the van Hove function, and ergodicity breaking parameter. Python scripts and shell scripts for this are included in sherlock_scripts.
Python code for generating figures in the main text and supporting information of the manuscript is included manuscript_figures_[micro/macro].py. The micro script deals with atomistic hopping events, and the macro script deals with statistical descriptors, including the key quantity diffusion kernel correlation.
The "micro" script compiles a .csv listing of all simulation planes, which is used by the "macro" script.
As it is not possible to include all data via GitHub, most figure-making blocks/cells will not immediately run.
Two sets of sample data are included. These are limited by GitHub constraints on the size of data files.
- Na beta''-alumina 300 K. 20 ns of center-of-mass trajectory, 20 ns of hopping events, extracted statistical descriptors (from 100 ns of trajectories).
- Ag beta-alumina 300 K. 20 ns of center-of-mass trajectory, 10 ns of hopping events, extracted statistical descriptors (from 100 ns of trajectories).