Perovskite Dynamics Analysis (PDynA) is a Python package for analysis of perovskite structural dynamics.
The Python documentation of PDynA can be found at https://pdyna.readthedocs.io/en/latest/.
-
The input to the code is molecular dynamics (MD) trajectories, currently readable formats are VASP-XDATCAR, XYZ, EXTXYZ, PDB (proteindatabank), ASE-Trajectory and LAMMPS dump files. The core class of PDynA is the
Trajectory
class. -
The structure recognition functions will automatically detect the constituent octahedral network and organic A-site molecules, and process the analysis.
-
The output is a selected set of the following properties: (pseudo-cubic) lattice parameter, octahedral distortion and tilting, time-averaged structure, A-site molecular orientation, A-site spatial displacement, radial distribution functions.
-
The octahedral distortion and tilting calculation is the core feature of this package, which can quantitatively examine the dynamic behaviour of perovskite in terms of how octahedra tilt and distort, as well as the spatial correlation of these properties (for example equivalent to Glazer notation in the 3C corner-sharing polytype).
-
The octahedral distortion and tilting calculation can be accelerated with parallelization by joblib, this can be tuned with the input parameter
multi_thread
.
- pdyna library containing:
- core.py: Contains the dataclass Trajectory and Frame as well as their related functions.
- structural.py: Handles structure recognition and property calculations.
- analysis.py: A collection of tools for computing and visualizing the output.
- io.py: The IO to input data files.
The main language is Python 3 and has been tested using Python 3.8+, with the following dependencies:
- Numpy
- Matplotlib
- Pymatgen
- Scipy
- ASE
1. From GitHub
PDynA can be installed with the following commands:
Clone the repository (or download manually)
git clone https://github.com/WMD-group/PDynA.git
cd to PDynA directory with the setup.py
file
cd pdyna
Install the package with pip
pip install .
2. From PyPI
Alternatively, PDynA can be installed directly from PyPI, with:
pip install pdyna
To start using the code, the user must define the B and X site elements in the BX6 octahedra. This can be done by defining the B and X sites species and their relative distances with the additional entry system_overwrite
in the dynamics
, for example,
traj.dynamics(..., system_overwrite=user_system)`
The input user_system
is a dict with four entries:
{'B-sites': ['Sn'], 'X-sites': ['I','Br'], 'fpg_val_BB': [[a,b], [c,d]], 'fpg_val_BX': [[a,b], [c,d]]}
The four numbers in B-B (fpg_val_BB
) and B-X (fpg_val_BX
) connectivity are:
c: NN1 distance of the connected pair
d: NN2 distance of the connected pair
a: lower bound of a range of distance that covers both and only NN1 and NN2
b: upper bound of a range of distance that covers both and only NN1 and NN2
If you are not sure about the relative distances in your system, you can run the system_test
function with one of your trajectories by calling:
traj = Trajectory(filetype,(file_path, MDtup))
traj.system_test(B_sites=['Sn'],X_sites=['I','Br']) # with Sn(I/Br) system as an example
This will give you two plots as follows:
An alternative way is to directly define the same quantities at the beginning of the Trajectory
class if you are working with only one system throughout.
_Xsite_species = ['Cl','Br','I']
_Bsite_species = ['Pb']
_fpg_val_BB = [[5,13.5], [8.2,12]] # the values for above plot
_fpg_val_BX = [[1,8], [3,6.8]] # the values for above plot
Two molecular dynamics trajectories are given in the examples
folder.
We encourage users to run orthogonal (the vectors connecting adjacent B-atoms are parallel to the principal axes) perovskite structures with corner-sharing connectivity to get the best performance of structure matching, when possible.
1. VASP-XDATCAR trajectory
The minimal example is based on a VASP-XDATCAR format trajectory, containing about 100 atoms.
Copy all the files under examples/minimal
to your working directory and run the script pdyna_example.py
.
2. LAMMPS trajectory
The second example is a larger LAMMPS trajectory (about 20,000 atoms), covering more functions in PDynA.
Copy all the files under examples/lammps
to your working directory, unzip lammps_example_mapbbr3.out.gz
and run the script example_lammps.py
.
For your own LAMMPS trajectory, please use a custom dump style with this command: dump dump_name all custom dump_frequency ./dump.out id element x y z
3. XYZ/PDB/ASE trajectory
Reading of these file types can be done similarly by running traj = Trajectory(filetype,(file_path, MDtup))
, where filetype
is one of "xyz", "pdb" or "ase-traj", the MD settings MDtup = (Ti, Tf, timestep, nblock)
(timestep
is MD time step in femtosecond, nblock
is the frequency of frame saving).
4. Parameters explained
A full list of all the parameters is shown in the examples/full.py
.
The computed dynamic properties are all stored in the Trajectory object and can all be accessed easily by calling traj.Tilting
, traj.Tilting_Corr
, traj.Distortion
, etc, which are octahedral tilting, octahedral tilting NN1 correlations (effective Glazor notation), octahedral distortions, respectively. The dimension of these arrays are (N_frames,N_octahedra,property_dimension).
@article{pdyna,
title={Structural dynamics descriptors for metal halide perovskites},
author={Liang, Xia and Klarbring, Johan and Baldwin, William J and Li, Zhenzhu and Cs{\'a}nyi, G{\'a}bor and Walsh, Aron},
journal={J. Phys. Chem. C},
volume={127},
number={38},
pages={19141--19151},
year={2023}
}
Python code and original data tables are licensed under the MIT License.
Please use the Issue Tracker to report bugs or request features in the first instance. For other queries about any aspect of the code, please contact Xia Liang by e-mail: xia.liang16[at]imperial.ac.uk.
- Xia Liang (Department of Materials, Imperial College London)