Tutorial-0X: Miscellaneous Assists

Some of the calculations required to obtain the data to run e.g. AMSET and Phono3py can be very expensive. To reduce that burden, there are a few basic functions to make calculations more efficient. The first is specific to VASP, but the rest are more general.

Zero-Weighted k-points

AMSET and BoltzTraP require dense k-point grids to get accurate results. Not all k-points are created equal, however, so we have provided a tool to combine two KPOINTS files, the converged KPOINTS file which must be weighted, and a second file of less equal k-points which can be zero- weighted, to increase the population without costing so much as their bretheren.

tp.setup.get_kpoint('weighted_KPOINTS', 'unweighted_KPOINTS')

tp gen kpoints -k weighted_KPOINTS -z unweighted_KPOINTS

When setting KPAR, unweighted k-points should not be considered. Our KPAR generator suggests suitable KPAR values, ignoring the zero- weighted k-points.

tp.setup.get_kpar('KPOINTS')

tp gen kpar -k KPOINTS

Phonopy Config Files

If you generate phonopy data from the command line, configuration files are useful to save time regenerating and record inputs. ThermoParser will generate these for you based on a POSCAR and your inputs.

tp.setup.get_band_conf('supercell size')
tp.setup.get_dos_conf('supercell size')

tp gen band-conf 'supercell size'
tp gen dos-conf 'supercell size'

The required argument 'dim', i.e. the supercell size, can be a string or an array (the latter in python only), and accepts 1x1, 3x1, 3x3 and 6x1 arrays, that is to say 2, '2 2 2', '2 0 0  0 2 0  0 0 2' and '2 2 2 0 0 0' all give the same result.

Target Lattice Thermal Conducitivity

The kappa-target plot shows what lattice thermal conductivity would be required to achieve a specified ZT. If it’s too low, you may not want to bother with the expensive third-order+ phonon caculations!

Merge

tp.data.utilities.merge uses the tp metadata to combine multiple data dictionaries, so one can obtain denser data for memory- intensive calculations (such as AMSET) by running multiple times and merging the data dictionaries before plotting.