Tutorial-04: Layered Plots

On top of the information in tutorial-03, this adds a plot of the phonon mean free path at each band and q-point in the simulation. The colour in this case corresponds to the band index, although a range of more informative variables can be projected onto the colour axis with add_projected_waterfall. This colouring can also be removed by specifying a single colour rather than a colourmap. This contains much the same information as the broadening in tutorial-02, with smaller values showing more scattering and lower lattice thermal conductivity.

As well as plotting graphs on neighbouring axes, it can be informative to plot them on the same axes to highlight links, for example how the mean free paths of phonons in BaSnO₃ dip where Ba has a strong contribution to the DoS, or how a phonon dispersion converges with supercell size. Although this can start to make scripts more complicated, and is not yet supported by the CLI (with the exception of phonon convergence), it is often worth the effort.

The python version of this code is:

#!/usr/bin/env python3

import tp

scs = '222 333 444 555'.split()
pfiles = ['../data/basno3/band-{}.yaml'.format(s) for s in scs]
kfile = '../data/basno3/kappa-m363636.hdf5'
dfile = '../data/basno3/projected_dos.dat'
poscar = '../data/basno3/POSCAR'

direction = 'avg'
temperature = 300
waterfall = 'mean_free_path'
quantities = ['waterfall', waterfall]

colour = 'winter_r'
colours = {'Ba': '#ffcf06',
           'Sn': '#59c605',
           'O':  '#00b1f7'}
cmap = 'viridis'

# You can ignore down to line 23!
from os import path
if not path.isfile(kfile) or (path.getsize(kfile) < 1024*1024*100):
    raise Exception('File not found, please use get-data.sh in the folder above.')
# Stop ignoring!

# Axes
fig, ax, add_legend = tp.axes.small.two_h()

# Load
dispersions = [tp.data.load.phonopy_dispersion(f) for f in pfiles]
kappa = tp.data.load.phono3py(kfile, quantities=quantities)
dos = tp.data.load.phonopy_dos(dfile, poscar=poscar)

# Plot
tp.plot.phonons.add_multi(ax[0], dispersions, colour=colour, label=scs)
tp.plot.frequency.format_waterfall(ax[1], kappa, waterfall, direction=direction,
                                   temperature=temperature, invert=True)
tp.plot.frequency.add_dos(ax[1], dos, colour=colours, scale=True, main=False,
                          alpha=0.6, line=False, invert=True)
tp.plot.frequency.add_waterfall(ax[1], kappa, waterfall, colour=cmap,
                                direction=direction, temperature=temperature,
                                invert=True)

# Formatting

tp.plot.utilities.set_locators(ax[1], x='log', y='linear')
axlabels = tp.settings.labels()
ax[1].set_xlabel(axlabels['mean_free_path'])
ax[1].set_ylabel(axlabels['frequency'])
add_legend()

# Save
fig.savefig('tutorial-04.png')

Axes (line 29)

Here we use two_h. h stands for horizontal, and there is a corresponding vertical set of axes, two_v. They also come with space for colourbars, by adding _colourbars to their names.

Load (line 32)

We need lists of files here, we’ve used list comprehensions to generate the file names and load them, but you could also use something like:

from glob import glob
pfiles = glob('band-*.yaml')

Although in that case, you would also need to define the labels.

Plot (lines 38-44)

add_dos and add_cumkappa have the arguments main and scale, which enable them to be used on the same axes as plots with other axis scales. main causes the axes ticks, labels and limits to be set for the plot, so turning it off doesn’t interfere with the current configuration. scale scales the data to the axes, rather than the axes to the data, so everything is visible. If both main and scale are set, the y-axis (or x, if invert) is scaled to percent.

It can look better to have the waterfall plot above the DoS, but the DoS should be scaled to the waterfall axes. Therefore, there is a function, format_waterfall, which sets the scale without plotting the waterfall plot. Running format_waterfall, then add_dos, then add_waterfall solves this, however it doesn’t work properly if line=True.

Formatting (lines 48-51)

Setting invert removes the y-axis labels and shortens the x-axis one in anticipation of being a DoS-style set of axes, but this can be reversed with some helper functions: set_locators sets the axis scales and tick locators for each axis, and also has a DoS argument, which removes the ticks and tick labels and the y axis label, while all the default labels can be accessed with settings.labels. These will be covered more in Tutorial-05.