# coding: utf-8
# Copyright (c) Max-Planck-Institut für Eisenforschung GmbH - Computational Materials Design (CM) Department
# Distributed under the terms of "New BSD License", see the LICENSE file.
import codecs
import pickle
import posixpath
from typing import Optional
import numpy as np
import scipy.constants
import structuretoolkit
from phonopy import Phonopy
from phonopy.file_IO import write_FORCE_CONSTANTS
from phonopy.units import VaspToTHz
from pyiron_base import JobGenerator, state
from pyiron_snippets.deprecate import deprecate
from pyiron_snippets.import_alarm import ImportAlarm
from pyiron_atomistics.atomistics.master.parallel import AtomisticParallelMaster
from pyiron_atomistics.atomistics.structure.atoms import ase_to_pyiron
from pyiron_atomistics.atomistics.structure.phonopy import (
publication as phonopy_publication,
)
__author__ = "Jan Janssen, Yury Lysogorskiy"
__copyright__ = (
"Copyright 2021, Max-Planck-Institut für Eisenforschung GmbH - "
"Computational Materials Design (CM) Department"
)
__version__ = "1.0"
__maintainer__ = "Jan Janssen"
__email__ = "janssen@mpie.de"
__status__ = "development"
__date__ = "Sep 1, 2017"
[docs]
class Thermal:
"""
Holds thermal properties that are the results of the phonopy calculation.
Args:
temperatures (ndarray): temperatures at which the other quantities are evaluated, units of Kelvin
free_energies (ndarray): free energies in the quasi-harmonic approximation from phonon contributions, units of
electron volts
entropy (ndarray): vibrational entropy calculated from the above free energy
cv (ndarray): heat capacity at constant volume, units of kJ/K/mol
"""
[docs]
def __init__(self, temperatures, free_energies, entropy, cv):
self.temperatures = temperatures
self.free_energies = free_energies
self.entropy = entropy
self.cv = cv
[docs]
class PhonopyJobGenerator(JobGenerator):
@property
def parameter_list(self):
"""
Returns:
(list)
"""
return [
[
"{}_{}".format(self._master.ref_job.job_name, ind),
self._restore_magmoms(
ase_to_pyiron(structuretoolkit.common.phonopy_to_atoms(sc))
),
]
for ind, sc in enumerate(self._master.phonopy.supercells_with_displacements)
]
def _restore_magmoms(self, structure):
"""
Args:
structure (pyiron.atomistics.structure.atoms): input structure
Returns:
structure (pyiron_atomistics.atomistics.structure.atoms): output structure with magnetic moments
"""
if self._master.structure.has("initial_magmoms"):
magmoms = self._master.structure.get_initial_magnetic_moments()
magmoms = np.tile(
magmoms,
np.prod(np.diagonal(self._master._phonopy_supercell_matrix())).astype(
int
),
)
structure.set_initial_magnetic_moments(magmoms)
return structure
[docs]
@staticmethod
def job_name(parameter):
return parameter[0]
[docs]
def modify_job(self, job, parameter):
job.structure = parameter[1]
return job
[docs]
class PhonopyJob(AtomisticParallelMaster):
"""
Phonopy wrapper for the calculation of free energy in the framework of quasi harmonic approximation.
Example:
>>> from pyiron_atomistics import Project
>>> pr = Project('my_project')
>>> lmp = pr.create_job('Lammps', 'lammps')
>>> lmp.structure = pr.create_structure('Fe', 'bcc', 2.832)
>>> phono = lmp.create_job('PhonopyJob', 'phonopy')
>>> phono.run()
Get output via `get_thermal_properties()`.
Note:
- This class does not consider the thermal expansion. For this, use `QuasiHarmonicJob` (find more in its
docstring)
- Depending on the value given in `job.input['interaction_range']`, this class automatically changes the number of
atoms. The input parameters of the reference job might have to be set appropriately (e.g. use `k_mesh_density`
for DFT instead of setting k-points directly).
- The structure used in the reference job should be a relaxed structure.
- Theory behind it: https://en.wikipedia.org/wiki/Quasi-harmonic_approximation
"""
[docs]
def __init__(self, project, job_name):
super(PhonopyJob, self).__init__(project, job_name)
self.__version__ = "0.0.1"
self.input["interaction_range"] = (10.0, "Minimal size of supercell, Ang")
self.input["displacement"] = (0.01, "atoms displacement, Ang")
self.input["dos_mesh"] = (20, "mesh size for DOS calculation")
self.input["primitive_matrix"] = None
self.input["number_of_snapshots"] = (
None,
"int or None, optional. Number of snapshots of supercells with random displacements. Random displacements "
"are generated displacing all atoms in random directions with a fixed displacement distance specified by "
"'distance' parameter, i.e., all atoms in supercell are displaced with the same displacement distance in "
"direct space. (Copied directly from phonopy docs. Requires the alm package to work.)",
)
self.input["eigenvectors"] = (
False,
"whether or not to save the eigenvectors of the dynamical matrix at each irreducible qpoint, "
"which are necessary to obtain the projected DOS (which can be done with a post-processing script). "
"They are accessed using job.phonopy.get_mesh_dict()['eigenvectors']. WARNING: "
"Setting this to True can result in a very large .h5 file (~GBs) or hit memory limits. "
"Only set to True if you know what you are doing!",
)
self.input["tetrahedron_method"] = (
True,
"use the tetrahedron method for the BZ integration. "
"If set to False, the Gaussian smearing method is used. "
"The 'sigma' value for the smearing is determined internally by phonopy. "
"See https://phonopy.github.io/phonopy/setting-tags.html#sigma .",
)
self.phonopy = None
self._job_generator = PhonopyJobGenerator(self)
self._disable_phonopy_pickle = False
state.publications.add(phonopy_publication())
@property
def phonopy_pickling_disabled(self):
return self._disable_phonopy_pickle
@phonopy_pickling_disabled.setter
def phonopy_pickling_disabled(self, disable):
self._disable_phonopy_pickle = disable
@property
def _phonopy_unit_cell(self):
if self.structure is not None:
return structuretoolkit.common.atoms_to_phonopy(self.structure)
else:
return None
def _enable_phonopy(self):
if self.phonopy is None:
if self.structure is not None:
self.phonopy = Phonopy(
unitcell=self._phonopy_unit_cell,
supercell_matrix=self._phonopy_supercell_matrix(),
primitive_matrix=self.input["primitive_matrix"],
)
self.phonopy.generate_displacements(
distance=self.input["displacement"],
number_of_snapshots=self.input["number_of_snapshots"],
)
self.to_hdf()
else:
raise ValueError("No reference job/ No reference structure found.")
def list_structures(self):
if self.structure is not None:
self._enable_phonopy()
return [struct for _, struct in self._job_generator.parameter_list]
else:
return []
def _phonopy_supercell_matrix(self):
if self.structure is not None:
supercell_range = np.ceil(
self.input["interaction_range"]
/ np.array(
[np.linalg.norm(vec) for vec in self._phonopy_unit_cell.cell]
)
)
return np.eye(3) * supercell_range
else:
return np.eye(3)
[docs]
def run_static(self):
# Initialise the phonopy object before starting the first calculation.
self._enable_phonopy()
super(PhonopyJob, self).run_static()
[docs]
def run_if_interactive(self):
self._enable_phonopy()
super(PhonopyJob, self).run_if_interactive()
[docs]
def to_hdf(self, hdf=None, group_name=None):
"""
Store the PhonopyJob in an HDF5 file
Args:
hdf (ProjectHDFio): HDF5 group object - optional
group_name (str): HDF5 subgroup name - optional
"""
super(PhonopyJob, self).to_hdf(hdf=hdf, group_name=group_name)
if self.phonopy is not None and not self._disable_phonopy_pickle:
with self.project_hdf5.open("output") as hdf5_output:
hdf5_output["phonopy_pickeled"] = codecs.encode(
pickle.dumps(self.phonopy), "base64"
).decode()
[docs]
def from_hdf(self, hdf=None, group_name=None):
"""
Restore the PhonopyJob from an HDF5 file
Args:
hdf (ProjectHDFio): HDF5 group object - optional
group_name (str): HDF5 subgroup name - optional
"""
super(PhonopyJob, self).from_hdf(hdf=hdf, group_name=group_name)
with self.project_hdf5.open("output") as hdf5_output:
if "phonopy_pickeled" in hdf5_output.list_nodes():
self.phonopy = pickle.loads(
codecs.decode(hdf5_output["phonopy_pickeled"].encode(), "base64")
)
if "dos_total" in hdf5_output.list_nodes():
self._dos_total = hdf5_output["dos_total"]
if "dos_energies" in hdf5_output.list_nodes():
self._dos_energies = hdf5_output["dos_energies"]
[docs]
def collect_output(self):
"""
Returns:
"""
if self.ref_job.server.run_mode.interactive:
forces_lst = self.project_hdf5.inspect(self.child_ids[0])[
"output/generic/forces"
]
else:
pr_job = self.project_hdf5.project.open(self.job_name + "_hdf5")
forces_lst = [
pr_job.inspect(job_name)["output/generic/forces"][-1]
for job_name in self._get_jobs_sorted()
]
self.phonopy.forces = forces_lst
self.phonopy.produce_force_constants(
fc_calculator=None if self.input["number_of_snapshots"] is None else "alm"
)
self.phonopy.run_mesh(
mesh=[self.input["dos_mesh"]] * 3,
with_eigenvectors=self.input["eigenvectors"],
)
mesh_dict = self.phonopy.get_mesh_dict()
self.phonopy.run_total_dos(
use_tetrahedron_method=self.input["tetrahedron_method"]
)
dos_dict = self.phonopy.get_total_dos_dict()
self.to_hdf()
with self.project_hdf5.open("output") as hdf5_out:
hdf5_out["dos_total"] = dos_dict["total_dos"]
hdf5_out["dos_energies"] = dos_dict["frequency_points"]
hdf5_out["qpoints"] = mesh_dict["qpoints"]
hdf5_out["supercell_matrix"] = self._phonopy_supercell_matrix()
hdf5_out["displacement_dataset"] = self.phonopy.dataset
hdf5_out["dynamical_matrix"] = (
self.phonopy.dynamical_matrix.dynamical_matrix
)
hdf5_out["force_constants"] = self.phonopy.force_constants
[docs]
def write_phonopy_force_constants(self, file_name="FORCE_CONSTANTS", cwd=None):
"""
Args:
file_name:
cwd:
Returns:
"""
if cwd is not None:
file_name = posixpath.join(cwd, file_name)
write_FORCE_CONSTANTS(
force_constants=self.phonopy.force_constants, filename=file_name
)
[docs]
def get_hesse_matrix(self):
"""
Returns:
"""
unit_conversion = (
scipy.constants.physical_constants["Hartree energy in eV"][0]
/ scipy.constants.physical_constants["Bohr radius"][0] ** 2
* scipy.constants.angstrom**2
)
force_shape = np.shape(self.phonopy.force_constants)
force_reshape = force_shape[0] * force_shape[2]
return (
np.transpose(self.phonopy.force_constants, (0, 2, 1, 3)).reshape(
(force_reshape, force_reshape)
)
/ unit_conversion
)
[docs]
def get_thermal_properties(self, t_min=1, t_max=1500, t_step=50, temperatures=None):
"""
Returns thermal properties at constant volume in the given temperature range. Can only be called after job
successfully ran.
Args:
t_min (float): minimum sample temperature
t_max (float): maximum sample temperature
t_step (int): tempeature sample interval
temperatures (array_like, float): custom array of temperature samples, if given t_min, t_max, t_step are
ignored.
Returns:
:class:`Thermal`: thermal properties as returned by Phonopy
"""
self.phonopy.run_thermal_properties(
t_step=t_step, t_max=t_max, t_min=t_min, temperatures=temperatures
)
tp_dict = self.phonopy.get_thermal_properties_dict()
kJ_mol_to_eV = 1000 / scipy.constants.Avogadro / scipy.constants.electron_volt
return Thermal(
tp_dict["temperatures"],
tp_dict["free_energy"] * kJ_mol_to_eV,
tp_dict["entropy"],
tp_dict["heat_capacity"],
)
@property
def dos_total(self):
"""
ndarray of float: Value of the DOS.
"""
return self["output/dos_total"]
@property
def dos_energies(self):
"""
ndarray of float: Energies at which the DOS is sampled.
"""
return self["output/dos_energies"]
@property
def dynamical_matrix(self):
"""
Returns:
"""
return np.real_if_close(
self.phonopy.get_dynamical_matrix().get_dynamical_matrix()
)
[docs]
def dynamical_matrix_at_q(self, q):
"""
Args:
q:
Returns:
"""
return np.real_if_close(self.phonopy.get_dynamical_matrix_at_q(q))
[docs]
@deprecate(ax="Use axis instead")
def plot_dos(self, ax=None, *args, axis=None, **kwargs):
"""
Plot the DOS.
If "label" is present in `kwargs` a legend is added to the plot automatically.
Args:
axis (optional): matplotlib axis to use, if None create a new one
ax: deprecated alias for axis
*args: passed to `axis.plot`
**kwargs: passed to `axis.plot`
Returns:
matplotlib.axes._subplots.AxesSubplot: axis with the plot
"""
try:
import pylab as plt
except ImportError:
import matplotlib.pyplot as plt
if ax is not None and axis is None:
axis = ax
if axis is None:
_, axis = plt.subplots(1, 1)
axis.plot(
self["output/dos_energies"], self["output/dos_total"], *args, **kwargs
)
axis.set_xlabel("Frequency [THz]")
axis.set_ylabel("DOS")
axis.set_title("Phonon DOS vs Energy")
if "label" in kwargs:
axis.legend()
return ax
[docs]
def get_band_structure(
self, npoints=101, with_eigenvectors=False, with_group_velocities=False
):
"""
Calculate band structure with automatic path through reciprocal space.
Can only be called after job is finished.
Args:
npoints (int, optional): Number of sample points between high symmetry points.
with_eigenvectors (boolean, optional): Calculate eigenvectors, too
with_group_velocities (boolean, optional): Calculate group velocities, too
Returns:
:class:`dict` of the results from phonopy under the following keys
- 'qpoints': list of (npoints, 3), samples paths in reciprocal space
- 'distances': list of (npoints,), distance along the paths in reciprocal space
- 'frequencies': list of (npoints, band), phonon frequencies
- 'eigenvectors': list of (npoints, band, band//3, 3), phonon eigenvectors
- 'group_velocities': list of (npoints, band), group velocities
where band is the number of bands (number of atoms * 3). Each entry is a list of arrays, and each array
corresponds to one path between two high-symmetry points automatically picked by Phonopy and may be of
different length than other paths. As compared to the phonopy output this method also reshapes the
eigenvectors so that they directly have the same shape as the underlying structure.
Raises:
:exception:`ValueError`: method is called on a job that is not finished or aborted
"""
if not self.status.finished:
raise ValueError(
"Job must be successfully run, before calling this method."
)
self.phonopy.auto_band_structure(
npoints,
with_eigenvectors=with_eigenvectors,
with_group_velocities=with_group_velocities,
)
results = self.phonopy.get_band_structure_dict()
if results["eigenvectors"] is not None:
# see https://phonopy.github.io/phonopy/phonopy-module.html#eigenvectors for the way phonopy stores the
# eigenvectors
results["eigenvectors"] = [
e.transpose(0, 2, 1).reshape(*e.shape[:2], -1, 3)
for e in results["eigenvectors"]
]
return results
[docs]
def plot_band_structure(
self, axis=None, *args, label: Optional[str] = None, **kwargs
):
"""
Plot bandstructure calculated with :meth:`.get_bandstructure`.
If :meth:`.get_bandstructure` hasn't been called before, it is automatically called with the default arguments.
If `label` is passed a legend is added automatically.
Args:
axis (matplotlib axis, optional): plot to this axis, if not given a new one is created.
*args: passed through to matplotlib.pyplot.plot when plotting the dispersion
label (str, optional): label for dispersion line
**kwargs: passed through to matplotlib.pyplot.plot when plotting the dispersion
Returns:
matplib axis: the axis the figure has been drawn to, if axis is given the same object is returned
"""
# label is it's own argument because if you try to pass it via **kwargs every line would get the label giving a
# messy legend
try:
import pylab as plt
except ImportError:
import matplotlib.pyplot as plt
if axis is None:
_, axis = plt.subplots(1, 1)
try:
results = self.phonopy.get_band_structure_dict()
except RuntimeError:
results = self.get_band_structure()
distances = results["distances"]
frequencies = results["frequencies"]
# HACK: strictly speaking this breaks phonopy API and could bite us
path_connections = self.phonopy._band_structure.path_connections
labels = self.phonopy._band_structure.labels
if "color" not in kwargs:
kwargs["color"] = "black"
offset = 0
tick_positions = [distances[0][0]]
for di, fi, ci in zip(distances, frequencies, path_connections):
axis.axvline(
tick_positions[-1], color="black", linestyle="dotted", alpha=0.5
)
line, *_ = axis.plot(offset + di, fi, *args, **kwargs)
tick_positions.append(di[-1] + offset)
if not ci:
offset += 0.05
axis.axvline(
tick_positions[-1], color="black", linestyle="dotted", alpha=0.5
)
tick_positions.append(di[-1] + offset)
if label is not None:
line.set_label(label)
axis.legend()
axis.set_xticks(tick_positions[:-1])
axis.set_xticklabels(labels)
axis.set_xlabel("Bandpath")
axis.set_ylabel("Frequency [THz]")
axis.set_title("Bandstructure")
return axis
[docs]
def validate_ready_to_run(self):
if self.ref_job._generic_input["calc_mode"] != "static":
raise ValueError(
"Phonopy reference jobs should be static calculations, but got {}".format(
self.ref_job._generic_input["calc_mode"]
)
)
if self.input["number_of_snapshots"] is not None:
with ImportAlarm(
"Phonopy with random snapshot displacement relies on the alm module but this unavailable. Please "
"ensure your python environment contains alm, e.g. by running `conda install -c conda-forge alm`."
"Note: at time of writing alm is only available for Linux and OSX systems."
) as import_alarm:
import alm
import_alarm.warn_if_failed()
super().validate_ready_to_run()
