# coding: utf-8
# Distributed under the terms of the MIT License.
""" This submodule implements some useful functions for real/reciprocal
cell manipulation, symmetry checking and sampling (e.g. grids and paths.)
"""
import numpy as np
from periodictable import elements
EPS = 1e-12
[docs]def abc2cart(lattice_abc):
""" Converts lattice parameters into Cartesian lattice vectors.
Parameters:
lattice_abc (list): [[a, b, c], [alpha, beta, gamma]]
Returns:
list: Cartesian lattice vectors.
"""
assert len(lattice_abc) == 2
assert len(lattice_abc[0]) == 3
assert len(lattice_abc[1]) == 3
a = lattice_abc[0][0]
b = lattice_abc[0][1]
c = lattice_abc[0][2]
deg2rad = np.pi/180
alpha = lattice_abc[1][0] * deg2rad
beta = lattice_abc[1][1] * deg2rad
gamma = lattice_abc[1][2] * deg2rad
lattice_cart = []
lattice_cart.append([a, 0.0, 0.0])
# vec(b) = (b np.cos(gamma), b np.sin(gamma), 0)
bx = b*np.cos(gamma)
by = b*np.sin(gamma)
tol = 1e-12
if abs(bx) < tol:
bx = 0.0
if abs(by) < tol:
by = 0.0
cx = c*np.cos(beta)
if abs(cx) < tol:
cx = 0.0
cy = c*(np.cos(alpha)-np.cos(beta)*np.cos(gamma))/np.sin(gamma)
if abs(cy) < tol:
cy = 0.0
cz = np.sqrt(c**2 - cx**2 - cy**2)
lattice_cart.append([bx, by, 0.0])
lattice_cart.append([cx, cy, cz])
return lattice_cart
[docs]def cart2abcstar(lattice_cart):
""" Convert lattice_cart =[[a1,a2,a3],[b1,b2,b3],[c1,c2,c3]]
to the reciprocal of the lattice vectors, NOT the reciprocal lattice vectors.
Parameters:
lattice_cart (list): Cartesian lattice vectors.
Returns:
list: lattice parameters [[a,b,c],[alpha,beta,gamma]]
"""
return np.asarray(real2recip(lattice_cart)) / (2*np.pi)
[docs]def cart2volume(lattice_cart):
""" Convert lattice_cart to cell volume.
Parameters:
lattice_cart (list): Cartesian lattice vectors.
Returns:
float: cell volume in Angstrom^3.
"""
lattice_cart = np.asarray(lattice_cart)
vol = np.abs(np.dot(np.cross(lattice_cart[0], lattice_cart[1]), lattice_cart[2]))
return vol
[docs]def cart2abc(lattice_cart):
""" Convert Cartesian lattice vectors to lattice parametres.
Parameters:
lattice_cart (list): Cartesian lattice vectors.
Returns:
list: lattice parameters [[a,b,c],[alpha,beta,gamma]].
"""
vecs = lattice_cart
lattice_abc = []
a, b, c = (np.sqrt(sum([val**2 for val in vec])) for vec in vecs)
lattice_abc.append([a, b, c])
# np.cos(alpha) = b.c /|b * c|
radians2deg = 180.0/np.pi
cos_alpha = sum([val_b*val_c for (val_c, val_b) in zip(vecs[2], vecs[1])]) / (b*c)
cos_beta = sum([val_c*val_a for (val_c, val_a) in zip(vecs[2], vecs[0])]) / (c*a)
cos_gamma = sum([val_a*val_b for (val_a, val_b) in zip(vecs[0], vecs[1])]) / (a*b)
alpha = radians2deg * np.arccos(cos_alpha)
beta = radians2deg * np.arccos(cos_beta)
gamma = radians2deg * np.arccos(cos_gamma)
lattice_abc.append([alpha, beta, gamma])
return lattice_abc
[docs]def frac2cart(lattice_cart, positions_frac):
""" Convert positions_frac block into positions_abs.
Parameters:
lattice_cart (list): Cartesian lattice vectors.
positions_frac (list): list of fractional position vectors.
Returns:
list: list of absolute position vectors.
"""
_positions_frac = np.asarray(positions_frac)
reshaped = False
if len(np.shape(_positions_frac)) == 1:
reshaped = True
_positions_frac = _positions_frac.reshape((1, 3))
_lattice_cart = np.asarray(lattice_cart)
positions_abs = switch_coords(_lattice_cart, _positions_frac)
if reshaped:
positions_abs = positions_abs.reshape(-1)
return positions_abs.tolist()
[docs]def wrap_frac_coords(positions, remove=False):
""" Wrap the given fractional coordinates back into the cell.
Parameters:
positions (list): list of fractional position vectors, or
a single position.
Keyword arguments:
remove (bool): if True, removes points exterior to the cell.
Returns:
list: list of wrapped fractional position vectors.
"""
from copy import deepcopy
wrapped = deepcopy(positions)
single = False
if len(wrapped) == 3 and isinstance(wrapped[0], float):
wrapped = [wrapped]
single = True
if remove:
to_remove = len(wrapped)*[False]
for ind, pos in enumerate(wrapped):
for k in range(3):
if pos[k] >= 1 or pos[k] < 0:
if remove:
to_remove[ind] = True
break
else:
wrapped[ind][k] %= 1
if remove:
wrapped = [wrapped[ind] for ind, res in enumerate(to_remove) if not res]
if single:
return wrapped[0]
return wrapped
[docs]def switch_coords(lattice, pos, norm=None):
""" Act on coordinates with the relevant lattice
vectors to switch from fractional to absolute coordinates.
Parameters:
lattice (np.ndarray(3, 3)): either lattice_cart or reciprocal lattice_cart
pos (np.ndarray(3, :)): input positions to convert
Keyword arguments:
norm (float): divide final coordinates by normalisation factor, e.g.
2*np.pi when lattice is recip and positions are cartesian.
Returns:
np.ndarray(3, :): converted positions
"""
new_pos = np.zeros_like(pos)
for ni, _ in enumerate(pos):
for j in range(3):
new_pos[ni] += lattice[j] * pos[ni][j]
if norm is not None:
new_pos /= norm
return new_pos
[docs]def cart2frac(lattice_cart, positions_abs):
""" Convert positions_abs block into positions_frac (and equivalent
in reciprocal space).
Parameters:
lattice_cart (list): Cartesian lattice vectors.
positions_abs (list): list of absolute position vectors.
Returns:
list: list of fractional position vectors with the same shape
as the input list.
"""
_positions_abs = np.asarray(positions_abs, dtype=np.float64)
reshaped = False
if len(np.shape(_positions_abs)) == 1:
reshaped = True
_positions_abs = _positions_abs.reshape((1, 3))
recip_lat = np.asarray(real2recip(lattice_cart))
recip_lat = recip_lat.T
positions_frac = switch_coords(recip_lat, _positions_abs, norm=2*np.pi)
if reshaped:
positions_frac = positions_frac.reshape(-1)
return positions_frac.tolist()
[docs]def real2recip(real_lat):
""" Convert the real lattice in Cartesian basis to
the reciprocal space lattice.
Parameters:
real_lat (list): Cartesian lattice vectors.
Returns:
list: Cartesian lattice vectors of reciprocal lattice.
"""
real_lat = np.asarray(real_lat)
recip_lat = np.zeros((3, 3))
volume = np.dot(real_lat[0], np.cross(real_lat[1], real_lat[2]))
recip_lat[0] = (2*np.pi)*np.cross(real_lat[1], real_lat[2])
recip_lat[1] = (2*np.pi)*np.cross(real_lat[2], real_lat[0])
recip_lat[2] = (2*np.pi)*np.cross(real_lat[0], real_lat[1])
recip_lat /= volume
return recip_lat.tolist()
[docs]def calc_mp_grid(lattice_cart, spacing):
""" Return correct Monkhorst-Pack grid based on lattice
vectors and desired spacing.
Parameters:
lattice_cart (list): Cartesian lattice vectors.
spacing (float): desired maximum grid spacing.
Returns:
list: list of 3 integers defining the MP grid.
"""
recip_lat = real2recip(lattice_cart)
recip_len = np.zeros((3))
recip_len = np.sqrt(np.sum(np.power(recip_lat, 2), axis=1))
mp_grid = recip_len / (2 * np.pi * spacing)
return [int(np.ceil(elem)) for elem in mp_grid]
[docs]def shift_to_include_gamma(mp_grid):
""" Calculate the shift required to include $\\Gamma$.
in the Monkhorst-Pack grid.
Parameters:
mp_grid (:obj:`list` of :obj:`int`): number of grid points
in each reciprocal space direction.
Returns:
:obj:`list` of :obj:`float`: shift required to include $\\Gamma$.
"""
shift = [0, 0, 0]
for ind, val in enumerate(mp_grid):
if val % 2 == 0:
shift[ind] = 1.0/(val*2)
return shift
[docs]def shift_to_exclude_gamma(mp_grid):
""" Calculate the shift required to exclude $\\Gamma$.
in the Monkhorst-Pack grid. Returns the "minimal shift", i.e. only
one direction will be shifted.
Parameters:
mp_grid (:obj:`list` of :obj:`int`): number of grid points
in each reciprocal space direction.
Returns:
:obj:`list` of :obj:`float`: shift required to exclude $\\Gamma$.
"""
shift = [0, 0, 0]
if all([val % 2 == 1 for val in mp_grid]):
for ind, val in enumerate(mp_grid):
if val % 2 == 1:
shift[ind] = 1.0/(val*2)
break
return shift
[docs]def get_best_mp_offset_for_cell(doc):
""" Calculates the "best" kpoint_mp_offset to use for the passed
cell. If the crystal has a hexagonal space group, then the offset
returned will shift the grid to include $\\Gamma$ point, and vice
versa for non-hexagonal cells.
Parameters:
doc (dict): matador document to consider, containing structural
information and a "kpoints_mp_spacing" key.
Returns:
:obj:`list` of :obj:`float`: the desired kpoint_mp_offset.
"""
gamma = False
if '6' in get_spacegroup_spg(doc, symprec=1e-5):
gamma = True
print(doc)
if 'lattice_cart' not in doc or 'kpoints_mp_spacing' not in doc:
raise RuntimeError('Unable to calculate offset without lattice or spacing')
mp_grid = calc_mp_grid(doc['lattice_cart'], doc['kpoints_mp_spacing'])
if gamma:
return shift_to_include_gamma(mp_grid)
return shift_to_exclude_gamma(mp_grid)
[docs]def calc_mp_spacing(real_lat, mp_grid, prec=3):
""" Convert real lattice in Cartesian basis and the
kpoint_mp_grid into a grid spacing.
Parameters:
real_lat (list): Cartesian lattice vectors.
mp_grid (:obj:`list` of :obj:`int`): 3 integers defining the MP grid.
Keyword arguments:
prec (int): desired decimal precision of output.
Returns:
float: mp_spacing rounded to `prec`.
"""
recip_lat = real2recip(real_lat)
recip_len = np.zeros((3))
recip_len = np.sqrt(np.sum(np.power(recip_lat, 2), axis=1))
spacing = recip_len / (2*np.pi*np.asarray(mp_grid))
max_spacing = np.max(spacing)
exponent = round(np.log10(max_spacing) - prec)
return round(max_spacing + 0.5*10**exponent, prec)
[docs]def get_seekpath_kpoint_path(
doc, standardize=True, explicit=True, spacing=0.01, threshold=1e-7, debug=False, symmetry_tol=None
):
""" Return the conventional kpoint path of the relevant crystal system
according to the definitions by "HKPOT" in
Comp. Mat. Sci. 128, 2017:
http://dx.doi.org/10.1016/j.commatsci.2016.10.015
Parameters:
doc (dict/tuple): matador doc or spglib tuple to find kpoint path for.
Keyword arguments:
spacing (float): desired kpoint spacing
threshold (float): internal seekpath threshold
symmetry_tol (float): spglib symmetry tolerance
Returns:
dict: standardized version of input doc
list: list of kpoint positions
dict: full dictionary of all seekpath results
"""
try:
from seekpath import get_explicit_k_path, get_path
except ImportError:
raise ImportError("SeeK-Path dependency missing, please install it with `pip install seekpath`.")
if symmetry_tol is None:
symmetry_tol = 1e-5
if isinstance(doc, tuple):
spg_structure = doc
else:
if standardize:
spg_structure = doc2spg(standardize_doc_cell(doc, symprec=symmetry_tol))
else:
spg_structure = doc2spg(doc)
if explicit:
seekpath_results = get_explicit_k_path(spg_structure,
reference_distance=spacing,
with_time_reversal=True,
symprec=symmetry_tol,
threshold=threshold)
kpt_path = seekpath_results['explicit_kpoints_rel']
else:
seekpath_results = get_path(spg_structure)
kpt_path = []
primitive_doc = dict()
primitive_doc['lattice_cart'] = seekpath_results['primitive_lattice']
primitive_doc['positions_frac'] = seekpath_results['primitive_positions']
primitive_doc['atom_types'] = [str(elements[i]) for i in seekpath_results['primitive_types']]
primitive_doc['num_atoms'] = len(primitive_doc['atom_types'])
primitive_doc['lattice_abc'] = cart2abc(primitive_doc['lattice_cart'])
primitive_doc['cell_volume'] = cart2volume(primitive_doc['lattice_cart'])
if debug:
print('Found lattice type {}'.format(seekpath_results['bravais_lattice_extended']))
print('Old lattice:\n', np.asarray(doc['lattice_cart']))
print('Contained {} atoms'.format(doc['num_atoms']))
print('New lattice:\n', np.asarray(primitive_doc['lattice_cart']))
print('Contains {} atoms'.format(primitive_doc['num_atoms']))
print('k-point path contains {} points.'.format(len(kpt_path)))
if 'site_occupancy' in doc:
if min(doc['site_occupancy']) < 1 - EPS:
print('Ignoring any site occupancy found in this cell.')
primitive_doc['site_occupancy'] = [1 for atom in primitive_doc['atom_types']]
return primitive_doc, kpt_path, seekpath_results
[docs]def doc2spg(doc, check_occ=True):
""" Return an spglib input tuple from a matador doc.
Parameters:
doc (dict or :class:`Crystal`): matador document or Crystal object.
Keyword arguments:
check_occ (bool): check for partial occupancy and raise an error if
present.
Returns:
tuple: spglib-style tuple of lattice, positions and types.
"""
from matador.utils.chem_utils import get_atomic_number
try:
if 'lattice_cart' not in doc:
doc['lattice_cart'] = abc2cart(doc['lattice_abc'])
except KeyError:
raise RuntimeError("doc2spg failed: unable to find lattice in document.")
required_keys = ("lattice_cart", "positions_frac", "atom_types")
if all(key in doc for key in required_keys):
if 'lattice_cart' not in doc:
doc['lattice_cart'] = abc2cart(doc['lattice_abc'])
cell = (doc['lattice_cart'],
doc['positions_frac'],
[get_atomic_number(elem) for elem in doc['atom_types']])
if check_occ and np.min(doc.get('site_occupancy', [1.0])) < 1.0:
raise RuntimeError("spglib does not support partial occupancy.")
return cell
else:
raise RuntimeError(f"Unable to use doc2spg, one of {required_keys} was missing.")
[docs]def get_space_group_label_latex(label):
""" Return the LaTeX format of the passed space group label. Takes
any string, leaves the first character upright, italicses the rest,
handles subscripts and bars over numbers.
Parameters:
label (str): a given space group in "standard" plain text format,
e.g. P-63m.
Returns:
str: the best attempt to convert the label to LaTeX.
"""
latex_label = '$'
if not isinstance(label, str):
raise RuntimeError("Space group label must be a string, not {}".format(label))
skip = False
for ind, char in enumerate(label):
if skip:
skip = False
continue
if char == '-':
# add the next char inside the bar
latex_label += "\\bar{" + label[ind+1] + "}"
skip = True
else:
latex_label += char
latex_label += "$"
return latex_label
[docs]def standardize_doc_cell(doc, primitive=True, symprec=1e-2):
""" Return standardized cell data from matador doc.
Parameters:
doc (dict or :class:`Crystal`): matador document or Crystal object.
Keyword arguments:
primitive (bool): whether to reduce cell to primitive.
symprec (float): spglib symmetry tolerance.
Returns:
dict: matador document containing standardized cell.
"""
import spglib as spg
from matador.crystal import Crystal
from matador.utils.chem_utils import get_atomic_symbol
from copy import deepcopy
spg_cell = doc2spg(doc)
spg_standardized = spg.standardize_cell(spg_cell, to_primitive=primitive, symprec=symprec)
if not isinstance(doc, Crystal):
std_doc = deepcopy(doc)
else:
std_doc = deepcopy(doc._data)
std_doc['lattice_cart'] = [list(vec) for vec in spg_standardized[0]]
std_doc['lattice_abc'] = cart2abc(std_doc['lattice_cart'])
std_doc['positions_frac'] = [list(atom) for atom in spg_standardized[1]]
std_doc['atom_types'] = [get_atomic_symbol(atom) for atom in spg_standardized[2]]
std_doc['site_occupancy'] = len(std_doc['positions_frac']) * [1]
std_doc['cell_volume'] = cart2volume(std_doc['lattice_cart'])
std_doc['space_group'] = get_spacegroup_spg(std_doc, symprec=symprec)
# if the original document was a crystal, return a new one
if isinstance(doc, Crystal):
std_doc = Crystal(std_doc)
return std_doc
[docs]def get_spacegroup_spg(doc, symprec=0.01, check_occ=True):
""" Return spglib spacegroup for a cell.
Parameters:
doc (dict or :class:`Crystal`): matador document or Crystal object.
Keyword arguments:
symprec (float): spglib symmetry tolerance.
Returns:
str: spacegroup symbol of structure.
"""
import spglib as spg
spg_cell = doc2spg(doc, check_occ=check_occ)
space_group = spg.get_spacegroup(spg_cell, symprec=symprec)
if space_group is None:
raise RuntimeError('Spglib was unable to calculate space group.')
return space_group.split(' ')[0]
[docs]def add_noise(doc, amplitude=0.1):
""" Add random noise to the positions of structure contained in doc.
Useful for force convergence tests.
Parameters:
doc (dict): dictionary containing matador structure.
Keyword arguments:
amplitude (float): maximum amplitude of noise vector.
Raises:
KeyError if (`lattice_cart` and `positions_frac`) or `positions_abs`
are missing.
Returns:
dict: the randomised structure.
"""
poscart = np.asarray(doc.get('positions_abs') or frac2cart(doc['lattice_cart'], doc['positions_frac']))
for atom in poscart:
noise_vector = np.random.rand(3)
noise_vector /= np.sqrt(np.sum(noise_vector**2))
noise_vector *= amplitude*(np.random.rand())
atom += noise_vector
doc['positions_abs'] = poscart.tolist()
doc['positions_frac'] = cart2frac(doc['lattice_cart'], doc['positions_abs'])
return doc
[docs]def calc_pairwise_distances_pbc(poscart, images, lattice, rmax,
poscart_b=None, compress=False, debug=False,
filter_zero=False, per_image=False):
""" Calculate PBC distances with SciPy's cdist, given the
image cell vectors.
Parameters:
poscart (numpy.ndarray): list or array of absolute atomic coordinates.
images: iterable of lattice vector multiples (e.g. [2, -1, 3])
required to obtain the translation to desired image cells.
lattice (:obj:`list` if :obj:`list`): list of lattice vectors of
the real cell.
rmax (float): maximum value after which to mask the array.
Keyword arguments:
poscart_b (numpy.ndarray): absolute positions of another type of
atom, where only A-B distances will be calculated.
debug (bool): print timing data and how many distances were masked.
compress (bool): whether or not to compressed the output array,
useful when e.g. creating PDFs but not when atom ID is important.
filter_zero (bool): whether or not to filter out the "self-interaction"
zero distances.
per_image (bool): return a list of distances per image, as opposed to
one large flat. This preserves atom IDs for use elsewhere.
Returns:
distances (numpy.ndarray): pairwise 2-D d_ij masked array with values
or stripped 1-D array containing just the distances, or a list of
numpy arrays if per_image is True.
"""
from scipy.spatial.distance import cdist
import time
if debug:
start = time.time()
_lattice = np.asarray(lattice)
_poscart = np.asarray(poscart)
image_distances = []
if poscart_b is None:
_poscart_b = _poscart
else:
_poscart_b = np.asarray(poscart_b)
num_pairs = len(_poscart) * len(_poscart_b)
if per_image:
for image_ind, prod in enumerate(images):
distances = cdist(_poscart, _poscart_b + prod @ _lattice)
distances = np.ma.masked_where(distances > rmax, distances, copy=False)
image_distances.append(distances)
distances = image_distances
else:
array_size = (len(_poscart) * len(images) * len(_poscart_b))
distances = np.empty(array_size)
for image_ind, prod in enumerate(images):
distances[image_ind*num_pairs:(image_ind+1)*num_pairs] = cdist(_poscart, _poscart_b + prod @ _lattice).flatten()
distances = np.ma.masked_where(distances > rmax, distances, copy=False)
if filter_zero:
distances = np.ma.masked_where(distances < EPS, distances, copy=False)
if debug:
print('Calculated: {}, Used: {}, Ignored: {}'.format(len(distances),
np.ma.count(distances),
np.ma.count_masked(distances)))
if compress:
distances = distances.compressed()
if debug:
end = time.time()
print('Calculated distances in {} s'.format(end - start))
return distances
[docs]def create_simple_supercell(doc, extension, standardize=False, symmetric=False):
""" Return a document with new supercell, given extension vector.
Parameters:
doc (dict): matador doc to construct cell from.
extension (:obj:`tuple` of :obj:`int`): multiplicity of each lattice vector,
e.g. (2,2,1).
Keyword arguments:
standardize (bool): whether or not to use spglib to standardize the cell first.
symmetric (bool): whether or not centre the new cell on the origin.
Returns:
supercell_doc (dict): matador document containing supercell.
"""
from itertools import product
from copy import deepcopy
if not all([elem >= 1 for elem in extension]) or extension == (1, 1, 1):
raise RuntimeError('Bad/null supercell {} requested...'.format(extension))
supercell_doc = deepcopy(doc)
# standardize cell with spglib
if standardize:
supercell_doc = standardize_doc_cell(supercell_doc)
# copy new doc and delete data that will not be corrected
keys_to_copy = set(['positions_frac', 'lattice_cart', 'atom_types', 'source', 'stoichiometry'])
supercell_doc = {key: supercell_doc[key] for key in keys_to_copy}
for i, elem in enumerate(extension):
for k in range(3):
supercell_doc['lattice_cart'][i][k] = elem * supercell_doc['lattice_cart'][i][k]
for ind, atom in enumerate(supercell_doc['positions_frac']):
supercell_doc['positions_frac'][ind][i] /= elem
images = product(*[list(range(elem)) for elem in extension])
new_positions = []
new_atoms = []
for image in images:
if image != (0, 0, 0):
for ind, atom in enumerate(supercell_doc['atom_types']):
new_pos = deepcopy(supercell_doc['positions_frac'][ind])
for i, elem in enumerate(image):
new_pos[i] += image[i] / extension[i]
new_positions.append(new_pos)
new_atoms.append(atom)
supercell_doc['atom_types'].extend(new_atoms)
supercell_doc['positions_frac'].extend(new_positions)
if symmetric:
supercell_doc['positions_frac'] = np.asarray(supercell_doc['positions_frac'])
supercell_doc['positions_frac'] -= np.mean(supercell_doc['positions_frac'], axis=0)
supercell_doc['positions_frac'] = supercell_doc['positions_frac'].tolist()
supercell_doc['num_atoms'] = len(supercell_doc['atom_types'])
supercell_doc['cell_volume'] = cart2volume(supercell_doc['lattice_cart'])
supercell_doc['lattice_abc'] = cart2abc(supercell_doc['lattice_cart'])
assert np.isclose(supercell_doc['cell_volume'],
np.prod(extension)*cart2volume(doc['lattice_cart']))
return supercell_doc