"""Tools for generating new random starting candidates."""
import numpy as np
from ase import Atoms
from ase.build import molecule
from ase.data import atomic_numbers
from ase.ga.utilities import (
atoms_too_close,
atoms_too_close_two_sets,
closest_distances_generator,
)
[docs]
class StartGenerator:
"""Class for generating random starting candidates.
Its basic task consists of randomly placing atoms or
molecules within a predescribed box, while respecting
certain minimal interatomic distances.
Depending on the problem at hand, certain box vectors
may not be known or chosen beforehand, and hence also
need to be generated at random. Common cases include
bulk crystals, films and chains, with respectively
3, 2 and 1 unknown cell vectors.
Parameters:
slab: Atoms object
Specifies the cell vectors and periodic boundary conditions
to be applied to the randomly generated structures.
Any included atoms (e.g. representing an underlying slab)
are copied to these new structures.
Variable cell vectors (see number_of_variable_cell_vectors)
will be ignored because these will be generated at random.
blocks: list
List of building units for the structure. Each item can be:
* an integer: representing a single atom by its atomic number,
* a string: for a single atom (a chemical symbol) or a
molecule (name recognized by ase.build.molecule),
* an Atoms object,
* an (A, B) tuple or list where A is any of the above
and B is the number of A units to include.
A few examples:
>>> blocks = ['Ti'] * 4 + ['O'] * 8
>>> blocks = [('Ti', 4), ('O', 8)]
>>> blocks = [('CO2', 3)] # 3 CO2 molecules
>>> co = Atoms('CO', positions=[[0, 0, 0], [1.4, 0, 0]])
>>> blocks = [(co, 3)]
Each individual block (single atom or molecule) in the
randomly generated candidates is given a unique integer
tag. These can be used to preserve the molecular identity
of these subunits.
blmin: dict or float
Dictionary with minimal interatomic distances.
If a number is provided instead, the dictionary will
be generated with this ratio of covalent bond radii.
Note: when preserving molecular identity (see use_tags),
the blmin dict will (naturally) only be applied
to intermolecular distances (not the intramolecular
ones).
number_of_variable_cell_vectors: int (default 0)
The number of variable cell vectors (0, 1, 2 or 3).
To keep things simple, it is the 'first' vectors which
will be treated as variable, i.e. the 'a' vector in the
univariate case, the 'a' and 'b' vectors in the bivariate
case, etc.
box_to_place_in: [list, list of lists] (default None)
The box in which the atoms can be placed.
The default (None) means the box is equal to the
entire unit cell of the 'slab' object.
In many cases, however, smaller boxes are desired
(e.g. for adsorbates on a slab surface or for isolated
clusters). Then, box_to_place_in can be set as
[p0, [v1, v2, v3]] with positions being generated as
p0 + r1 * v1 + r2 * v2 + r3 + v3.
In case of one or more variable cell vectors,
the corresponding items in p0/v1/v2/v3 will be ignored.
box_volume: int or float or None (default)
Initial guess for the box volume in cubic Angstrom
(used in generating the variable cell vectors).
Typical values in the solid state are 8-12 A^3 per atom.
If there are no variable cell vectors, the default None
is required (box volume equal to the box_to_place_in
volume).
splits: dict or None
Splitting scheme for increasing the translational symmetry
in the random candidates, based on:
* `Lyakhov, Oganov, Valle, Comp. Phys. Comm. 181 (2010) 1623-32`__
__ http://dx.doi.org/10.1016/j.cpc.2010.06.007
This should be a dict specifying the relative probabilities
for each split, written as tuples. For example,
>>> splits = {(2,): 3, (1,): 1}
This means that, for each structure, either a splitting
factor of 2 is applied to one randomly chosen axis,
or a splitting factor of 1 is applied (i.e., no splitting).
The probability ratio of the two scenararios will be 3:1,
i.e. 75% chance for the former and 25% chance for the latter
splitting scheme. Only the directions in which the 'slab'
object is periodic are eligible for splitting.
To e.g. always apply splitting factors of 2 and 3 along two
randomly chosen axes:
>>> splits = {(2, 3): 1}
By default, no splitting is applied (splits = None = {(1,): 1}).
cellbounds: ase.ga.utilities.CellBounds instance
Describing limits on the cell shape, see
:class:`~ase.ga.utilities.CellBounds`.
Note that it only make sense to impose conditions
regarding cell vectors which have been marked as
variable (see number_of_variable_cell_vectors).
test_dist_to_slab: bool (default True)
Whether to make sure that the distances between
the atoms and the slab satisfy the blmin.
test_too_far: bool (default True)
Whether to also make sure that there are no isolated
atoms or molecules with nearest-neighbour bond lengths
larger than 2x the value in the blmin dict.
rng: Random number generator
By default numpy.random.
"""
def __init__(self, slab, blocks, blmin, number_of_variable_cell_vectors=0,
box_to_place_in=None, box_volume=None, splits=None,
cellbounds=None, test_dist_to_slab=True, test_too_far=True,
rng=np.random):
self.slab = slab
self.blocks = []
for item in blocks:
if isinstance(item, (tuple, list)):
assert len(item) == 2, 'Item length %d != 2' % len(item)
block, count = item
else:
block, count = item, 1
# Convert block into Atoms object
if isinstance(block, Atoms):
pass
elif block in atomic_numbers:
block = Atoms(block)
elif isinstance(block, str):
block = molecule(block)
elif block in atomic_numbers.values():
block = Atoms(numbers=[block])
else:
raise ValueError('Cannot parse this block:', block)
# Add to self.blocks, taking into account that
# we want to group the same blocks together.
# This is important for the cell splitting.
for i, (b, c) in enumerate(self.blocks):
if block == b:
self.blocks[i][1] += count
break
else:
self.blocks.append([block, count])
if isinstance(blmin, dict):
self.blmin = blmin
else:
numbers = np.unique([b.get_atomic_numbers() for b in self.blocks])
self.blmin = closest_distances_generator(
numbers,
ratio_of_covalent_radii=blmin)
self.number_of_variable_cell_vectors = number_of_variable_cell_vectors
assert self.number_of_variable_cell_vectors in range(4)
if len(self.slab) > 0:
msg = 'Including atoms in the slab only makes sense'
msg += ' if there are no variable unit cell vectors'
assert self.number_of_variable_cell_vectors == 0, msg
for i in range(self.number_of_variable_cell_vectors):
msg = f'Unit cell {("abc"[i])}-vector is marked as variable '
msg += 'and slab must then also be periodic in this direction'
assert self.slab.pbc[i], msg
if box_to_place_in is None:
p0 = np.array([0., 0., 0.])
cell = self.slab.get_cell()
self.box_to_place_in = [p0, [cell[0, :], cell[1, :], cell[2, :]]]
else:
self.box_to_place_in = box_to_place_in
if box_volume is None:
assert self.number_of_variable_cell_vectors == 0
box_volume = abs(np.linalg.det(self.box_to_place_in[1]))
else:
assert self.number_of_variable_cell_vectors > 0
self.box_volume = box_volume
assert self.box_volume > 0
if splits is None:
splits = {(1,): 1}
tot = sum(v for v in splits.values())
self.splits = {k: v * 1. / tot for k, v in splits.items()}
self.cellbounds = cellbounds
self.test_too_far = test_too_far
self.test_dist_to_slab = test_dist_to_slab
self.rng = rng
def get_new_candidate(self, maxiter=None):
"""Returns a new candidate.
maxiter: upper bound on the total number of times
the random position generator is called
when generating the new candidate.
By default (maxiter=None) no such bound
is imposed. If the generator takes too
long time to create a new candidate, it
may be suitable to specify a finite value.
When the bound is exceeded, None is returned.
"""
pbc = self.slab.get_pbc()
# Choose cell splitting
r = self.rng.random()
cumprob = 0
for split, prob in self.splits.items():
cumprob += prob
if cumprob > r:
break
# Choose direction(s) along which to split
# and by how much
directions = [i for i in range(3) if pbc[i]]
repeat = [1, 1, 1]
if len(directions) > 0:
for number in split:
d = self.rng.choice(directions)
repeat[d] = number
repeat = tuple(repeat)
# Generate the 'full' unit cell
# for the eventual candidates
cell = self.generate_unit_cell(repeat)
if self.number_of_variable_cell_vectors == 0:
assert np.allclose(cell, self.slab.get_cell())
# Make the smaller 'box' in which we are
# allowed to place the atoms and which will
# then be repeated to fill the 'full' unit cell
box = np.copy(cell)
for i in range(self.number_of_variable_cell_vectors, 3):
box[i] = np.array(self.box_to_place_in[1][i])
box /= np.array([repeat]).T
# Here we gather the (reduced) number of blocks
# to put in the smaller box, and the 'surplus'
# occurring when the block count is not divisible
# by the number of repetitions.
# E.g. if we have a ('Ti', 4) block and do a
# [2, 3, 1] repetition, we employ a ('Ti', 1)
# block in the smaller box and delete 2 out 6
# Ti atoms afterwards
nrep = int(np.prod(repeat))
blocks, ids, surplus = [], [], []
for i, (block, count) in enumerate(self.blocks):
count_part = int(np.ceil(count * 1. / nrep))
blocks.extend([block] * count_part)
surplus.append(nrep * count_part - count)
ids.extend([i] * count_part)
N_blocks = len(blocks)
# Shuffle the ordering so different blocks
# are added in random order
order = np.arange(N_blocks)
self.rng.shuffle(order)
blocks = [blocks[i] for i in order]
ids = np.array(ids)[order]
# Add blocks one by one until we have found
# a valid candidate
blmin = self.blmin
blmin_too_far = {key: 2 * val for key, val in blmin.items()}
niter = 0
while maxiter is None or niter < maxiter:
cand = Atoms('', cell=box, pbc=pbc)
for i in range(N_blocks):
atoms = blocks[i].copy()
atoms.set_tags(i)
atoms.set_pbc(pbc)
atoms.set_cell(box, scale_atoms=False)
while maxiter is None or niter < maxiter:
niter += 1
cop = atoms.get_positions().mean(axis=0)
pos = np.dot(self.rng.random((1, 3)), box)
atoms.translate(pos - cop)
if len(atoms) > 1:
# Apply a random rotation to multi-atom blocks
phi, theta, psi = 360 * self.rng.random(3)
atoms.euler_rotate(phi=phi, theta=0.5 * theta, psi=psi,
center=pos)
if not atoms_too_close_two_sets(cand, atoms, blmin):
cand += atoms
break
else:
# Reached maximum iteration number
# Break out of the for loop above
cand = None
break
if cand is None:
# Exit the main while loop
break
# Rebuild the candidate after repeating,
# randomly deleting surplus blocks and
# sorting back to the original order
cand_full = cand.repeat(repeat)
tags_full = cand_full.get_tags()
for i in range(nrep):
tags_full[len(cand) * i:len(cand) * (i + 1)] += i * N_blocks
cand_full.set_tags(tags_full)
cand = Atoms('', cell=cell, pbc=pbc)
ids_full = np.tile(ids, nrep)
tag_counter = 0
if len(self.slab) > 0:
tag_counter = int(max(self.slab.get_tags())) + 1
for i, (block, count) in enumerate(self.blocks):
tags = np.where(ids_full == i)[0]
bad = self.rng.choice(tags, size=surplus[i], replace=False)
for tag in tags:
if tag not in bad:
select = [a.index for a in cand_full if a.tag == tag]
atoms = cand_full[select] # is indeed a copy!
atoms.set_tags(tag_counter)
assert len(atoms) == len(block)
cand += atoms
tag_counter += 1
for i in range(self.number_of_variable_cell_vectors, 3):
cand.positions[:, i] += self.box_to_place_in[0][i]
# By construction, the minimal interatomic distances
# within the structure should already be respected
assert not atoms_too_close(cand, blmin, use_tags=True), \
'This is not supposed to happen; please report this bug'
if self.test_dist_to_slab and len(self.slab) > 0:
if atoms_too_close_two_sets(self.slab, cand, blmin):
continue
if self.test_too_far:
tags = cand.get_tags()
for tag in np.unique(tags):
too_far = True
indices_i = np.where(tags == tag)[0]
indices_j = np.where(tags != tag)[0]
too_far = not atoms_too_close_two_sets(cand[indices_i],
cand[indices_j],
blmin_too_far)
if too_far and len(self.slab) > 0:
# the block is too far from the rest
# but might still be sufficiently
# close to the slab
too_far = not atoms_too_close_two_sets(cand[indices_i],
self.slab,
blmin_too_far)
if too_far:
break
else:
too_far = False
if too_far:
continue
# Passed all the tests
cand = self.slab + cand
cand.set_cell(cell, scale_atoms=False)
break
else:
# Reached max iteration count in the while loop
return None
return cand
def generate_unit_cell(self, repeat):
"""Generates a random unit cell.
For this, we use the vectors in self.slab.cell
in the fixed directions and randomly generate
the variable ones. For such a cell to be valid,
it has to satisfy the self.cellbounds constraints.
The cell will also be such that the volume of the
box in which the atoms can be placed (box limits
described by self.box_to_place_in) is equal to
self.box_volume.
Parameters:
repeat: tuple of 3 integers
Indicates by how much each cell vector
will later be reduced by cell splitting.
This is used to ensure that the original
cell is large enough so that the cell lengths
of the smaller cell exceed the largest
(X,X)-minimal-interatomic-distance in self.blmin.
"""
# Find the minimal cell length 'Lmin'
# that we need in order to ensure that
# an added atom or molecule will never
# be 'too close to itself'
Lmin = 0.
for atoms, count in self.blocks:
dist = atoms.get_all_distances(mic=False, vector=False)
num = atoms.get_atomic_numbers()
for i in range(len(atoms)):
dist[i, i] += self.blmin[(num[i], num[i])]
for j in range(i):
bl = self.blmin[(num[i], num[j])]
dist[i, j] += bl
dist[j, i] += bl
L = np.max(dist)
if L > Lmin:
Lmin = L
# Generate a suitable unit cell
valid = False
while not valid:
cell = np.zeros((3, 3))
for i in range(self.number_of_variable_cell_vectors):
# on-diagonal values
cell[i, i] = self.rng.random() * np.cbrt(self.box_volume)
cell[i, i] *= repeat[i]
for j in range(i):
# off-diagonal values
cell[i, j] = (self.rng.random() - 0.5) * cell[i - 1, i - 1]
# volume scaling
for i in range(self.number_of_variable_cell_vectors, 3):
cell[i] = self.box_to_place_in[1][i]
if self.number_of_variable_cell_vectors > 0:
volume = abs(np.linalg.det(cell))
scaling = self.box_volume / volume
scaling **= 1. / self.number_of_variable_cell_vectors
cell[:self.number_of_variable_cell_vectors] *= scaling
for i in range(self.number_of_variable_cell_vectors, 3):
cell[i] = self.slab.get_cell()[i]
# bounds checking
valid = True
if self.cellbounds is not None:
if not self.cellbounds.is_within_bounds(cell):
valid = False
if valid:
for i in range(3):
if np.linalg.norm(cell[i]) < repeat[i] * Lmin:
assert self.number_of_variable_cell_vectors > 0
valid = False
return cell