Source code for ase_ga.particle_mutations

# fmt: off

from operator import itemgetter

import numpy as np

from ase import Atoms
from ase_ga.offspring_creator import OffspringCreator
from ase_ga.utilities import get_distance_matrix, get_nndist


[docs] class Mutation(OffspringCreator): """Base class for all particle mutation type operators. Do not call this class directly.""" def __init__(self, num_muts=1, rng=np.random): OffspringCreator.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'Mutation' self.min_inputs = 1
[docs] @classmethod def get_atomic_configuration(cls, atoms, elements=None, eps=4e-2): """Returns the atomic configuration of the particle as a list of lists. Each list contain the indices of the atoms sitting at the same distance from the geometrical center of the particle. Highly symmetrical particles will often have many atoms in each shell. For further elaboration see: J. Montejano-Carrizales and J. Moran-Lopez, Geometrical characteristics of compact nanoclusters, Nanostruct. Mater., 1, 5, 397-409 (1992) Parameters: elements: Only take into account the elements specified in this list. Default is to take all elements into account. eps: The distance allowed to separate elements within each shell.""" atoms = atoms.copy() if elements is None: e = list(set(atoms.get_chemical_symbols())) else: e = elements atoms.set_constraint() atoms.center() geo_mid = np.array([(atoms.get_cell() / 2.)[i][i] for i in range(3)]) dists = [(np.linalg.norm(geo_mid - atoms[i].position), i) for i in range(len(atoms))] dists.sort(key=itemgetter(0)) atomic_conf = [] old_dist = -10. for dist, i in dists: if abs(dist - old_dist) > eps: atomic_conf.append([i]) else: atomic_conf[-1].append(i) old_dist = dist sorted_elems = sorted(set(atoms.get_chemical_symbols())) if e is not None and sorted(e) != sorted_elems: for shell in atomic_conf: torem = [] for i in shell: if atoms[i].symbol not in e: torem.append(i) for i in torem: shell.remove(i) return atomic_conf
[docs] @classmethod def get_list_of_possible_permutations(cls, atoms, l1, l2): """Returns a list of available permutations from the two lists of indices, l1 and l2. Checking that identical elements are not permuted.""" possible_permutations = [] for i in l1: for j in l2: if atoms[int(i)].symbol != atoms[int(j)].symbol: possible_permutations.append((i, j)) return possible_permutations
[docs] class RandomMutation(Mutation): """Moves a random atom the supplied length in a random direction.""" def __init__(self, length=2., num_muts=1, rng=np.random): Mutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'RandomMutation' self.length = length
[docs] def mutate(self, atoms): """ Does the actual mutation. """ tbm = self.rng.choice(range(len(atoms))) indi = Atoms() for a in atoms: if a.index == tbm: a.position += self.random_vector(self.length, rng=self.rng) indi.append(a) return indi
[docs] def get_new_individual(self, parents): f = parents[0] indi = self.initialize_individual(f) indi.info['data']['parents'] = [f.info['confid']] to_mut = f.copy() for _ in range(self.num_muts): to_mut = self.mutate(to_mut) for atom in to_mut: indi.append(atom) return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))
[docs] @classmethod def random_vector(cls, length, rng=np.random): """return random vector of certain length""" vec = np.array([rng.random() * 2 - 1 for _ in range(3)]) vl = np.linalg.norm(vec) return np.array([v * length / vl for v in vec])
[docs] class RandomPermutation(Mutation): """Permutes two random atoms. Parameters: num_muts: the number of times to perform this operation. rng: Random number generator By default numpy.random. """ def __init__(self, elements=None, num_muts=1, rng=np.random): Mutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'RandomPermutation' self.elements = elements
[docs] def get_new_individual(self, parents): f = parents[0].copy() diffatoms = len(set(f.numbers)) assert diffatoms > 1, 'Permutations with one atomic type is not valid' indi = self.initialize_individual(f) indi.info['data']['parents'] = [f.info['confid']] for _ in range(self.num_muts): RandomPermutation.mutate(f, self.elements, rng=self.rng) for atom in f: indi.append(atom) return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))
[docs] @classmethod def mutate(cls, atoms, elements=None, rng=np.random): """Do the actual permutation.""" if elements is None: indices = range(len(atoms)) else: indices = [a.index for a in atoms if a.symbol in elements] i1 = rng.choice(indices) i2 = rng.choice(indices) while atoms[i1].symbol == atoms[i2].symbol: i2 = rng.choice(indices) atoms.symbols[[i1, i2]] = atoms.symbols[[i2, i1]]
[docs] class COM2surfPermutation(Mutation): """The Center Of Mass to surface (COM2surf) permutation operator described in S. Lysgaard et al., Top. Catal., 2014, 57 (1-4), pp 33-39 Parameters: elements: which elements should be included in this permutation, for example: include all metals and exclude all adsorbates min_ratio: minimum ratio of each element in the core or surface region. If elements=[a, b] then ratio of a is Na / (Na + Nb) (N: Number of). If less than minimum ratio is present in the core, the region defining the core will be extended until the minimum ratio is met, and vice versa for the surface region. It has the potential reach the recursive limit if an element has a smaller total ratio in the complete particle. In that case remember to decrease this min_ratio. num_muts: the number of times to perform this operation. rng: Random number generator By default numpy.random. """ def __init__(self, elements=None, min_ratio=0.25, num_muts=1, rng=np.random): Mutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'COM2surfPermutation' self.min_ratio = min_ratio self.elements = elements
[docs] def get_new_individual(self, parents): f = parents[0].copy() diffatoms = len(set(f.numbers)) assert diffatoms > 1, 'Permutations with one atomic type is not valid' indi = self.initialize_individual(f) indi.info['data']['parents'] = [f.info['confid']] for _ in range(self.num_muts): elems = self.elements COM2surfPermutation.mutate(f, elems, self.min_ratio, rng=self.rng) for atom in f: indi.append(atom) return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))
[docs] @classmethod def mutate(cls, atoms, elements, min_ratio, rng=np.random): """Performs the COM2surf permutation.""" ac = atoms.copy() if elements is not None: del ac[[a.index for a in ac if a.symbol not in elements]] syms = ac.get_chemical_symbols() for el in set(syms): assert syms.count(el) / float(len(syms)) > min_ratio atomic_conf = Mutation.get_atomic_configuration(atoms, elements=elements) core = COM2surfPermutation.get_core_indices(atoms, atomic_conf, min_ratio) shell = COM2surfPermutation.get_shell_indices(atoms, atomic_conf, min_ratio) permuts = Mutation.get_list_of_possible_permutations(atoms, core, shell) chosen = rng.randint(len(permuts)) swap = list(permuts[chosen]) atoms.symbols[swap] = atoms.symbols[swap[::-1]]
[docs] @classmethod def get_core_indices(cls, atoms, atomic_conf, min_ratio, recurs=0): """Recursive function that returns the indices in the core subject to the min_ratio constraint. The indices are found from the supplied atomic configuration.""" elements = list({atoms[i].symbol for subl in atomic_conf for i in subl}) core = [i for subl in atomic_conf[:1 + recurs] for i in subl] while len(core) < 1: recurs += 1 core = [i for subl in atomic_conf[:1 + recurs] for i in subl] for elem in elements: ratio = len([i for i in core if atoms[i].symbol == elem]) / float(len(core)) if ratio < min_ratio: return COM2surfPermutation.get_core_indices(atoms, atomic_conf, min_ratio, recurs + 1) return core
[docs] @classmethod def get_shell_indices(cls, atoms, atomic_conf, min_ratio, recurs=0): """Recursive function that returns the indices in the surface subject to the min_ratio constraint. The indices are found from the supplied atomic configuration.""" elements = list({atoms[i].symbol for subl in atomic_conf for i in subl}) shell = [i for subl in atomic_conf[-1 - recurs:] for i in subl] while len(shell) < 1: recurs += 1 shell = [i for subl in atomic_conf[-1 - recurs:] for i in subl] for elem in elements: ratio = len([i for i in shell if atoms[i].symbol == elem]) / float(len(shell)) if ratio < min_ratio: return COM2surfPermutation.get_shell_indices(atoms, atomic_conf, min_ratio, recurs + 1) return shell
class _NeighborhoodPermutation(Mutation): """Helper class that holds common functions to all permutations that look at the neighborhoods of each atoms.""" @classmethod def get_possible_poor2rich_permutations(cls, atoms, inverse=False, recurs=0, distance_matrix=None): dm = distance_matrix if dm is None: dm = get_distance_matrix(atoms) # Adding a small value (0.2) to overcome slight variations # in the average bond length nndist = get_nndist(atoms, dm) + 0.2 same_neighbors = {} def f(x): return x[1] for i, atom in enumerate(atoms): same_neighbors[i] = 0 neighbors = [j for j in range(len(dm[i])) if dm[i][j] < nndist] for n in neighbors: if atoms[n].symbol == atom.symbol: same_neighbors[i] += 1 sorted_same = sorted(same_neighbors.items(), key=f) if inverse: sorted_same.reverse() poor_indices = [j[0] for j in sorted_same if abs(j[1] - sorted_same[0][1]) <= recurs] rich_indices = [j[0] for j in sorted_same if abs(j[1] - sorted_same[-1][1]) <= recurs] permuts = Mutation.get_list_of_possible_permutations(atoms, poor_indices, rich_indices) if len(permuts) == 0: _NP = _NeighborhoodPermutation return _NP.get_possible_poor2rich_permutations(atoms, inverse, recurs + 1, dm) return permuts
[docs] class Poor2richPermutation(_NeighborhoodPermutation): """The poor to rich (Poor2rich) permutation operator described in S. Lysgaard et al., Top. Catal., 2014, 57 (1-4), pp 33-39 Permutes two atoms from regions short of the same elements, to regions rich in the same elements. (Inverse of Rich2poorPermutation) Parameters: elements: Which elements to take into account in this permutation rng: Random number generator By default numpy.random. """ def __init__(self, elements=[], num_muts=1, rng=np.random): _NeighborhoodPermutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'Poor2richPermutation' self.elements = elements
[docs] def get_new_individual(self, parents): f = parents[0].copy() diffatoms = len(set(f.numbers)) assert diffatoms > 1, 'Permutations with one atomic type is not valid' indi = self.initialize_individual(f) indi.info['data']['parents'] = [f.info['confid']] for _ in range(self.num_muts): Poor2richPermutation.mutate(f, self.elements, rng=self.rng) for atom in f: indi.append(atom) return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))
@classmethod def mutate(cls, atoms, elements, rng=np.random): _NP = _NeighborhoodPermutation # indices = [a.index for a in atoms if a.symbol in elements] ac = atoms.copy() del ac[[atom.index for atom in ac if atom.symbol not in elements]] permuts = _NP.get_possible_poor2rich_permutations(ac) chosen = rng.randint(len(permuts)) swap = list(permuts[chosen]) atoms.symbols[swap] = atoms.symbols[swap[::-1]]
[docs] class Rich2poorPermutation(_NeighborhoodPermutation): """ The rich to poor (Rich2poor) permutation operator described in S. Lysgaard et al., Top. Catal., 2014, 57 (1-4), pp 33-39 Permutes two atoms from regions rich in the same elements, to regions short of the same elements. (Inverse of Poor2richPermutation) Parameters: elements: Which elements to take into account in this permutation rng: Random number generator By default numpy.random. """ def __init__(self, elements=None, num_muts=1, rng=np.random): _NeighborhoodPermutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'Rich2poorPermutation' self.elements = elements
[docs] def get_new_individual(self, parents): f = parents[0].copy() diffatoms = len(set(f.numbers)) assert diffatoms > 1, 'Permutations with one atomic type is not valid' indi = self.initialize_individual(f) indi.info['data']['parents'] = [f.info['confid']] if self.elements is None: elems = list(set(f.get_chemical_symbols())) else: elems = self.elements for _ in range(self.num_muts): Rich2poorPermutation.mutate(f, elems, rng=self.rng) for atom in f: indi.append(atom) return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))
@classmethod def mutate(cls, atoms, elements, rng=np.random): _NP = _NeighborhoodPermutation ac = atoms.copy() del ac[[atom.index for atom in ac if atom.symbol not in elements]] permuts = _NP.get_possible_poor2rich_permutations(ac, inverse=True) chosen = rng.randint(len(permuts)) swap = list(permuts[chosen]) atoms.symbols[swap] = atoms.symbols[swap[::-1]]
[docs] class SymmetricSubstitute(Mutation): """Permute all atoms within a subshell of the symmetric particle. The atoms within a subshell all have the same distance to the center, these are all equivalent under the particle point group symmetry. """ def __init__(self, elements=None, num_muts=1, rng=np.random): Mutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'SymmetricSubstitute' self.elements = elements
[docs] def substitute(self, atoms): """Does the actual substitution""" atoms = atoms.copy() aconf = self.get_atomic_configuration(atoms, elements=self.elements) itbm = self.rng.randint(0, len(aconf) - 1) to_element = self.rng.choice(self.elements) for i in aconf[itbm]: atoms[i].symbol = to_element return atoms
[docs] def get_new_individual(self, parents): f = parents[0] indi = self.substitute(f) indi = self.initialize_individual(f, indi) indi.info['data']['parents'] = [f.info['confid']] return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))
[docs] class RandomSubstitute(Mutation): """Substitutes one atom with another atom type. The possible atom types are supplied in the parameter elements""" def __init__(self, elements=None, num_muts=1, rng=np.random): Mutation.__init__(self, num_muts=num_muts, rng=rng) self.descriptor = 'RandomSubstitute' self.elements = elements
[docs] def substitute(self, atoms): """Does the actual substitution""" atoms = atoms.copy() if self.elements is None: elems = list(set(atoms.get_chemical_symbols())) else: elems = self.elements[:] possible_indices = [a.index for a in atoms if a.symbol in elems] itbm = self.rng.choice(possible_indices) elems.remove(atoms[itbm].symbol) new_symbol = self.rng.choice(elems) atoms[itbm].symbol = new_symbol return atoms
[docs] def get_new_individual(self, parents): f = parents[0] indi = self.substitute(f) indi = self.initialize_individual(f, indi) indi.info['data']['parents'] = [f.info['confid']] return (self.finalize_individual(indi), self.descriptor + ':Parent {}'.format(f.info['confid']))