def test_add_candidates():
import pytest
from ase.build import fcc111
from ase.ga.data import PrepareDB
from ase.ga.data import DataConnection
from ase.ga.offspring_creator import OffspringCreator
from ase.ga import set_raw_score
import os
db_file = 'gadb.db'
if os.path.isfile(db_file):
os.remove(db_file)
db = PrepareDB(db_file)
slab1 = fcc111('Ag', size=(2, 2, 2))
db.add_unrelaxed_candidate(slab1)
slab2 = fcc111('Cu', size=(2, 2, 2))
set_raw_score(slab2, 4)
db.add_relaxed_candidate(slab2)
assert slab2.info['confid'] == 3
db = DataConnection(db_file)
assert db.get_number_of_unrelaxed_candidates() == 1
slab3 = db.get_an_unrelaxed_candidate()
old_confid = slab3.info['confid']
slab3[0].symbol = 'Au'
db.add_unrelaxed_candidate(slab3, 'mutated: Parent {0}'.format(old_confid))
new_confid = slab3.info['confid']
# confid should update when using add_unrelaxed_candidate
assert old_confid != new_confid
slab3[1].symbol = 'Au'
db.add_unrelaxed_step(slab3, 'mutated: Parent {0}'.format(new_confid))
# confid should not change when using add_unrelaxed_step
assert slab3.info['confid'] == new_confid
with pytest.raises(AssertionError):
db.add_relaxed_step(slab3)
set_raw_score(slab3, 3)
db.add_relaxed_step(slab3)
slab4 = OffspringCreator.initialize_individual(slab1,
fcc111('Au', size=(2, 2, 2)))
set_raw_score(slab4, 67)
db.add_relaxed_candidate(slab4)
assert slab4.info['confid'] == 7
more_slabs = []
for m in ['Ni', 'Pd', 'Pt']:
slab = fcc111(m, size=(2, 2, 2))
slab = OffspringCreator.initialize_individual(slab1, slab)
set_raw_score(slab, sum(slab.get_masses()))
more_slabs.append(slab)
db.add_more_relaxed_candidates(more_slabs)
assert more_slabs[1].info['confid'] == 9
os.remove(db_file)
def prepare_ga(dbfile='godb.db', splits={(2,): 1}, N=20):
blocks = [('Pd', 4), ('OH', 8)] # the building blocks
volume = 50. * 4 # volume in angstrom^3
l = [list(Atoms(block).numbers)*count for block, count in blocks]
stoichiometry = [item for sublist in l for item in sublist]
atom_numbers = list(set(stoichiometry))
blmin = closest_distances_generator(atom_numbers=atom_numbers,
ratio_of_covalent_radii=0.6)
blmin[(1, 8)] = blmin[(8, 1)] = 2.0
cellbounds = CellBounds(bounds={'phi': [0.2 * 180., 0.8 * 180.],
'chi': [0.2 * 180., 0.8 * 180.],
'psi': [0.2 *180., 0.8 * 180.],
'a': [2, 8], 'b': [2, 8], 'c': [2, 8]})
# create the starting population
sg = StartGenerator(blocks, blmin, volume, cellbounds=cellbounds,
splits=splits)
# create the database to store information in
da = PrepareDB(db_file_name=dbfile, stoichiometry=stoichiometry)
for i in range(N):
a = sg.get_new_candidate()
a.set_initial_magnetic_moments(magmoms=None)
niggli_reduce(a)
da.add_unrelaxed_candidate(a)
return
def test_create_database():
from ase.ga.data import PrepareDB
from ase.ga.data import DataConnection
import os
import numpy as np
db_file = 'gadb.db'
if os.path.isfile(db_file):
os.remove(db_file)
from ase.build import fcc111
atom_numbers = np.array([78, 78, 79, 79])
slab = fcc111('Ag', size=(4, 4, 2), vacuum=10.)
PrepareDB(db_file_name=db_file,
simulation_cell=slab,
stoichiometry=atom_numbers)
assert os.path.isfile(db_file)
dc = DataConnection(db_file)
slab_get = dc.get_slab()
an_get = dc.get_atom_numbers_to_optimize()
assert len(slab) == len(slab_get)
assert np.all(slab.numbers == slab_get.numbers)
assert np.all(slab.get_positions() == slab_get.get_positions())
assert np.all(an_get == atom_numbers)
os.remove(db_file)
def prepare_ga(dbfile='godb.db', N=20):
''' This method creates a database with the desired number
of randomly generated structures.
'''
blocks = [('Si', 7)] # the building blocks
l = [list(Atoms(block).numbers) * count for block, count in blocks]
stoichiometry = [int(item) for sublist in l for item in sublist]
atom_numbers = list(set(stoichiometry))
# This dictionary will be used to check that the shortest
# Si-Si distances are above a certain threshold
# (here 1.5 Angstrom):
blmin = {(14, 14): 1.5}
# This defines the cubic simulation cell:
slab = Atoms('', positions=np.zeros((0, 3)), cell=[16] * 3)
# This defines the smaller box in which the
# initial coordinates are allowed to vary
density = 0.12 # in atoms per cubic Angstrom
aspect_ratios = np.array([1.0, 1.0, 1.0])
v = len(stoichiometry) / density
l = np.cbrt(v / np.product(aspect_ratios))
cell = np.identity(3) * aspect_ratios * l
p0 = 0.5 * (np.diag(slab.get_cell() - cell))
box = [p0, cell]
# Seed the random number generators using the system time,
# to ensure that no two runs produce the same results:
np.random.seed()
seed()
# Generate the random structures and add them to the database:
sg = StartGenerator(slab=slab, atom_numbers=stoichiometry,
closest_allowed_distances=blmin,
box_to_place_in=box)
da = PrepareDB(db_file_name=dbfile,
simulation_cell=slab,
stoichiometry=stoichiometry)
for i in range(N):
a = sg.get_new_candidate()
da.add_unrelaxed_candidate(a)
return
def prepare_ga(dbfile='godb.db', N=20):
"""
This method creates a database with the desired number
of randomly generated structures.
"""
Z = atomic_numbers['Si']
atom_numbers = [Z] * 7
# This dictionary will be used to check that the shortest
# Si-Si distances are above a certain threshold
# (here 1.5 Angstrom):
blmin = {(Z, Z): 1.5}
# This defines the cubic simulation cell. In case there
big_cell = np.identity(3) * 16.
slab = Atoms('', cell=big_cell)
# This defines the smaller box in which the initial
# coordinates are allowed to be generated
density = 0.1 # in atoms per cubic Angstrom
L = np.cbrt(len(atom_numbers) / density)
small_cell = np.identity(3) * L
p0 = 0.5 * np.diag(big_cell - small_cell)
box = [p0, small_cell]
# Seed the random number generators using the system time,
# to ensure that no two runs produce the same results:
np.random.seed()
seed()
# Generate the random structures and add them to the database:
sg = StartGenerator(slab=slab, atom_numbers=atom_numbers,
closest_allowed_distances=blmin,
box_to_place_in=box)
da = PrepareDB(db_file_name=dbfile,
simulation_cell=slab,
stoichiometry=atom_numbers)
for i in range(N):
a = sg.get_new_candidate()
da.add_unrelaxed_candidate(a)
return
def test_create_database(tmp_path):
db_file = tmp_path / 'gadb.db'
atom_numbers = np.array([78, 78, 79, 79])
slab = fcc111('Ag', size=(4, 4, 2), vacuum=10.)
PrepareDB(db_file_name=db_file,
simulation_cell=slab,
stoichiometry=atom_numbers)
assert os.path.isfile(db_file)
dc = DataConnection(db_file)
slab_get = dc.get_slab()
an_get = dc.get_atom_numbers_to_optimize()
assert len(slab) == len(slab_get)
assert np.all(slab.numbers == slab_get.numbers)
assert np.all(slab.get_positions() == slab_get.get_positions())
assert np.all(an_get == atom_numbers)
def test_database_logic(seed, testdir):
from ase.ga.data import PrepareDB
from ase.ga.data import DataConnection
from ase.ga.startgenerator import StartGenerator
from ase.ga.utilities import closest_distances_generator
from ase.ga import set_raw_score
import numpy as np
from ase.build import fcc111
from ase.constraints import FixAtoms
# set up the random number generator
rng = np.random.RandomState(seed)
slab = fcc111('Au', size=(4, 4, 2), vacuum=10.0, orthogonal=True)
slab.set_constraint(FixAtoms(mask=slab.positions[:, 2] <= 10.))
# define the volume in which the adsorbed cluster is optimized
# the volume is defined by a corner position (p0)
# and three spanning vectors (v1, v2, v3)
pos = slab.get_positions()
cell = slab.get_cell()
p0 = np.array([0., 0., max(pos[:, 2]) + 2.])
v1 = cell[0, :] * 0.8
v2 = cell[1, :] * 0.8
v3 = cell[2, :]
v3[2] = 3.
# define the closest distance between two atoms of a given species
blmin = closest_distances_generator(atom_numbers=[47, 79],
ratio_of_covalent_radii=0.7)
# Define the composition of the atoms to optimize
atom_numbers = 2 * [47] + 2 * [79]
# create the starting population
sg = StartGenerator(slab=slab,
blocks=atom_numbers,
blmin=blmin,
box_to_place_in=[p0, [v1, v2, v3]],
rng=rng)
# generate the starting population
starting_population = [sg.get_new_candidate() for i in range(20)]
d = PrepareDB(db_file_name=db_file,
simulation_cell=slab,
stoichiometry=atom_numbers)
for a in starting_population:
d.add_unrelaxed_candidate(a)
# and now for the actual test
dc = DataConnection(db_file)
dc.get_slab()
dc.get_atom_numbers_to_optimize()
assert dc.get_number_of_unrelaxed_candidates() == 20
a1 = dc.get_an_unrelaxed_candidate()
dc.mark_as_queued(a1)
assert dc.get_number_of_unrelaxed_candidates() == 19
assert len(dc.get_all_candidates_in_queue()) == 1
set_raw_score(a1, 0.0)
dc.add_relaxed_step(a1)
assert dc.get_number_of_unrelaxed_candidates() == 19
assert len(dc.get_all_candidates_in_queue()) == 0
assert len(dc.get_all_relaxed_candidates()) == 1
a2 = dc.get_an_unrelaxed_candidate()
dc.mark_as_queued(a2)
confid = a2.info['confid']
assert dc.get_all_candidates_in_queue()[0] == confid
dc.remove_from_queue(confid)
assert len(dc.get_all_candidates_in_queue()) == 0
# (and not intramolecularly):
Z = atomic_numbers['N']
blmin = closest_distances_generator(atom_numbers=[Z],
ratio_of_covalent_radii=1.3)
# The bounds for the randomly generated unit cells:
cellbounds = CellBounds(bounds={'phi': [30, 150], 'chi': [30, 150],
'psi': [30, 150], 'a': [3, 50],
'b': [3, 50], 'c': [3, 50]})
# The familiar 'slab' object, here only providing
# the PBC as there are no atoms or cell vectors
# that need to be applied.
slab = Atoms('', pbc=True)
# create the starting population
sg = StartGenerator(slab, blocks, blmin, box_volume=box_volume,
cellbounds=cellbounds, splits=splits,
number_of_variable_cell_vectors=3,
test_too_far=False)
# Initialize the database
da = PrepareDB(db_file_name='gadb.db', simulation_cell=slab,
stoichiometry=[Z] * 16)
# Generate the new structures
# and add them to the database
for i in range(N):
a = sg.get_new_candidate()
da.add_unrelaxed_candidate(a)
from ase.ga.data import PrepareDB
from ase.ga.data import DataConnection
import os
import numpy as np
db_file = 'gadb.db'
if os.path.isfile(db_file):
os.remove(db_file)
from ase.lattice.surface import fcc111
atom_numbers = np.array([78, 78, 79, 79])
slab = fcc111('Ag', size=(4, 4, 2), vacuum=10.)
d = PrepareDB(db_file_name=db_file,
simulation_cell=slab,
stoichiometry=atom_numbers)
assert os.path.isfile(db_file)
dc = DataConnection(db_file)
slab_get = dc.get_slab()
an_get = dc.get_atom_numbers_to_optimize()
assert len(slab) == len(slab_get)
assert np.all(slab.numbers == slab_get.numbers)
assert np.all(slab.get_positions() == slab_get.get_positions())
assert np.all(an_get == atom_numbers)
os.remove(db_file)
e = slab.get_potential_energy()
e_per_atom = e / len(slab)
refs[m] = e_per_atom
print('{0} = {1:.3f} eV/atom'.format(m, e_per_atom))
# The mixing energy for the pure slab is 0 by definition
set_raw_score(slab, 0.0)
pure_slabs.append(slab)
# The population size should be at least the number of different compositions
pop_size = 2 * len(slab)
# We prepare the db and write a few constants that we are going to use later
db = PrepareDB('hull.db',
population_size=pop_size,
reference_energies=refs,
metals=metals,
lattice_constants=lattice_constants)
# We add the pure slabs to the database as relaxed because we have already
# set the raw_score
for slab in pure_slabs:
db.add_relaxed_candidate(slab,
atoms_string=''.join(slab.get_chemical_symbols()))
# Now we create the rest of the candidates for the initial population
for i in range(pop_size - 2):
# How many of each metal is picked at random, making sure that
# we do not pick pure slabs
nA = random.randint(0, len(slab) - 2)
nB = len(slab) - 2 - nA
# controls the level of translational symmetry (within the unit
# cell) of the randomly generated structures. Here a 1:1 ratio
# of splitting factors 2 and 1 is used:
splits = {(2,): 1, (1,): 1}
# There will hence be a 50% probability that a candidate
# is constructed by repeating an randomly generated Ag12
# structure along a randomly chosen axis. In the other 50%
# of cases, no cell cell splitting will be applied.
# The 'slab' object in the GA serves as a template
# in the creation of new structures, which inherit
# the slab's atomic positions (if any), cell vectors
# (if specified), and periodic boundary conditions.
# Here only the last property is relevant:
slab = Atoms('', pbc=True)
# Initialize the random structure generator
sg = StartGenerator(slab, blocks, blmin, box_volume=volume,
number_of_variable_cell_vectors=3,
cellbounds=cellbounds, splits=splits)
# Create the database
da = PrepareDB(db_file_name='gadb.db',
stoichiometry=[Z] * 24)
# Generate N random structures
# and add them to the database
for i in range(N):
a = sg.get_new_candidate()
da.add_unrelaxed_candidate(a)
from ase.test import must_raise
from ase.build import fcc111
from ase.ga.data import PrepareDB
from ase.ga.data import DataConnection
from ase.ga.offspring_creator import OffspringCreator
from ase.ga import set_raw_score
import os
db_file = 'gadb.db'
if os.path.isfile(db_file):
os.remove(db_file)
db = PrepareDB(db_file)
slab1 = fcc111('Ag', size=(2, 2, 2))
db.add_unrelaxed_candidate(slab1)
slab2 = fcc111('Cu', size=(2, 2, 2))
set_raw_score(slab2, 4)
db.add_relaxed_candidate(slab2)
assert slab2.info['confid'] == 3
db = DataConnection(db_file)
assert db.get_number_of_unrelaxed_candidates() == 1
slab3 = db.get_an_unrelaxed_candidate()
old_confid = slab3.info['confid']
slab3[0].symbol = 'Au'
db.add_unrelaxed_candidate(slab3, 'mutated: Parent {0}'.format(old_confid))
new_confid = slab3.info['confid']
def test_basic_example_main_run(seed, testdir):
# set up the random number generator
rng = np.random.RandomState(seed)
# create the surface
slab = fcc111('Au', size=(4, 4, 1), vacuum=10.0, orthogonal=True)
slab.set_constraint(FixAtoms(mask=len(slab) * [True]))
# define the volume in which the adsorbed cluster is optimized
# the volume is defined by a corner position (p0)
# and three spanning vectors (v1, v2, v3)
pos = slab.get_positions()
cell = slab.get_cell()
p0 = np.array([0., 0., max(pos[:, 2]) + 2.])
v1 = cell[0, :] * 0.8
v2 = cell[1, :] * 0.8
v3 = cell[2, :]
v3[2] = 3.
# Define the composition of the atoms to optimize
atom_numbers = 2 * [47] + 2 * [79]
# define the closest distance two atoms of a given species can be to each other
unique_atom_types = get_all_atom_types(slab, atom_numbers)
blmin = closest_distances_generator(atom_numbers=unique_atom_types,
ratio_of_covalent_radii=0.7)
# create the starting population
sg = StartGenerator(slab=slab,
blocks=atom_numbers,
blmin=blmin,
box_to_place_in=[p0, [v1, v2, v3]],
rng=rng)
# generate the starting population
population_size = 5
starting_population = [sg.get_new_candidate() for i in range(population_size)]
# from ase.visualize import view # uncomment these lines
# view(starting_population) # to see the starting population
# create the database to store information in
d = PrepareDB(db_file_name=db_file,
simulation_cell=slab,
stoichiometry=atom_numbers)
for a in starting_population:
d.add_unrelaxed_candidate(a)
# XXXXXXXXXX This should be the beginning of a new test,
# but we are using some resources from the precious part.
# Maybe refactor those things as (module-level?) fixtures.
# Change the following three parameters to suit your needs
population_size = 5
mutation_probability = 0.3
n_to_test = 5
# Initialize the different components of the GA
da = DataConnection('gadb.db')
atom_numbers_to_optimize = da.get_atom_numbers_to_optimize()
n_to_optimize = len(atom_numbers_to_optimize)
slab = da.get_slab()
all_atom_types = get_all_atom_types(slab, atom_numbers_to_optimize)
blmin = closest_distances_generator(all_atom_types,
ratio_of_covalent_radii=0.7)
comp = InteratomicDistanceComparator(n_top=n_to_optimize,
pair_cor_cum_diff=0.015,
pair_cor_max=0.7,
dE=0.02,
mic=False)
pairing = CutAndSplicePairing(slab, n_to_optimize, blmin, rng=rng)
mutations = OperationSelector([1., 1., 1.],
[MirrorMutation(blmin, n_to_optimize, rng=rng),
RattleMutation(blmin, n_to_optimize, rng=rng),
PermutationMutation(n_to_optimize, rng=rng)],
rng=rng)
# Relax all unrelaxed structures (e.g. the starting population)
while da.get_number_of_unrelaxed_candidates() > 0:
a = da.get_an_unrelaxed_candidate()
a.calc = EMT()
print('Relaxing starting candidate {0}'.format(a.info['confid']))
dyn = BFGS(a, trajectory=None, logfile=None)
dyn.run(fmax=0.05, steps=100)
set_raw_score(a, -a.get_potential_energy())
da.add_relaxed_step(a)
# create the population
population = Population(data_connection=da,
population_size=population_size,
comparator=comp,
rng=rng)
# test n_to_test new candidates
for i in range(n_to_test):
print('Now starting configuration number {0}'.format(i))
a1, a2 = population.get_two_candidates()
a3, desc = pairing.get_new_individual([a1, a2])
if a3 is None:
continue
da.add_unrelaxed_candidate(a3, description=desc)
# Check if we want to do a mutation
if rng.rand() < mutation_probability:
a3_mut, desc = mutations.get_new_individual([a3])
if a3_mut is not None:
da.add_unrelaxed_step(a3_mut, desc)
a3 = a3_mut
# Relax the new candidate
a3.calc = EMT()
dyn = BFGS(a3, trajectory=None, logfile=None)
dyn.run(fmax=0.05, steps=100)
set_raw_score(a3, -a3.get_potential_energy())
da.add_relaxed_step(a3)
population.update()
write('all_candidates.traj', da.get_all_relaxed_candidates())
import random
from ase import Atoms
from ase.ga.data import PrepareDB
metals = ['Al', 'Au', 'Cu', 'Ag', 'Pd', 'Pt', 'Ni']
population_size = 10
# Create database
db = PrepareDB('fcc_alloys.db', population_size=population_size, metals=metals)
# Create starting population
for i in xrange(population_size):
atoms_string = [random.choice(metals) for _ in xrange(4)]
db.add_unrelaxed_candidate(Atoms(atoms_string),
atoms_string=''.join(atoms_string))