def testForceAgainstDefaultNeighborList(self):
atoms = Diamond(symbol="C", pbc=False)
builtin = torchani.neurochem.Builtins()
calculator = torchani.ase.Calculator(builtin.species,
builtin.aev_computer,
builtin.models,
builtin.energy_shifter)
default_neighborlist_calculator = torchani.ase.Calculator(
builtin.species, builtin.aev_computer, builtin.models,
builtin.energy_shifter, True)
atoms.set_calculator(calculator)
dyn = Langevin(atoms, 5 * units.fs, 50 * units.kB, 0.002)
def test_energy(a=atoms):
a = a.copy()
a.set_calculator(calculator)
e1 = a.get_potential_energy()
a.set_calculator(default_neighborlist_calculator)
e2 = a.get_potential_energy()
self.assertLess(abs(e1 - e2), tol)
dyn.attach(test_energy, interval=1)
dyn.run(500)
def testEqualOneModelPTI(self):
calculator_pti = self.model_pti[0].ase()
calculator = self.model[0].ase()
atoms = Diamond(symbol="C", pbc=True)
atoms_pti = Diamond(symbol="C", pbc=True)
atoms.set_calculator(calculator)
atoms_pti.set_calculator(calculator_pti)
self.assertEqual(atoms.get_potential_energy(), atoms_pti.get_potential_energy())
def testWithNumericalForceWithPBCEnabled(self):
atoms = Diamond(symbol="C", pbc=True)
calculator = torchani.models.ANI1x().ase()
atoms.set_calculator(calculator)
dyn = Langevin(atoms, 5 * units.fs, 30000000 * units.kB, 0.002)
dyn.run(100)
f = torch.from_numpy(atoms.get_forces())
fn = get_numeric_force(atoms, 0.001)
df = (f - fn).abs().max()
avgf = f.abs().mean()
if avgf > 0:
self.assertLess(df / avgf, 0.1)
def testWithNumericalForceWithPBCEnabled(self):
# Run a Langevin thermostat dynamic for 100 steps and after the dynamic
# check once that the numerical and analytical force agree to a given
# relative tolerance
atoms = Diamond(symbol="C", pbc=True)
calculator = self.model.ase()
atoms.set_calculator(calculator)
dyn = Langevin(atoms, 5 * units.fs, 30000000 * units.kB, 0.002)
dyn.run(100)
f = atoms.get_forces()
fn = get_numeric_force(atoms, 0.001)
self.assertEqual(f, fn, rtol=0.1, atol=0.1)
def _testForce(self, pbc):
atoms = Diamond(symbol="C", pbc=pbc)
builtin = torchani.neurochem.Builtins()
calculator = torchani.ase.Calculator(
builtin.species, builtin.aev_computer,
builtin.models, builtin.energy_shifter)
atoms.set_calculator(calculator)
dyn = Langevin(atoms, 5 * units.fs, 30000000 * units.kB, 0.002)
dyn.run(100)
f = torch.from_numpy(atoms.get_forces())
fn = get_numeric_force(atoms, 0.001)
df = (f - fn).abs().max()
avgf = f.abs().mean()
if avgf > 0:
self.assertLess(df / avgf, 0.1)
def test_pbc(self):
a = Diamond('Si', latticeconstant=5.432, size=[2,2,2])
sx, sy, sz = a.get_cell().diagonal()
a.set_calculator(Tersoff())
e1 = a.get_potential_energy()
a.set_pbc([True,True,False])
e2 = a.get_potential_energy()
a.set_pbc(True)
a.set_cell([sx,sy,2*sz])
e3 = a.get_potential_energy()
self.assertEqual(e2, e3)
# This should give the unrelaxed surface energy
esurf = (e2-e1)/(2*sx*sy) * Jm2
self.assertTrue(abs(esurf-2.309) < 0.001)
def test_pbc(self):
a = Diamond('Si', latticeconstant=5.432, size=[2,2,2])
sx, sy, sz = a.get_cell().diagonal()
a.set_calculator(Tersoff())
e1 = a.get_potential_energy()
a.set_pbc([True,True,False])
e2 = a.get_potential_energy()
a.set_pbc(True)
a.set_cell([sx,sy,2*sz])
e3 = a.get_potential_energy()
self.assertEqual(e2, e3)
# This should give the unrelaxed surface energy
esurf = (e2-e1)/(2*sx*sy) * Jm2
self.assertTrue(abs(esurf-2.309) < 0.001)
def test_calculator():
"""
Take ASE structure, PySCF object,
and run through ASE calculator interface.
This allows other ASE methods to be used with PySCF;
here we try to compute an equation of state.
"""
ase_atom=Diamond(symbol='C', latticeconstant=3.5668)
# Set up a cell; everything except atom; the ASE calculator will
# set the atom variable
cell = pbcgto.Cell()
cell.h=ase_atom.cell
cell.basis = 'gth-szv'
cell.pseudo = 'gth-pade'
cell.gs=np.array([8,8,8])
cell.verbose = 0
# Set up the kind of calculation to be done
# Additional variables for mf_class are passed through mf_dict
mf_class=pbcdft.RKS
mf_dict = { 'xc' : 'lda,vwn' }
# Once this is setup, ASE is used for everything from this point on
ase_atom.set_calculator(pyscf_ase.PySCF(molcell=cell, mf_class=mf_class, mf_dict=mf_dict))
print "ASE energy", ase_atom.get_potential_energy()
print "ASE energy (should avoid re-evaluation)", ase_atom.get_potential_energy()
# Compute equation of state
ase_cell=ase_atom.cell
volumes = []
energies = []
for x in np.linspace(0.95, 1.2, 5):
ase_atom.set_cell(ase_cell * x, scale_atoms = True)
print "[x: %f, E: %f]" % (x, ase_atom.get_potential_energy())
volumes.append(ase_atom.get_volume())
energies.append(ase_atom.get_potential_energy())
eos = EquationOfState(volumes, energies)
v0, e0, B = eos.fit()
print(B / kJ * 1.0e24, 'GPa')
eos.plot('eos.png')
def MD():
old_stdout = sys.stdout
sys.stdout = mystdout = StringIO()
orig_stdout = sys.stdout
f = open('out.txt', 'w')
sys.stdout = f
#f = open('out_' + str(temp) + '.txt', 'w')
#sys.stdout = f
Tmin = 1000
Tmax = 15000
a0 = 5.43
N = 1
atoms = Diamond(symbol='Si', latticeconstant=a0)
atoms *= (N, N, N)
atoms.set_calculator(model.calculator)
MaxwellBoltzmannDistribution(atoms, Tmin * units.kB) ###is this allowed?
Stationary(atoms) # zero linear momentum
ZeroRotation(atoms)
def printenergy(a=atoms):
epot = a.get_potential_energy() / len(a)
ekin = a.get_kinetic_energy() / len(a)
print(ekin)
traj = Trajectory('Si.traj', 'w', atoms)
for temp in np.linspace(Tmin, Tmax, 5):
dyn = Langevin(atoms, 5 * units.fs, units.kB * temp, 0.002)
dyn.attach(traj.write, interval=1)
dyn.attach(printenergy, interval=1)
dyn.run(100)
sys.stdout = orig_stdout
f.close()
print_crack_system(directions)
# now, we build system aligned with requested crystallographic orientation
unit_slab = Diamond(directions=directions,
size=(1, 1, 1),
symbol='Si',
pbc=True,
latticeconstant=a0)
if (hasattr(params, 'check_rotated_elastic_constants') and
# Check that elastic constants of the rotated system
# lead to same Young's modulus and Poisson ratio
params.check_rotated_elastic_constants):
unit_slab.set_calculator(params.calc)
C_r1 = measure_triclinic_elastic_constants(unit_slab,
optimizer=FIRE,
fmax=params.bulk_fmax)
R = np.array([ np.array(x)/np.linalg.norm(x) for x in directions ])
C_r2 = rotate_elastic_constants(C, R)
for C_r in [C_r1, C_r2]:
S_r = np.linalg.inv(C_r)
E_r = 1./S_r[1,1]
print('Young\'s modulus from C_r: %.1f GPa' % (E_r / units.GPa))
assert (abs(E_r - E)/units.GPa < 1e-3)
nu_r = -S_r[1,2]/S_r[1,1]
print('Possion ratio from C_r: %.3f' % (nu_r))
params.crack_direction, params.cleavage_plane, params.crack_front
]
print_crack_system(directions)
# now, we build system aligned with requested crystallographic orientation
unit_slab = Diamond(directions=directions,
size=(1, 1, 1),
symbol='Si',
pbc=True,
latticeconstant=a0)
if (hasattr(params, 'check_rotated_elastic_constants') and
# Check that elastic constants of the rotated system
# lead to same Young's modulus and Poisson ratio
params.check_rotated_elastic_constants):
unit_slab.set_calculator(params.calc)
C_r1 = measure_triclinic_elastic_constants(unit_slab,
optimizer=FIRE,
fmax=params.bulk_fmax)
R = np.array([np.array(x) / np.linalg.norm(x) for x in directions])
C_r2 = rotate_elastic_constants(C, R)
for C_r in [C_r1, C_r2]:
S_r = np.linalg.inv(C_r)
E_r = 1. / S_r[1, 1]
print('Young\'s modulus from C_r: %.1f GPa' % (E_r / units.GPa))
assert (abs(E_r - E) / units.GPa < 1e-3)
nu_r = -S_r[1, 2] / S_r[1, 1]
print('Possion ratio from C_r: %.3f' % (nu_r))
cell.a = ase_atom.cell
cell.basis = 'gth-szv'
cell.pseudo = 'gth-pade'
cell.verbose = 0
# Set up the kind of calculation to be done
# Additional variables for mf_class are passed through mf_dict
# E.g. gamma-point SCF calculation can be set to
mf_class = pbcdft.RKS
# SCF with k-point sampling can be set to
mf_class = lambda cell: pbcdft.KRKS(cell, kpts=cell.make_kpts([2, 2, 2]))
mf_dict = {'xc': 'lda,vwn'}
# Once this is setup, ASE is used for everything from this point on
ase_atom.set_calculator(
pyscf_ase.PySCF(molcell=cell, mf_class=mf_class, mf_dict=mf_dict))
print("ASE energy", ase_atom.get_potential_energy())
print("ASE energy (should avoid re-evaluation)",
ase_atom.get_potential_energy())
# Compute equation of state
ase_cell = ase_atom.cell
volumes = []
energies = []
for x in np.linspace(0.95, 1.2, 5):
ase_atom.set_cell(ase_cell * x, scale_atoms=True)
print "[x: %f, E: %f]" % (x, ase_atom.get_potential_energy())
volumes.append(ase_atom.get_volume())
energies.append(ase_atom.get_potential_energy())
eos = EquationOfState(volumes, energies)
class TestFitElasticConstants(matscipytest.MatSciPyTestCase):
"""
Tests of elastic constant calculation.
We test with a cubic Silicon lattice, since this also has most of the
symmetries of the lower-symmetry crystal families
"""
def setUp(self):
self.pot = quippy.Potential('IP SW',
param_str="""
<SW_params n_types="2" label="PRB_31_plus_H">
<comment> Stillinger and Weber, Phys. Rev. B 31 p 5262 (1984), extended for other elements </comment>
<per_type_data type="1" atomic_num="1" />
<per_type_data type="2" atomic_num="14" />
<per_pair_data atnum_i="1" atnum_j="1" AA="0.0" BB="0.0"
p="0" q="0" a="1.0" sigma="1.0" eps="0.0" />
<per_pair_data atnum_i="1" atnum_j="14" AA="8.581214" BB="0.0327827"
p="4" q="0" a="1.25" sigma="2.537884" eps="2.1672" />
<per_pair_data atnum_i="14" atnum_j="14" AA="7.049556277" BB="0.6022245584"
p="4" q="0" a="1.80" sigma="2.0951" eps="2.1675" />
<!-- triplet terms: atnum_c is the center atom, neighbours j and k -->
<per_triplet_data atnum_c="1" atnum_j="1" atnum_k="1"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="1" atnum_j="1" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="1" atnum_j="14" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="14" atnum_j="1" atnum_k="1"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="14" atnum_j="1" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="14" atnum_j="14" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
</SW_params>
""")
self.fmax = 1e-4
self.at0 = Diamond('Si', latticeconstant=5.43)
self.at0.set_calculator(self.pot)
# relax initial positions and unit cell
FIRE(StrainFilter(self.at0, mask=[1, 1, 1, 0, 0, 0]),
logfile=None).run(fmax=self.fmax)
self.C_ref = np.array(
[[151.4276439, 76.57244456, 76.57244456, 0., 0., 0.],
[76.57244456, 151.4276439, 76.57244456, 0., 0., 0.],
[76.57244456, 76.57244456, 151.4276439, 0., 0., 0.],
[0., 0., 0., 109.85498798, 0., 0.],
[0., 0., 0., 0., 109.85498798, 0.],
[0., 0., 0., 0., 0., 109.85498798]])
self.C_err_ref = np.array(
[[1.73091718, 1.63682097, 1.63682097, 0., 0., 0.],
[1.63682097, 1.73091718, 1.63682097, 0., 0., 0.],
[1.63682097, 1.63682097, 1.73091718, 0., 0., 0.],
[0., 0., 0., 1.65751232, 0., 0.],
[0., 0., 0., 0., 1.65751232, 0.],
[0., 0., 0., 0., 0., 1.65751232]])
self.C_ref_relaxed = np.array(
[[151.28712587, 76.5394162, 76.5394162, 0., 0., 0.],
[76.5394162, 151.28712587, 76.5394162, 0., 0., 0.],
[76.5394162, 76.5394162, 151.28712587, 0., 0., 0.],
[0., 0., 0., 56.32421772, 0., 0.],
[0., 0., 0., 0., 56.32421772, 0.],
[0., 0., 0., 0., 0., 56.32421772]])
self.C_err_ref_relaxed = np.array(
[[1.17748661, 1.33333615, 1.33333615, 0., 0., 0.],
[1.33333615, 1.17748661, 1.33333615, 0., 0., 0.],
[1.33333615, 1.33333615, 1.17748661, 0., 0., 0.],
[0., 0., 0., 0.18959684, 0., 0.],
[0., 0., 0., 0., 0.18959684, 0.],
[0., 0., 0., 0., 0., 0.18959684]])
def test_measure_triclinic_unrelaxed(self):
# compare to brute force method without relaxation
C = measure_triclinic_elastic_constants(self.at0,
delta=1e-2,
optimizer=None)
self.assertArrayAlmostEqual(C / units.GPa, self.C_ref, tol=0.2)
def test_measure_triclinic_relaxed(self):
# compare to brute force method with relaxation
C = measure_triclinic_elastic_constants(self.at0,
delta=1e-2,
optimizer=FIRE,
fmax=self.fmax)
self.assertArrayAlmostEqual(C / units.GPa,
self.C_ref_relaxed,
tol=0.2)
def testcubic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'cubic',
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa, self.C_ref, tol=1e-2)
if abs(C_err).max() > 0.:
self.assertArrayAlmostEqual(C_err / units.GPa,
self.C_err_ref,
tol=1e-2)
def testcubic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'cubic',
optimizer=FIRE,
fmax=self.fmax,
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa,
self.C_ref_relaxed,
tol=1e-2)
if abs(C_err).max() > 0.:
self.assertArrayAlmostEqual(C_err / units.GPa,
self.C_err_ref_relaxed,
tol=1e-2)
def testorthorhombic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'orthorhombic',
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa, self.C_ref, tol=0.1)
def testorthorhombic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'orthorhombic',
optimizer=FIRE,
fmax=self.fmax,
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa,
self.C_ref_relaxed,
tol=0.1)
def testmonoclinic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'monoclinic',
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa, self.C_ref, tol=0.2)
def testmonoclinic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'monoclinic',
optimizer=FIRE,
fmax=self.fmax,
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa,
self.C_ref_relaxed,
tol=0.2)
def testtriclinic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'triclinic',
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa, self.C_ref, tol=0.2)
def testtriclinic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0,
'triclinic',
optimizer=FIRE,
fmax=self.fmax,
verbose=False,
graphics=False)
self.assertArrayAlmostEqual(C / units.GPa,
self.C_ref_relaxed,
tol=0.2)
# Now let's create a calculator from builtin models:
calculator = my_torchani.models.ANI1ccx().ase()
# .. note::
# Regardless of the dtype you use in your model, when converting it to ASE
# calculator, it always automatically the dtype to ``torch.float64``. The
# reason for this behavior is, at many cases, the rounding error is too
# large for structure minimization. If you insist on using
# ``torch.float32``, do the following instead:
#
# .. code-block:: python
#
# calculator = torchani.models.ANI1ccx().ase(dtype=torch.float32)
# Now let's set the calculator for ``atoms``:
atoms.set_calculator(calculator)
# Now let's minimize the structure:
print("Begin minimizing...")
opt = BFGS(atoms)
opt.run(fmax=0.001)
print()
# Now create a callback function that print interesting physical quantities:
def printenergy(a=atoms):
"""Function to print the potential, kinetic and total energy."""
epot = a.get_potential_energy() / len(a)
ekin = a.get_kinetic_energy() / len(a)
print('Energy per atom: Epot = %.3feV Ekin = %.3feV (T=%3.0fK) '
'Etot = %.3feV' % (epot, ekin, ekin / (1.5 * units.kB), epot + ekin))
cell.a=ase_atom.cell
cell.basis = 'gth-szv'
cell.pseudo = 'gth-pade'
cell.verbose = 0
# Set up the kind of calculation to be done
# Additional variables for mf_class are passed through mf_dict
# E.g. gamma-point SCF calculation can be set to
mf_class = pbcdft.RKS
# SCF with k-point sampling can be set to
mf_class = lambda cell: pbcdft.KRKS(cell, kpts=cell.make_kpts([2,2,2]))
mf_dict = { 'xc' : 'lda,vwn' }
# Once this is setup, ASE is used for everything from this point on
ase_atom.set_calculator(pyscf_ase.PySCF(molcell=cell, mf_class=mf_class, mf_dict=mf_dict))
print("ASE energy", ase_atom.get_potential_energy())
print("ASE energy (should avoid re-evaluation)", ase_atom.get_potential_energy())
# Compute equation of state
ase_cell=ase_atom.cell
volumes = []
energies = []
for x in np.linspace(0.95, 1.2, 5):
ase_atom.set_cell(ase_cell * x, scale_atoms = True)
print "[x: %f, E: %f]" % (x, ase_atom.get_potential_energy())
volumes.append(ase_atom.get_volume())
energies.append(ase_atom.get_potential_energy())
eos = EquationOfState(volumes, energies)
v0, e0, B = eos.fit()
from ase.lattice.cubic import Diamond
from ase.md.velocitydistribution import MaxwellBoltzmannDistribution
from ase.md.verlet import VelocityVerlet
from ase import units
size = 3
# Set up a crystal
atoms = Diamond(directions=[[1, 0, 0], [0, 1, 0], [0, 0, 1]],
symbol='Si',
size=(size, size, size),
pbc=True)
# Describe the interatomic interactions with the Effective Medium Theory
atoms.set_calculator(lammps['ref'])
np.random.seed(42)
# Set the momenta corresponding to T=300K
MaxwellBoltzmannDistribution(atoms, 1000 * units.kB)
# We want to run MD with constant energy using the VelocityVerlet algorithm.
dyn = VelocityVerlet(atoms, 1 * units.fs) # 5 fs time step.
def printenergy(a):
"""Function to print the potential, kinetic and total energy"""
epot = a.get_potential_energy() / len(a)
ekin = a.get_kinetic_energy() / len(a)
delta = np.max(np.abs(a.get_forces() - lammps['our'].get_forces(a)))
print('Energy per atom: Epot = %.3feV Ekin = %.3feV (T=%3.0fK) '
# set of utility routines specific this this model/testing framework
from utilities import relax_atoms, relax_atoms_cell
# the current model
import model
a0 = 5.44 # initial guess at lattice constant, cell will be relaxed below
fmax = 0.01 # maximum force following relaxtion [eV/A]
if not hasattr(model, 'bulk_reference'):
# set up the a
bulk = Diamond(symbol='Si', latticeconstant=a0)
# specify that we will use model.calculator to compute forces, energies and stresses
bulk.set_calculator(model.calculator)
# use one of the routines from utilities module to relax the initial
# unit cell and atomic positions
bulk = relax_atoms_cell(bulk, tol=fmax, traj_file=None)
else:
bulk = model.bulk_reference.copy()
bulk.set_calculator(model.calculator)
a0 = bulk.cell[0, 0] # get lattice constant from relaxed bulk
print "got a0 ", a0
bulk = Diamond(symbol="Si",
latticeconstant=a0,
directions=[[1, -1, 0], [1, 0, -1], [1, 1, 1]])
bulk.set_calculator(model.calculator)
from utilities import relax_atoms, relax_atoms_cell
# the current model
import model
np.random.seed(75)
a0 = 5.44
fmax = 0.01 # maximum force following relaxtion [eV/A]
if not hasattr(model, 'bulk_reference'):
# set up the a
bulk = Diamond(symbol='Si', latticeconstant=a0)
# specify that we will use model.calculator to compute forces, energies and stresses
bulk.set_calculator(model.calculator)
# use one of the routines from utilities module to relax the initial
# unit cell and atomic positions
bulk = relax_atoms_cell(bulk, tol=fmax, traj_file=None)
else:
bulk = model.bulk_reference.copy()
bulk.set_calculator(model.bulk_reference.calc)
print "BOB doing bulk pot en"
e0 = bulk.get_potential_energy()/float(len(bulk))
print "got e0 ", e0
def interface_energy(e0, filename):
class TestFitElasticConstants(matscipytest.MatSciPyTestCase):
"""
Tests of elastic constant calculation.
We test with a cubic Silicon lattice, since this also has most of the
symmetries of the lower-symmetry crystal families
"""
def setUp(self):
self.pot = quippy.Potential('IP SW', param_str="""
<SW_params n_types="2" label="PRB_31_plus_H">
<comment> Stillinger and Weber, Phys. Rev. B 31 p 5262 (1984), extended for other elements </comment>
<per_type_data type="1" atomic_num="1" />
<per_type_data type="2" atomic_num="14" />
<per_pair_data atnum_i="1" atnum_j="1" AA="0.0" BB="0.0"
p="0" q="0" a="1.0" sigma="1.0" eps="0.0" />
<per_pair_data atnum_i="1" atnum_j="14" AA="8.581214" BB="0.0327827"
p="4" q="0" a="1.25" sigma="2.537884" eps="2.1672" />
<per_pair_data atnum_i="14" atnum_j="14" AA="7.049556277" BB="0.6022245584"
p="4" q="0" a="1.80" sigma="2.0951" eps="2.1675" />
<!-- triplet terms: atnum_c is the center atom, neighbours j and k -->
<per_triplet_data atnum_c="1" atnum_j="1" atnum_k="1"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="1" atnum_j="1" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="1" atnum_j="14" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="14" atnum_j="1" atnum_k="1"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="14" atnum_j="1" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
<per_triplet_data atnum_c="14" atnum_j="14" atnum_k="14"
lambda="21.0" gamma="1.20" eps="2.1675" />
</SW_params>
""")
self.fmax = 1e-4
self.at0 = Diamond('Si', latticeconstant=5.43)
self.at0.set_calculator(self.pot)
# relax initial positions and unit cell
FIRE(StrainFilter(self.at0, mask=[1,1,1,0,0,0]), logfile=None).run(fmax=self.fmax)
self.C_ref = np.array([[ 151.4276439 , 76.57244456, 76.57244456, 0. , 0. , 0. ],
[ 76.57244456, 151.4276439 , 76.57244456, 0. , 0. , 0. ],
[ 76.57244456, 76.57244456, 151.4276439 , 0. , 0. , 0. ],
[ 0. , 0. , 0. , 109.85498798, 0. , 0. ],
[ 0. , 0. , 0. , 0. , 109.85498798, 0. ],
[ 0. , 0. , 0. , 0. , 0. , 109.85498798]])
self.C_err_ref = np.array([[ 1.73091718, 1.63682097, 1.63682097, 0. , 0. , 0. ],
[ 1.63682097, 1.73091718, 1.63682097, 0. , 0. , 0. ],
[ 1.63682097, 1.63682097, 1.73091718, 0. , 0. , 0. ],
[ 0. , 0. , 0. , 1.65751232, 0. , 0. ],
[ 0. , 0. , 0. , 0. , 1.65751232, 0. ],
[ 0. , 0. , 0. , 0. , 0. , 1.65751232]])
self.C_ref_relaxed = np.array([[ 151.28712587, 76.5394162 , 76.5394162 , 0. ,
0. , 0. ],
[ 76.5394162 , 151.28712587, 76.5394162 , 0. ,
0. , 0. ],
[ 76.5394162 , 76.5394162 , 151.28712587, 0. ,
0. , 0. ],
[ 0. , 0. , 0. , 56.32421772,
0. , 0. ],
[ 0. , 0. , 0. , 0. ,
56.32421772, 0. ],
[ 0. , 0. , 0. , 0. ,
0. , 56.32421772]])
self.C_err_ref_relaxed = np.array([[ 1.17748661, 1.33333615, 1.33333615, 0. , 0. ,
0. ],
[ 1.33333615, 1.17748661, 1.33333615, 0. , 0. ,
0. ],
[ 1.33333615, 1.33333615, 1.17748661, 0. , 0. ,
0. ],
[ 0. , 0. , 0. , 0.18959684, 0. ,
0. ],
[ 0. , 0. , 0. , 0. , 0.18959684,
0. ],
[ 0. , 0. , 0. , 0. , 0. ,
0.18959684]])
def test_measure_triclinic_unrelaxed(self):
# compare to brute force method without relaxation
C = measure_triclinic_elastic_constants(self.at0, delta=1e-2, optimizer=None)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref, tol=0.2)
def test_measure_triclinic_relaxed(self):
# compare to brute force method with relaxation
C = measure_triclinic_elastic_constants(self.at0, delta=1e-2, optimizer=FIRE,
fmax=self.fmax)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref_relaxed, tol=0.2)
def testcubic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'cubic', verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref, tol=1e-2)
if abs(C_err).max() > 0.:
self.assertArrayAlmostEqual(C_err/units.GPa, self.C_err_ref, tol=1e-2)
def testcubic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'cubic',
optimizer=FIRE, fmax=self.fmax,
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref_relaxed, tol=1e-2)
if abs(C_err).max() > 0.:
self.assertArrayAlmostEqual(C_err/units.GPa, self.C_err_ref_relaxed, tol=1e-2)
def testorthorhombic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'orthorhombic',
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref, tol=0.1)
def testorthorhombic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'orthorhombic',
optimizer=FIRE, fmax=self.fmax,
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref_relaxed, tol=0.1)
def testmonoclinic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'monoclinic',
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref, tol=0.2)
def testmonoclinic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'monoclinic',
optimizer=FIRE, fmax=self.fmax,
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref_relaxed, tol=0.2)
def testtriclinic_unrelaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'triclinic',
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref, tol=0.2)
def testtriclinic_relaxed(self):
C, C_err = fit_elastic_constants(self.at0, 'triclinic',
optimizer=FIRE, fmax=self.fmax,
verbose=False, graphics=False)
self.assertArrayAlmostEqual(C/units.GPa, self.C_ref_relaxed, tol=0.2)
