def compute_gradient(self, xyz):
geom2 = self.rhf.geometry.update_xyz(xyz)
rhf2 = est.RHF(self.rhf.options.copy().set_values({
'geometry': geom2,
}))
rhf2.compute_energy(Dguess=self.rhf.tensors['D'])
return rhf2.compute_gradient()
def compute_energy(self, xyz):
geom2 = self.rhf.geometry.update_xyz(xyz)
rhf2 = est.RHF(self.rhf.options.copy().set_values({
'geometry': geom2,
}))
rhf2.compute_energy(Dguess=self.rhf.tensors['D'])
return ls.Tensor.array(rhf2.scalars['Escf'])
def compute_gradient(self, xyz):
geom2 = self.cisci.geometry.update_xyz(xyz)
rhf2 = est.RHF(self.cisci.reference.options.copy().set_values({
'geometry' : geom2,
}))
rhf2.compute_energy(Dguess=self.cisci.reference.tensors['D'])
cisci2 = est.CIS(self.cisci.options.copy().set_values({
'reference' : rhf2,
}))
cisci2.compute_energy()
return cisci2.compute_gradient(self.gradient_S, self.gradient_index)
def compute_energy(self, xyz):
geom2 = self.cisci.geometry.update_xyz(xyz)
rhf2 = est.RHF(self.cisci.reference.options.copy().set_values({
'geometry' : geom2,
}))
rhf2.compute_energy(Dguess=self.cisci.reference.tensors['D'])
cisci2 = est.CIS(self.cisci.options.copy().set_values({
'reference' : rhf2,
}))
cisci2.compute_energy()
return ls.Tensor.array(cisci2.evals[self.energy_S][self.energy_index])
def compute_overlap(self, xyz):
geom2 = self.casci.geometry.update_xyz(xyz)
rhf2 = est.RHF(self.casci.reference.options.copy().set_values({
'geometry':
geom2,
}))
rhf2.compute_energy(Dguess=self.casci.reference.tensors['D'])
casci2 = est.CASCI(self.casci.options.copy().set_values({
'reference':
rhf2,
}))
casci2.compute_energy()
O = est.CASCI.compute_overlap(self.casci, casci2, self.overlap_S)
# Phase the |J'> states to match the <I| states
for k in range(O.shape[1]):
if O[k, k] < 0.0:
O[:, k] *= -1.0
return O[self.overlap_indexA, self.overlap_indexB]
def run_guess(run_dir, run_xyz, run_basis, run_basis_2):
# Validity checks: { key : (delta, OK) }
checks = collections.OrderedDict()
start_time = ls.timings_header()
resources = ls.ResourceList.build()
molecule = ls.Molecule.from_xyz_file(run_dir + run_xyz)
geom = est.Geometry.build(
resources=resources,
molecule=molecule,
basisname=run_basis,
)
geom2 = est.Geometry.build(
resources=resources,
molecule=molecule,
basisname=run_basis_2,
)
# ==> RHF <== #
ref = est.RHF.from_options(
geometry=geom,
)
ref.compute_energy()
# => Identity Guess: should be the same <= #
ref2 = est.RHF(ref.options.copy().set_values({
'geometry' : geom,
}))
ref2.initialize()
Dguess, Cocc_mom, Cocc_act = est.RHF.guess_density(ref, ref2)
ref2.compute_energy(Dguess)
dD = np.max(np.abs(Dguess[...] - ref.tensors['D']))
checks['dD-RHF'] = (dD, dD < 1.0E-13)
# => Cast-Up Guess <= #
ref2 = est.RHF(ref.options.copy().set_values({
'geometry' : geom2,
}))
ref2.initialize()
Dguess, Cocc_mom, Cact_mom = est.RHF.guess_density(ref, ref2)
ref2.compute_energy(Dguess)
# Check if NO structure makes sense
n1s = ref.tensors['n']
n2s = ls.Tensor.eigh(ls.Tensor.chain([ref2.tensors['X'], ref2.tensors['S'], Dguess, ref2.tensors['S'], ref2.tensors['X']], [True, False, False, False, False]))[0]
dn = np.max([abs(n1 - n2) for n1, n2 in zip(sorted(-np.abs(n1s)), sorted(-np.abs(n2s)))])
checks['dn-RHF'] = (dn, dn < 1.0E-12)
# ==> FOMO-RHF <== #
ref = est.RHF.from_options(
geometry=geom,
fomo=True,
fomo_method='gaussian',
fomo_temp=0.3,
fomo_nocc=2,
fomo_nact=5,
)
ref.compute_energy()
# => Identity Guess: should be the same <= #
ref2 = est.RHF(ref.options.copy().set_values({
'geometry' : geom,
}))
ref2.initialize()
Dguess, Cocc_mom, Cact_mom = est.RHF.guess_density(ref, ref2)
ref2.compute_energy(Dguess)
checks['dD-FOMO-RHF'] = (dD, dD < 1.0E-13)
# => Cast-Up Guess <= #
ref2 = est.RHF(ref.options.copy().set_values({
'geometry' : geom2,
}))
ref2.initialize()
Dguess, Cocc_mom, Cact_mom = est.RHF.guess_density(ref, ref2)
ref2.compute_energy(Dguess)
# Check if NO structure makes sense
n1s = ref.tensors['n']
n2s = ls.Tensor.eigh(ls.Tensor.chain([ref2.tensors['X'], ref2.tensors['S'], Dguess, ref2.tensors['S'], ref2.tensors['X']], [True, False, False, False, False]))[0]
dn = np.max([abs(n1 - n2) for n1, n2 in zip(sorted(-np.abs(n1s)), sorted(-np.abs(n2s)))])
checks['dn-FOMO-RHF'] = (dn, dn < 1.0E-12)
ls.timings_footer(start_time)
print('Test Results:')
for key, value in checks.items():
print('%-20s: %11.3E %s' % (
key,
value[0],
'OK' if value[1] else 'BAD',
))
return all(x[1] for x in list(checks.values()))