def test_as_pyscf_method(self):
mol = gto.M(atom='''
O 0. 0. 0.
H 0. -0.757 0.587
H 0. 0.757 0.587
''', verbose=0)
mol1 = geometric_solver.optimize(scf.RHF(mol))
self.assertAlmostEqual(lib.finger(mol1.atom_coords()),
2.19951175503979, 4)
def test_optimize(self):
mol = gto.M(atom='''
O 0. 0. 0.
H 0. -0.757 0.587
H 0. 0.757 0.587
''',
symmetry=True, verbose=0)
mol1 = geometric_solver.optimize(scf.RHF(mol))
self.assertAlmostEqual(lib.finger(mol1.atom_coords()),
3.038506469458414, 4)
self.assertEqual(mol1.symmetry, 'C2v')
def test_optimize(self):
mol = gto.M(atom='''
O 0. 0. 0.
H 0. -0.757 0.587
H 0. 0.757 0.587
''',
symmetry=True,
verbose=0)
mol1 = geometric_solver.optimize(scf.RHF(mol))
self.assertAlmostEqual(lib.finger(mol1.atom_coords()),
3.038506469458414, 4)
self.assertEqual(mol1.symmetry, 'C2v')
def Q3OptimizerDriver(pre_energy, mol , HF, R, conv_params):
if pre_energy ==True:
opt_pre = optimize(pre_energy, **conv_params)
opt_geom = opt_pre.tofile(working_dir+'/opt-geom.xyz')
else:
if theory==KS:
if shell == O:
SPE =scf.UKS(mol).run()
else shell == R:
SPE =scf.RKS(mol).run()
else theory == HF:
if shell == O:
SPE =scf.UHF(mol).run()
else shell == R:
SPE =scf.RHF(mol).run()
def f(mol):
e, g = grad_scan(mol)
r = mol.atom_coords()
penalty = np.linalg.norm(r[0] - r[1])**2 * 0.1
e += penalty
g[0] += (r[0] - r[1]) * 2 * 0.1
g[1] -= (r[0] - r[1]) * 2 * 0.1
print('Customized |g|', np.linalg.norm(g))
return e, g
#
# Function as_pyscf_method is a wrapper that convert the "energy-gradients"
# function to berny_solver. The "energy-gradients" function takes the Mole
# object as geometry input, and returns the energy and gradients of that
# geometry.
#
fake_method = as_pyscf_method(mol, f)
new_mol = berny_solver.optimize(fake_method)
print('Old geometry (Bohr)')
print(mol.atom_coords())
print('New geometry (Bohr)')
print(new_mol.atom_coords())
#
# Geometry can be also optimized with geomeTRIC library
#
new_mol = geometric_solver.optimize(fake_method)
'''
import numpy
from pyscf import gto, scf, cc, qmmm
from pyscf.geomopt import berny_solver
mol = gto.M(atom='''
C 1.1879 -0.3829 0.0000
C 0.0000 0.5526 0.0000
O -1.1867 -0.2472 0.0000
H -1.9237 0.3850 0.0000
H 2.0985 0.2306 0.0000
H 1.1184 -1.0093 0.8869
H 1.1184 -1.0093 -0.8869
H -0.0227 1.1812 0.8852
H -0.0227 1.1812 -0.8852
''',
basis='3-21g')
numpy.random.seed(1)
coords = numpy.random.random((5, 3)) * 10
charges = (numpy.arange(5) + 1.) * -.001
mf = qmmm.mm_charge(scf.RHF(mol), coords, charges)
#mf.verbose=4
#mf.kernel()
mol1 = berny_solver.optimize(mf)
from pyscf.geomopt import geometric_solver
mycc = cc.CCSD(mf)
mol1 = geometric_solver.optimize(mycc)
'''
Use geomeTRIC library to optimize the molecular geometry.
'''
from pyscf import gto, scf
from pyscf.geomopt.geometric_solver import optimize
mol = gto.M(atom='N 0 0 0; N 0 0 1.2', basis='ccpvdz')
mf = scf.RHF(mol)
#
# geometry optimization for HF. There are two entries to invoke the geomeTRIC
# optimization
#
# method 1: import the optimize function from pyscf.geomopt.geometric_solver
mol_eq = optimize(mf)
print(mol_eq.atom_coords())
# method 2: create the optimizer from Gradients class
mol_eq = mf.Gradients().optimizer(solver='geomeTRIC').kernel()
#
# geometry optimization for CASSCF
#
from pyscf import mcscf
mf = scf.RHF(mol)
mc = mcscf.CASSCF(mf, 4, 4)
conv_params = {
'convergence_energy': 1e-4, # Eh
'convergence_grms': 3e-3, # Eh/Bohr
'convergence_gmax': 4.5e-3, # Eh/Bohr
print("Initial energy: {:.8e}".format(mc.e_tot[0]))
mc.nuc_grad_iroot = 0
def my_call(env):
carts = env['mol'].atom_coords() * BOHR
h2co_geom_analysis(carts)
conv_params = {
'convergence_energy': 1e-6, # Eh
'convergence_grms': 3e-5, # Eh/Bohr
'convergence_gmax': 4.5e-5, # Eh/Bohr
'convergence_drms': 1.2e-5, # Angstrom
'convergence_dmax': 1.8e-5, # Angstrom
}
conv, mol_eq = optimize(mc, callback=my_call, **conv_params)
molcas_geom = np.asarray([[0.550219, -0.000000, -0.000000],
[-0.690238, -0.000000, -0.000000],
[1.139489, -0.000000, 0.937479],
[1.139489, -0.000000, -0.937479]])
print(
"SA(2) tPBE(6,6)/6-31g optimized geometry of first root of formaldehdye:")
h2co_geom_analysis(mol_eq.atom_coords() * BOHR)
print(
"OpenMolcas's opinion using analytical gradient implementation (note OpenMolcas and PySCF have different quadrature grids):"
)
h2co_geom_analysis(molcas_geom)
#!/usr/bin/env python
#
# Author: Qiming Sun <*****@*****.**>
#
'''
Geometry optimization with solvent model
'''
from pyscf import gto, scf, dft
from pyscf import solvent
from pyscf.geomopt import geometric_solver
mol = gto.M(atom='''
C 0.000000 0.000000 -0.542500
O 0.000000 0.000000 0.677500
H 0.000000 0.9353074360871938 -1.082500
H 0.000000 -0.9353074360871938 -1.082500
''',
verbose=4)
mf = solvent.ddCOSMO(scf.RHF(mol))
new_mol = geometric_solver.optimize(mf)
Raises:
Q3_Error: Invalid Input
"""
conv_params = { # These are the default settings
'convergence_energy': 1e-6, # Eh
'convergence_grms': 3e-4, # Eh/Bohr
'convergence_gmax': 4.5e-4, # Eh/Bohr
'convergence_drms': 1.2e-3, # Angstrom
'convergence_dmax': 1.8e-3, # Angstrom
}
def Q3OptimizerDriver(pre_energy, mol , HF, R, conv_params):
if pre_energy ==True:
opt_pre = optimize(pre_energy, **conv_params)
opt_geom = opt_pre.tofile(working_dir+'/opt-geom.xyz')
else:
if theory==KS:
if shell == O:
SPE =scf.UKS(mol).run()
else shell == R:
SPE =scf.RKS(mol).run()
else theory == HF:
if shell == O:
SPE =scf.UHF(mol).run()
else shell == R:
SPE =scf.RHF(mol).run()
SPE_opt = optimize(SPE, **conv_params)
opt_geom = mol_eq.tofile(working_dir+'opt-geom.xyz')
mol = gto.M(atom='''
C 0.000000 0.000000 -0.542500
O 0.000000 0.000000 0.677500
H 0.000000 0.9353074360871938 -1.082500
H 0.000000 -0.9353074360871938 -1.082500
''',
basis='3-21g')
mf = scf.RHF(mol)
# Run analyze function in callback
def cb(envs):
mf = envs['g_scanner'].base
mf.analyze(verbose=4)
#
# Method 1: Pass callback to optimize function
#
geometric_solver.optimize(mf, callback=cb)
berny_solver.optimize(mf, callback=cb)
#
# Method 2: Add callback to geometry optimizer
#
opt = mf.nuc_grad_method().as_scanner().optimizer()
opt.callback = cb
opt.kernel()
'''
import numpy
from pyscf import gto, scf, cc, qmmm
from pyscf.geomopt import berny_solver
mol = gto.M(atom='''
C 1.1879 -0.3829 0.0000
C 0.0000 0.5526 0.0000
O -1.1867 -0.2472 0.0000
H -1.9237 0.3850 0.0000
H 2.0985 0.2306 0.0000
H 1.1184 -1.0093 0.8869
H 1.1184 -1.0093 -0.8869
H -0.0227 1.1812 0.8852
H -0.0227 1.1812 -0.8852
''',
basis='3-21g')
numpy.random.seed(1)
coords = numpy.random.random((5,3)) * 10
charges = (numpy.arange(5) + 1.) * -.001
mf = qmmm.mm_charge(scf.RHF(mol), coords, charges)
#mf.verbose=4
#mf.kernel()
mol1 = berny_solver.optimize(mf)
from pyscf.geomopt import geometric_solver
mycc = cc.CCSD(mf)
mol1 = geometric_solver.optimize(mycc)
#!/usr/bin/env python
'''
Use geomeTRIC library to optimize the molecular geometry.
'''
from pyscf import gto, scf
from pyscf.geomopt.geometric_solver import optimize
mol = gto.M(atom='N 0 0 0; N 0 0 1.2', basis='ccpvdz')
mf = scf.RHF(mol)
#
# geometry optimization for HF
#
mol_eq = optimize(mf)
print(mol_eq.atom_coords())
#
# geometry optimization for CASSCF
#
from pyscf import mcscf
mf = scf.RHF(mol)
mc = mcscf.CASSCF(mf, 4, 4)
conv_params = {
'convergence_energy': 1e-4, # Eh
'convergence_grms': 3e-3, # Eh/Bohr
'convergence_gmax': 4.5e-3, # Eh/Bohr
'convergence_drms': 1.2e-2, # Angstrom
'convergence_dmax': 1.8e-2, # Angstrom
}
