# Copyright 2014-2022 The PySCF Developers. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# Author: Abdelrahman Ahmed <>
# Samragni Banerjee <samragnibanerjee4@gmail.com>
# James Serna <jamcar456@gmail.com>
# Terrence Stahl <>
# Alexander Sokolov <alexander.y.sokolov@gmail.com>
'''
Unrestricted algebraic diagrammatic construction
'''
import numpy as np
from pyscf import lib, symm
from pyscf.lib import logger
from pyscf.adc import uadc_ao2mo
from pyscf.adc import uadc_amplitudes
from pyscf.adc import radc_ao2mo
from pyscf.adc import dfadc
from pyscf import __config__
from pyscf import df
from pyscf import scf
# Excited-state kernel
[docs]
def kernel(adc, nroots=1, guess=None, eris=None, verbose=None):
adc.method = adc.method.lower()
if adc.method not in ("adc(2)", "adc(2)-x", "adc(3)"):
raise NotImplementedError(adc.method)
cput0 = (logger.process_clock(), logger.perf_counter())
log = logger.Logger(adc.stdout, adc.verbose)
if adc.verbose >= logger.WARN:
adc.check_sanity()
adc.dump_flags()
if isinstance(adc._scf, scf.rohf.ROHF) and (adc.method_type == "ip" or adc.method_type == "ea"):
logger.warn(
adc, "EA/IP-ADC with the ROHF reference do not incorporate the occ-vir Fock matrix elements...")
if eris is None:
eris = adc.transform_integrals()
imds = adc.get_imds(eris)
matvec, diag = adc.gen_matvec(imds, eris)
guess = adc.get_init_guess(nroots, diag, ascending = True)
conv, adc.E, U = lib.linalg_helper.davidson1(
lambda xs : [matvec(x) for x in xs],
guess, diag, nroots=nroots, verbose=log, tol=adc.conv_tol,
max_cycle=adc.max_cycle, max_space=adc.max_space, tol_residual=adc.tol_residual)
adc.U = np.array(U).T.copy()
if adc.compute_properties:
adc.P,adc.X = adc.get_properties(nroots)
nfalse = np.shape(conv)[0] - np.sum(conv)
header = ("\n*************************************************************"
"\n ADC calculation summary"
"\n*************************************************************")
logger.info(adc, header)
for n in range(nroots):
print_string = ('%s root %d | Energy (Eh) = %14.10f | Energy (eV) = %12.8f ' %
(adc.method, n, adc.E[n], adc.E[n]*27.2114))
if adc.compute_properties:
print_string += ("| Spec. factor = %10.8f " % adc.P[n])
print_string += ("| conv = %s" % conv[n])
logger.info(adc, print_string)
if nfalse >= 1:
logger.warn(adc, "Davidson iterations for " + str(nfalse) + " root(s) did not converge!!!")
log.timer('ADC', *cput0)
return adc.E, adc.U, adc.P, adc.X
[docs]
class UADC(lib.StreamObject):
'''Ground state calculations
Attributes:
verbose : int
Print level. Default value equals to :class:`Mole.verbose`
max_memory : float or int
Allowed memory in MB. Default value equals to :class:`Mole.max_memory`
incore_complete : bool
Avoid all I/O. Default is False.
method : string
nth-order ADC method. Options are : ADC(2), ADC(2)-X, ADC(3). Default is ADC(2).
>>> mol = gto.M(atom = 'H 0 0 0; F 0 0 1.1', basis = 'ccpvdz')
>>> mf = scf.RHF(mol).run()
>>> myadc = adc.UADC(mf).run()
Saved results
e_corr : float
MPn correlation correction
e_tot : float
Total energy (HF + correlation)
t1, t2 :
T amplitudes t1[i,a], t2[i,j,a,b] (i,j in occ, a,b in virt)
'''
incore_complete = getattr(__config__, 'adc_uadc_UADC_incore_complete', False)
_keys = {
'tol_residual','conv_tol', 'e_corr', 'method', 'method_type', 'mo_coeff',
'mol', 'mo_energy_a', 'mo_energy_b', 'incore_complete',
'scf_energy', 'e_tot', 't1', 't2', 'frozen', 'chkfile',
'max_space', 'mo_occ', 'max_cycle', 'imds', 'with_df', 'compute_properties',
'approx_trans_moments', 'evec_print_tol', 'spec_factor_print_tol',
'E', 'U', 'P', 'X', 'ncvs', 'dip_mom', 'dip_mom_nuc',
'spin_c', 'f_ov'
}
def __init__(self, mf, frozen=None, mo_coeff=None, mo_occ=None):
from pyscf import gto
if 'dft' in str(mf.__module__):
raise NotImplementedError('DFT reference for UADC')
if mo_coeff is None:
mo_coeff = mf.mo_coeff
if mo_occ is None:
mo_occ = mf.mo_occ
self.mol = mf.mol
self._scf = mf
self.verbose = self.mol.verbose
self.stdout = self.mol.stdout
self.max_memory = mf.max_memory
self.max_space = getattr(__config__, 'adc_uadc_UADC_max_space', 12)
self.max_cycle = getattr(__config__, 'adc_uadc_UADC_max_cycle', 50)
self.conv_tol = getattr(__config__, 'adc_uadc_UADC_conv_tol', 1e-12)
self.tol_residual = getattr(__config__, 'adc_uadc_UADC_tol_residual', 1e-6)
self.scf_energy = mf.e_tot
self.frozen = frozen
self.incore_complete = self.incore_complete or self.mol.incore_anyway
self.f_ov = None
if isinstance(mf, scf.rohf.ROHF):
logger.info(mf, "\nROHF reference detected in ADC, semicanonicalizing the orbitals...")
mo_a = mo_coeff.copy()
nalpha = mf.mol.nelec[0]
nbeta = mf.mol.nelec[1]
h1e = mf.get_hcore()
dm = mf.make_rdm1()
vhf = mf.get_veff(mf.mol, dm)
fock_a = h1e + vhf[0]
fock_b = h1e + vhf[1]
if nalpha > nbeta:
ndocc = nbeta
nsocc = nalpha - nbeta
else:
ndocc = nalpha
nsocc = nbeta - nalpha
fock_a = np.dot(mo_a.T,np.dot(fock_a, mo_a))
fock_b = np.dot(mo_a.T,np.dot(fock_b, mo_a))
# Semicanonicalize Ca using fock_a, nocc_a -> Ca, mo_energy_a, U_a, f_ov_a
mo_a, mo_energy_a, f_ov_a, f_aa = self.semi_canonicalize_orbitals(
fock_a, ndocc + nsocc, mo_a)
# Semicanonicalize Cb using fock_b, nocc_b -> Cb, mo_energy_b, U_b, f_ov_b
mo_b, mo_energy_b, f_ov_b, f_bb = self.semi_canonicalize_orbitals(fock_b, ndocc, mo_a)
mo_coeff = [mo_a, mo_b]
f_ov = [f_ov_a, f_ov_b]
self.f_ov = f_ov
self.spin_c = True
self.mo_energy_a = mo_energy_a.copy()
self.mo_energy_b = mo_energy_b.copy()
else:
self.mo_energy_a = mf.mo_energy[0]
self.mo_energy_b = mf.mo_energy[1]
self.mo_coeff = mo_coeff
self.mo_occ = mo_occ
self.e_corr = None
self.t1 = None
self.t2 = None
self.imds = lambda:None
self._nocc = mf.nelec
self._nmo = (mo_coeff[0].shape[1], mo_coeff[1].shape[1])
self._nvir = (self._nmo[0] - self._nocc[0], self._nmo[1] - self._nocc[1])
self.chkfile = mf.chkfile
self.method = "adc(2)"
self.method_type = "ip"
self.with_df = None
self.compute_properties = True
self.approx_trans_moments = False
self.evec_print_tol = 0.1
self.spec_factor_print_tol = 0.1
self.ncvs = None
self.E = None
self.U = None
self.P = None
self.X = (None,)
self.spin_c = False
dip_ints = -self.mol.intor('int1e_r',comp=3)
dip_mom_a = np.zeros((dip_ints.shape[0], self._nmo[0], self._nmo[0]))
dip_mom_b = np.zeros((dip_ints.shape[0], self._nmo[1], self._nmo[1]))
for i in range(dip_ints.shape[0]):
dip = dip_ints[i,:,:]
dip_mom_a[i,:,:] = np.dot(mo_coeff[0].T, np.dot(dip, mo_coeff[0]))
dip_mom_b[i,:,:] = np.dot(mo_coeff[1].T, np.dot(dip, mo_coeff[1]))
self.dip_mom = []
self.dip_mom.append(dip_mom_a)
self.dip_mom.append(dip_mom_b)
charges = self.mol.atom_charges()
coords = self.mol.atom_coords()
self.dip_mom_nuc = lib.einsum('i,ix->x', charges, coords)
compute_amplitudes = uadc_amplitudes.compute_amplitudes
compute_energy = uadc_amplitudes.compute_energy
transform_integrals = uadc_ao2mo.transform_integrals_incore
[docs]
def semi_canonicalize_orbitals(self, f, nocc, C):
# Diagonalize occ-occ block
evals_oo, evecs_oo = np.linalg.eigh(f[:nocc, :nocc])
# Diagonalize virt-virt block
evals_vv, evecs_vv = np.linalg.eigh(f[nocc:, nocc:])
evals = np.hstack((evals_oo, evals_vv))
U = np.zeros_like(f)
U[:nocc, :nocc] = evecs_oo
U[nocc:, nocc:] = evecs_vv
C = np.dot(C, U)
transform_f = np.dot(U.T, np.dot(f, U))
f_ov = transform_f[:nocc, nocc:].copy()
return C, evals, f_ov, transform_f
[docs]
def dump_flags(self, verbose=None):
logger.info(self, '')
logger.info(self, '******** %s ********', self.__class__)
logger.info(self, 'max_space = %d', self.max_space)
logger.info(self, 'max_cycle = %d', self.max_cycle)
logger.info(self, 'conv_tol = %s', self.conv_tol)
logger.info(self, 'tol_residual = %s', self.tol_residual)
logger.info(self, 'max_memory %d MB (current use %d MB)',
self.max_memory, lib.current_memory()[0])
return self
[docs]
def dump_flags_gs(self, verbose=None):
logger.info(self, '')
logger.info(self, '******** %s ********', self.__class__)
logger.info(self, 'max_memory %d MB (current use %d MB)',
self.max_memory, lib.current_memory()[0])
return self
[docs]
def kernel_gs(self):
assert(self.mo_coeff is not None)
assert(self.mo_occ is not None)
self.method = self.method.lower()
if self.method not in ("adc(2)", "adc(2)-x", "adc(3)"):
raise NotImplementedError(self.method)
if self.verbose >= logger.WARN:
self.check_sanity()
self.dump_flags_gs()
nmo_a, nmo_b = self._nmo
nao = self.mo_coeff[0].shape[0]
nmo_pair = nmo_a * (nmo_a+1) // 2
nao_pair = nao * (nao+1) // 2
mem_incore = (max(nao_pair**2, nmo_a**4) + nmo_pair**2) * 2 * 8/1e6
mem_now = lib.current_memory()[0]
if getattr(self, 'with_df', None) or getattr(self._scf, 'with_df', None):
if getattr(self, 'with_df', None):
self.with_df = self.with_df
else:
self.with_df = self._scf.with_df
def df_transform():
return uadc_ao2mo.transform_integrals_df(self)
self.transform_integrals = df_transform
elif (self._scf._eri is None or
(mem_incore+mem_now >= self.max_memory and not self.incore_complete)):
def outcore_transform():
return uadc_ao2mo.transform_integrals_outcore(self)
self.transform_integrals = outcore_transform
eris = self.transform_integrals()
self.e_corr, self.t1, self.t2 = uadc_amplitudes.compute_amplitudes_energy(
self, eris=eris, verbose=self.verbose)
self._finalize()
return self.e_corr, self.t1, self.t2
[docs]
def kernel(self, nroots=1, guess=None, eris=None):
assert (self.mo_coeff is not None)
assert (self.mo_occ is not None)
self.method = self.method.lower()
if self.method not in ("adc(2)", "adc(2)-x", "adc(3)"):
raise NotImplementedError(self.method)
if self.verbose >= logger.WARN:
self.check_sanity()
self.dump_flags_gs()
nmo_a, nmo_b = self._nmo
nao = self.mo_coeff[0].shape[0]
nmo_pair = nmo_a * (nmo_a+1) // 2
nao_pair = nao * (nao+1) // 2
mem_incore = (max(nao_pair**2, nmo_a**4) + nmo_pair**2) * 2 * 8/1e6
mem_now = lib.current_memory()[0]
if getattr(self, 'with_df', None) or getattr(self._scf, 'with_df', None):
if getattr(self, 'with_df', None):
self.with_df = self.with_df
else:
self.with_df = self._scf.with_df
def df_transform():
return uadc_ao2mo.transform_integrals_df(self)
self.transform_integrals = df_transform
elif (self._scf._eri is None or
(mem_incore+mem_now >= self.max_memory and not self.incore_complete)):
def outcore_transform():
return uadc_ao2mo.transform_integrals_outcore(self)
self.transform_integrals = outcore_transform
eris = self.transform_integrals()
self.e_corr, self.t1, self.t2 = uadc_amplitudes.compute_amplitudes_energy(
self, eris=eris, verbose=self.verbose)
self._finalize()
self.method_type = self.method_type.lower()
if (self.method_type == "ea"):
e_exc, v_exc, spec_fac, X, adc_es = self.ea_adc(nroots=nroots, guess=guess, eris=eris)
elif(self.method_type == "ip"):
if not isinstance(self.ncvs, type(None)) and self.ncvs > 0:
e_exc, v_exc, spec_fac, X, adc_es = self.ip_cvs_adc(
nroots=nroots, guess=guess, eris=eris)
else:
e_exc, v_exc, spec_fac, X, adc_es = self.ip_adc(
nroots=nroots, guess=guess, eris=eris)
else:
raise NotImplementedError(self.method_type)
self._adc_es = adc_es
return e_exc, v_exc, spec_fac, X
def _finalize(self):
'''Hook for dumping results and clearing up the object.'''
logger.note(self, 'MP%s correlation energy of reference state (a.u.) = %.8f',
self.method[4], self.e_corr)
return self
[docs]
def ea_adc(self, nroots=1, guess=None, eris=None):
from pyscf.adc import uadc_ea
adc_es = uadc_ea.UADCEA(self)
e_exc, v_exc, spec_fac, x = adc_es.kernel(nroots, guess, eris)
return e_exc, v_exc, spec_fac, x, adc_es
[docs]
def ip_adc(self, nroots=1, guess=None, eris=None):
from pyscf.adc import uadc_ip
adc_es = uadc_ip.UADCIP(self)
e_exc, v_exc, spec_fac, x = adc_es.kernel(nroots, guess, eris)
return e_exc, v_exc, spec_fac, x, adc_es
[docs]
def ip_cvs_adc(self, nroots=1, guess=None, eris=None):
from pyscf.adc import uadc_ip_cvs
adc_es = uadc_ip_cvs.UADCIPCVS(self)
e_exc, v_exc, spec_fac, x = adc_es.kernel(nroots, guess, eris)
return e_exc, v_exc, spec_fac, x, adc_es
[docs]
def density_fit(self, auxbasis=None, with_df=None):
if with_df is None:
self.with_df = df.DF(self._scf.mol)
self.with_df.max_memory = self.max_memory
self.with_df.stdout = self.stdout
self.with_df.verbose = self.verbose
if auxbasis is None:
self.with_df.auxbasis = self._scf.with_df.auxbasis
else:
self.with_df.auxbasis = auxbasis
else:
self.with_df = with_df
return self
[docs]
def analyze(self):
self._adc_es.analyze()
[docs]
def compute_dyson_mo(self):
return self._adc_es.compute_dyson_mo()
[docs]
def make_rdm1(self):
return self._adc_es.make_rdm1()
if __name__ == '__main__':
from pyscf import gto
from pyscf import adc
r = 1.098
mol = gto.Mole()
mol.atom = [
['N', (0., 0. , -r/2 )],
['N', (0., 0. , r/2)],]
mol.basis = {'N':'aug-cc-pvdz'}
mol.verbose = 0
mol.build()
mf = scf.UHF(mol)
mf.conv_tol = 1e-12
mf.kernel()
myadc = adc.ADC(mf)
ecorr, t_amp1, t_amp2 = myadc.kernel_gs()
print(ecorr - -0.32201692499346535)
myadcip = adc.uadc_ip.UADCIP(myadc)
e,v,p = kernel(myadcip,nroots=3)
print("ADC(2) IP energies")
print (e[0] - 0.5434389897908212)
print (e[1] - 0.5434389942222756)
print (e[2] - 0.6240296265084732)
print("ADC(2) IP spectroscopic factors")
print (p[0] - 0.884404855445607)
print (p[1] - 0.8844048539643351)
print (p[2] - 0.9096460559671828)
myadcea = adc.uadc_ea.UADCEA(myadc)
e,v,p = kernel(myadcea,nroots=3)
print("ADC(2) EA energies")
print (e[0] - 0.09617819143037348)
print (e[1] - 0.09617819161265123)
print (e[2] - 0.12583269048810924)
print("ADC(2) EA spectroscopic factors")
print (p[0] - 0.991642716974455)
print (p[1] - 0.9916427170555298)
print (p[2] - 0.9817184409336244)
myadc = adc.ADC(mf)
myadc.method = "adc(3)"
ecorr, t_amp1, t_amp2 = myadc.kernel_gs()
print(ecorr - -0.31694173142858517)
myadcip = adc.uadc_ip.UADCIP(myadc)
e,v,p = kernel(myadcip,nroots=3)
print("ADC(3) IP energies")
print (e[0] - 0.5667526838174817)
print (e[1] - 0.5667526888293601)
print (e[2] - 0.6099995181296374)
print("ADC(3) IP spectroscopic factors")
print (p[0] - 0.9086596203469742)
print (p[1] - 0.9086596190173993)
print (p[2] - 0.9214613318791076)
myadcea = adc.uadc_ea.UADCEA(myadc)
e,v,p = kernel(myadcea,nroots=3)
print("ADC(3) EA energies")
print (e[0] - 0.09836545519235675)
print (e[1] - 0.09836545535587536)
print (e[2] - 0.12957093060942082)
print("ADC(3) EA spectroscopic factors")
print (p[0] - 0.9920495578633931)
print (p[1] - 0.992049557938337)
print (p[2] - 0.9819274864738444)
myadc.method = "adc(2)-x"
e,v,p = myadc.kernel(nroots=4)
print("ADC(2)-x IP energies")
print (e[0] - 0.5405255355249104)
print (e[1] - 0.5405255399061982)
print (e[2] - 0.62080267098272)
print (e[3] - 0.620802670982715)
myadc.method_type = "ea"
e,v,p = myadc.kernel(nroots=4)
print("ADC(2)-x EA energies")
print (e[0] - 0.09530653292650725)
print (e[1] - 0.09530653311305577)
print (e[2] - 0.1238833077840878)
print (e[3] - 0.12388330873739162)
