Time-dependent Hartree-Fock and density functional theory#

Modules: pyscf.tdscf, pyscf.pbc.tdscf

Introduction#

PySCF implements the time-dependent Hartree-Fock (TDHF) and time-dependent density functional theory (TDDFT) (frequency domain) linear response theories to compute excited-state energies and transition properties in the tdscf module. A minimal example that runs a TDDFT calculation is as follows

from pyscf import gto, scf, dft, tddft
mol = gto.Mole()
mol.build(
    atom = 'H 0 0 0; F 0 0 1.1',  # in Angstrom
    basis = '631g',
    symmetry = True,
    verbose = 4,
)

mf = dft.RKS(mol)
mf.xc = 'b3lyp'
mf.kernel()

mytd = tddft.TDDFT(mf)
mytd.nstates = 10
mytd.kernel()
mytd.analyze()

The example above computes the excitation energies, oscillator strengths and transition dipole moments of the ten lowest singlet exicted states.

Theory#

Using first-order time-dependent perturbation theory within HF or KS theory, one obtains the non-Hermitian TDHF or TDDFT equations for the excitation energies [51]:

\[\begin{split}\left(\!\!\begin{array}{ll} \mathbf{A} & \mathbf{B} \\ \mathbf{B}^\ast & \mathbf{A}^\ast \end{array}\!\!\right) \left(\!\!\begin{array}{c} \mathbf{X} \\ \mathbf{Y} \end{array}\!\!\right) = \omega \left(\!\!\begin{array}{lr} \mathbf{1} & \mathbf{0} \\ \mathbf{0} & -\mathbf{1} \end{array}\!\!\right) \left(\!\!\begin{array}{c} \mathbf{X} \\ \mathbf{Y} \end{array}\!\!\right) \;,\end{split}\]

where \(\mathbf{A}\) and \(\mathbf{B}\) are the orbital hessians which also appear in the stability analysis for reference states (see Stability analysis), \(\omega\) is the excitation energy, and \(\mathbf{X}\) and \(\mathbf{Y}\) represent the response of the density matrix. In cases where the system possesses a degenerate ground state or has triplet instabilities, the algorithms used to solve the above equations may be unstable. This can be solved by applying the Tamm-Dancoff approximation (TDA) [52], which simply neglects the \(\mathbf{B}\) and \(\mathbf{Y}\) matrices and leads to a Hermitian eigenvalue problem

\[\mathbf{AX} = \omega\mathbf{X} \;.\]

Methods#

For TDHF or TDDFT calculations, the reference state can be either restricted or unrestricted:

mytd = mol.RKS().run().TDDFT().run()
mytd = mol.UKS().run().TDDFT().run()

By default, only singlet excited states are computed. In order to compute triplet excited states, one needs to set the attribute singlet to False:

mytd.singlet = False
mytd.kernel()

One can also perform symmetry analysis by calling the analyze() method, which also computes the oscillator strengths and dipole moments:

mytd.analyze(verbose=4)

Property calculation#

Oscillator strengths#

Oscillator strengths for each excited state can be computed in both length and velocity gauges:

mytd.oscillator_strength(gauge='length')
mytd.oscillator_strength(gauge='velocity')

Higher order corrections [53] to the oscillator strength can also be included:

#include corrections due to magnetic dipole and electric quadruple
mytd.oscillator_strength(gauge='velocity', order=1)
#also include corrections due to magnetic quadruple and electric octupole
mytd.oscillator_strength(gauge='velocity', order=2)

Transition moments#

PySCF implements various types of transition moments between the reference SCF state and the TDHF or TDDFT excited states. These include:

  • electric dipole, quadrupole and octupole transition moments in both length and velocity gauges:

    mytd.transition_dipole()
mytd.transition_velocity_dipole()
mytd.transition_quadrupole()
mytd.transition_velocity_quadrupole()
mytd.transition_octupole()
mytd.transition_velocity_octupole()

  • magnetic dipole and quadrupole transition moments:

    mytd.transition_magnetic_dipole()
mytd.transition_magnetic_quadrupole()

Nuclear gradients#

Analytic nuclear gradients are available for TDHF and TDDFT, and they can be computed as follows:

tdg = mytd.Gradients()
g1 = tdg.kernel() #default will compute the gradients of first excited state
g1 = tdg.kernel(state=1) #first excited state
g2 = tdg.kernel(state=2) #second excited state

Natural transition orbital analysis#

Natural transition orbitals (NTOs) can be computed by singular value decomposition of the transition density matrix. In PySCF, these orbitals can be obtained as follows:

weights, nto_coeff = mytd.get_nto(state=1)

where nto_coeff are the coefficients for NTOs represented in AO basis, and they are ordered as occupied orbitals followed by virtual orbitals.