Local treatment of electron excitations in the EOM-CCSD method

Abstract

The Equation-of-Motion coupled cluster method restricted to single and double excitations (EOM-CCSD) and singlet excited states is formulated in a basis of nonorthogonal local orbitals. In the calculation of excited states only electron promotions from localized molecular orbitals into subspaces (excitation domains) of the local basis are allowed, which strongly reduces the number of EOM-CCSD amplitudes to be optimized. Furthermore, double excitations are neglected unless the excitation domains of the corresponding localized occupied orbitals are close to each other. Unlike in the local methods for the ground state, the excitation domains cannot be simply restricted to the atomic orbitals that are spatially close to the localized occupied orbitals. In the present paper the choice of the excitation domains is based on the analysis of wave functions computed by more approximate (and cheaper) methods like, e.g., configuration-interaction singles. The effect of various local approximations is investigated in detail, and it is found that a balanced description of the local configuration spaces describing the ground and excited states is essential to obtain accurate results. Using a single set of parameters for a given basis set, test calculations with the local EOM-CCSD method were performed for 14 molecules and 49 electronically excited states. The excitation energies computed by the local EOM-CCSD method reproduce the conventional EOM-CCSD excitation energies with an average error of 0.06 eV.

Ultraviolet photodissociation of OCS: Product energy and angular distributions

“…The TDM for C is also increased yielding a larger absorption cross section compared to the previous calculations. It is interesting to note that the new TDMs, which are reasonably converged with respect to the number of states averaged over in the CASSCF calculations, agree satisfactorily with the results obtained with the equations-of-motion coupled-cluster method 29 restricted to single and double excitations (EOM-CCSD). 30 The EOM-CCSD approach usually gives good agreement with full CI calculations when the excited states are adequately described by promotion of a single electron from the ground state and the ground state itself is well described by a single-reference calculation.…”

Local CC2 response method based on the Laplace transform: Orbital-relaxed first-order properties for excited states

“…48 for the LT-LMP2 method the Lagrangians in Eqs. (24) and (37) are not the proper energy Lagrangians, if the Laplace transformation is employed, but just approximations to it, while the proper LT-DF-LCC2 Lagrangians are impractical (cf. Eq.…”

“…To specify such restrictions a priori is rather straightforward for calculations of the ground state amplitudes, but more intricate for eigenvectors of excited states. 30,32,37,38 For example, a certain eigenvector may want to change in the course of the iterative Davidson diagonalization and acquire Rydberg or charge transfer character, if the virtual space allows for it.…”

Linear-Response Time-Dependent Embedded Mean-Field Theory

“…(13). (ai|bj) and (ai|f xc (r, r ′ )|bj) correspond to the two-electron repulsion integrals and the xc integrals, respectively,…”

“…For example, linear-scaling implementations based on linear response take advantage of spatial locality of either the atomic [5][6][7] or molecular [8][9][10][11][12][13] orbitals. An alternative strategy employs subsystem embedding to describe localized excitations, including TDDFT implementations using either fragment molecular orbitals 14 or frozen-density embedding, [15][16][17][18][19][20] as well as the QM/MM approach.…”