Benchmark of Electronically Excited States for Semiempirical Methods: MNDO, AM1, PM3, OM1, OM2, OM3, INDO/S, and INDO/S2

Abstract

Semiempirical configuration interaction (CI) calculations with eight different Hamiltonians are reported for a recently proposed benchmark set of 28 medium-sized organic molecules. Vertical excitation energies and one-electron properties are computed using the same geometries as in our previous ab initio benchmark study on electronically excited states. The CI calculations for the standard methods (MNDO, AM1, PM3) and for the orthogonalization-corrected methods (OM1, OM2, OM3) include single, double, triple, and quadruple excitations (CISDTQ) using the graphical unitary group approach (GUGA) as implemented in the MNDO code. The CIS calculations for the established INDO/S method and the reparametrized INDO/S2 variant employ a modified version of the ZINDO program. As compared to the best theoretical reference data from the ab initio benchmark, all currently applied semiempirical methods tend to underestimate the vertical excitation energies, but the errors are much larger in the case of the standard methods (MNDO, AM1, PM3). Overall, the mean absolute deviations relative to the theoretical best estimates are lowest for OM3, and only slightly higher for OM1 and OM2 (in the range of 0.4-0.5 eV). INDO/S performs similar to OM2 for the vertical excitation energies of singlet states, but deteriorates considerably for triplet states. The INDO/S2 reparametrization for oxygen improves the results for low-lying singlet states of oxygen-containing compounds, but makes them worse for high-lying singlets as well as for triplets. The ab initio reference data for oscillator strengths and excited-state dipole moments are again best reproduced by the orthogonalization-corrected approaches (OM1, OM2, OM3), which thus emerge as the most favorable semiempirical methods overall for treating valence excited states of large organic chromophores.

Assessment of approximate computational methods for conical intersections and branching plane vectors in organic molecules

“…The OM2 variant is the most complete model of the OMx family and offers a good compromise between accuracy and computational efficiency, especially for excited states. 89 It includes valence-shell orthogonalization corrections both in the one-center part 87 and in the two-center part of the core Hamiltonian. 88,90 The MOs obtained from the OM2 calculation are used as input in a subsequent GUGA-MRCI calculation.…”

On the applicability of time‐dependent density functional theory (TDDFT) and semiempirical methods to the computation of excited‐state potential energy surfaces of perylene‐based dye‐aggregates

“…For the computation of excited states the OMx methods devised by Thiel and coworkers have been proven to be especially successful, which can be attributed to the improved orbital energy gaps, due to the orthogonalization corrections. [21] Thus, they are superior to most other methods, which use the underlying MNDOscheme. They provide rather accurate results and are very costefficient.…”

Electronic excitations in long polyenes revisited

“…Because the description of the diatomic overlap has been greatly improved in the OM2 valence shell Hamiltonian, which has been developed and carefully evaluated by Thiel and co-workers for a series of different dye molecules and various levels of CI, 9,10 we expect that our following OM2/MRCI treatment can properly account for the excitation energies and oscillator strengths of the covalent and ionic excitations in the polyenes. Figure 2 illustrates the early computational results 7,11 obtained by PPP-MRCI calculations on the vertical excitation energies E i (N) of the low-lying singlet states S i in regularly alternating polyenes with N = 4, 6, .…”