Theoretical background, parametrization, and performance of the semiempirical configuration interaction singles (CIS) method MSINDO-sCIS designed for the calculation of optical spectra of large organic molecules are presented. The CIS Hamiltonian is modified by scaling of the Coulomb and exchange integrals and a semiempirical correction. For a recently proposed benchmark set of 28 medium-sized organic molecules, vertical excitation energies for singlet and triplet states are calculated and statistically evaluated. A full reparameterization of the MSINDO method for both ground and excited state properties was necessary. The results of the reparameterized MSINDO-sCIS method are compared to the currently best semiempirical method for excited states, OM3-CISDTQ, and to other standard methods, such as MNDO and INDO/S. The mean absolute deviation with respect to the theoretical best estimates (TBEs) for MSINDO-sCIS is 0.44 eV, comparable to the OM3 method but significantly smaller than for INDO/S. The computational effort is strongly reduced compared to OM3-CISDTQ and OM3-MRCISD, since only single excitations are taken into account. Higher excitations are implicitly included by parametrization and an empirical correction term. By application of the Davidson-Liu block diagonalization method, high computational efficiency is achieved. Furthermore, it is demonstrated that the MSINDO-sCIS method correctly describes charge-transfer (CT) states that represent a problem for time-dependent density functional theory (TD-DFT) methods.

The semiempirical SCF MO method MSINDO (modified symmetrically orthogonalized intermediate neglect of differential overlap) [T. Bredow and K. Jug, Electronic Encyclopedia of Computational Chemistry, 2004] is extended to the calculation of excited state properties through implementation of the configuration interaction singles (CIS) approach. MSINDO allows the calculation of periodic systems via the cyclic cluster model (CCM) [T. Bredow et al., J. Comput. Chem., 2001, 22, 89] which is a direct-space approach and therefore can be in principle combined with all molecular quantum-chemical techniques. The CIS equations are solved for a cluster with periodic boundary conditions using the Davidson-Liu iterative block diagonalization approach. As a proof-of-principle, MSINDO-CCM-CIS is applied for the calculation of optical spectra of ZnO and TiO(2), oxygen-defective rutile, and F-centers in NaCl. The calculated spectra are compared to available experimental and theoretical literature data. After re-adjustment of the empirical parameters the quantitative agreement with experiment is satisfactory. The present approximate approach is one of the first examples of a quantum-chemical methodology for solids where excited states are correctly described as n-electron state functions. After careful benchmark testing it will allow calculation of photophysical and photochemical processes relevant to materials science and catalysis.

Analytical expressions for the sCIS (scaled configuration interaction singles) and UCIS (unrestricted CIS) energy gradients are presented for the semiempirical method MSINDO. The theoretical background of the derivation of the analytical gradients is presented, and the implementation into the MSINDO program package is described. The computational efficiency of the underlying Z-vector method is greatly enhanced by making use of the transpose-free quasiminimal residual (TFQMR) algorithm. Benchmark timing tests are compared to the widely used TD-B3LYP approach. For a statistical evaluation of the accuracy of MSINDO-sCIS, geometry optimizations are performed for a small set of organic molecules in selected excited states. The obtained results are compared to CASPT2 and TD-B3LYP/TZVP. In order to demonstrate the applicability of the present approach to periodic systems within the cyclic cluster model, we present first calculations of the excited state structure of ethyne adsorbed on the NaCl (100) surface.

Despite the fact that the complexation of ammonium cations with ionophores like crown ethers plays an important role in biological and industrial processes, there is still a lack of theoretical methods to reproduce or even predict the host-guest complex structures or their thermodynamic stabilities in an accurate manner. Hence, the development of ionophores has often relied on a trial-and-error approach and the synthetic efforts associated with this have been enormous, so far. Therefore, theoretical methods for the reliable prediction of binding affinities of crown ether derivatives with ammonium ions would be an indispensable tool for the rational design of new receptors with tailored properties. Here, we suggest a computationally efficient but still accurate theoretical approach. It is tested for a model system consisting of 18-crown-6 ether and an ammonium cation, but is invented for application to much larger complexes. The accuracy of various approximate quantum-chemical methods, based on density functional theory (DFT) and many-body perturbation theory, is evaluated against the gold standard CCSD(T) in the basis set limit as internal reference. An important aspect is the consideration of dispersion interactions in DFT methods, for which the dispersion-correction by Grimme was employed. For all selected methods, the basis-set dependence of calculated interaction energies was investigated.