Abstract:
Besides the ab-initio calculation of NMR chemical shifts for supramolecular systems the present dissertation focuses on enhancing the efficiency of newly developed linear-scaling ab-initio quantum-chemical methods. In this context, the large prefactor occurring in large-basis calculations of the linear-scaling AO-MP2 (atomic-orbital Møller-Plesset second-order perturbation theory) method with rigorous integral upper bounds developed in our group could be lowered. This prefactor reduction is achieved via approximation of integrals by means of an auxiliary basis set within the resolution of the identity (RI) method.
The RI approximation was also applied in the context of a modified AO-MP2 method
based on a Cholesky decomposition of the pseudo-density matrices occurring in
AO-MP2 (Cholesky-decomposed density-based MP2, CDD-MP2), which is also linear-scaling and which has, especially for large basis sets, an inherently smaller prefactor than AO-MP2 due to restriction of the transformed indices to the occupied and virtual subspaces, respectively. Here, by employing the RI approximation a further substantial reduction of the prefactor could be achieved. For both AO-MP2 and CDD-MP2 the RI approximation was applied only to the Coulomb-type integral contraction. Even though both developed RI-based methods (RI-AO-MP2 and RI-CDD-MP2) show a reduced prefactor as compared to AO-MP2/CDD-MP2 for basis sets such as SVP and cc-pVTZ, the RI-CDD-MP2 method is clearly superior to RI-AO-MP2, as it is far more efficient and shows an earlier crossover to the conventional fifth-order scaling RI-MP2 method. For compact molecules, RI-CDD-MP2 shows gains of one to two orders of magnitude as compared to linear-scaling AO-MP2. Since there exists yet no rigorous way to restrict the auxiliary basis to significance regions, the RI-AO-MP2 and RI-CDD-MP2 methods developed in this work have the disadvantage that the linear-scaling behavior is lost and a quadratic scaling is observed
for the dominant steps (cubic-scaling steps, which also occur, are not yet dominant for the considered systems). Although this prevents application to very large systems, the scaling-behavior is significantly reduced as compared to conventional RI-MP2 (O(N5)), so that in particular by the RI-CDD-MP2 method large-basis calculations of the correlation energy become feasible for a wide variety of medium-sized to large systems.
The main subject of the second part of the present dissertation is the calculation of 1H NMR chemical shifts for supramolecular systems. As there is no simple relationship between structural parameters and NMR shieldings, quantum chemistry is an important tool for the interpretation of spectra, particularly in the case of more complex systems. In this regard, important contributions to structure estimations could be attained by the calculation of NMR chemical shifts, namely in the context of a study of the hexamer formation of a peptide synthesized in the group of Schrader, and for host-guest complexes with molecular tweezers and clips developed in the groups of Klärner and Schrader, respectively.