Seminar by Prof. Pavel Hobza
Academy of Sciences of the Czech Republic, Prague (Czech Republic)
- 27.Aug.2014 11:00
- 27.Aug.2014 12:00
- Lecture room 2009, Jülich GRS building (16.15)
Calculations of noncovalent interactions belong to the most difficult part of computational chemistry. Only the most advanced methods covering large portion of correlation energy provide reliable stabilization energies and geometries of various types of noncovalent complexes. Special position plays the coupled cluster CCSD(T) approach (covering single-, double- and triple excitations) which stands as the "gold standard" method of computational chemistry because of its outstanding accuracy for determination of various properties of noncovalent complexes. Accuracy of the method was tested by performing calculations of stabilization energy of various complexes at higher theoretical level, CCSDT, CCSDTQ, CCSDTQP and even full CI.
Highly accurate calculations going beyond the CCSD(T)/complete basis set (CBS) limit were used for determination of structure and binding energy (D0) of HF dimer. The calculated D0 agree excellently with the experimental value and the largest error originates in evaluation of ZPV energy. Harmonic approach (at the CCSD(T)/CBS level) provides substantial error and inclusion of anharmonicity is thus inevitable. Similar approach was used for determination of structure and stabilization energy of more than 20 small molecular clusters for which the experimental stabilization energy (D0) is known and also here the agreement between theory and experiment was fair. On the other hand, a significant difference between two independent experimental and theoretical D0 values was found for the anisol dimer which is the prototype of π - π stacked dimer. A possible source of this disagreement is discussed.
Databases based on CCSD(T) characteristics (e.g. S22, S66, L7, X40 and A24 databases from our laboratory) are used for design, verification and parametrization of newly developed computational techniques allowing to treat larger complexes. Special position is played by MP2.5/CBS method and semiempirical quantum mechanical method, both developed in our laboratory. The former technique, providing highly accurate characteristics for complexes having about 100 atoms was applied to guest - host complexes. The latter method is applicable to extended biomacromolecular complexes having more than 10,000 atoms. The PM6-D3H4X method provides very good estimates of binding energies for protein - ligand complexes stabilized by different noncovalent motifs like hydrogen or halogen bonding. The method is the base of our scoring function used in in silico drug design.