Dresden 2006 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe
CPP: Chemische Physik und Polymerphysik
CPP 29: Computational Techniques
CPP 29.1: Vortrag
Freitag, 31. März 2006, 10:30–10:45, ZEU 160
First-principles calculation of noncovalent interactions - Diffusion Monte Carlo applied to hydrogen bonding and aromatic stacking — •M. Fuchs1, C. Filippi2, J. Ireta1, and M. Scheffler1 — 1Fritz-Haber-Institut der MPG, Berlin — 2Universiteit Leiden (NL)
Van der Waals (dispersion) forces are important in many molecular phenomena, such as self-assembly of molecular crystals and peptide folding. Calculating this nonlocal correlation effect requires accurate electronic structure methods. Usual density-functional theory (DFT) with generalized gradient functionals fails unless empirical corrections are added that still need extensive validation. Quantum chemical approaches like MP2 or coupled cluster (CCSD(T)) are accurate, yet their unfavorable computational scaling limits them to rather small systems. Diffusion Monte Carlo (DMC) can provide accurate molecular total energies and remains feasible also for larger systems. Hence DMC promises to be useful for benchmarking simpler but more approximate approaches. Here we apply the fixed-node DMC method to (bio-)molecular model systems where dispersion forces are significant. Our DMC binding energies for (dimethyl-) formamide dimers and adenine-thymine DNA base pairs fully agree with data from CCSD(T), in particular for stacked DNA base pairs where MP2 is insufficient. In addition we consider benzene dimers in the stacked, slipped parallel, and T-shaped structures. Resolving their relative energies requires carefully optimized DMC trial wavefunctions. We discuss the role of the determinantal part and the correlation factor, optimizing them simultaneously using a fluctuation potential approach.