Events
Time: 4:30 p.m.
Place: UHW 11/1102
Volker Blum, Fritz-Haber-Institut der Max-Planck-Gesellschaft, Berlin, Germany
Quantum-mechanical first principles (e.g., density-functional theory) provide us with
an enormously successful, essentially accurate computational framework for the
potential energy surface that governs materials or molecular properties, chemical
reactions etc. One of the ongoing challenges is that we can reach benchmark-level
accuracy for small systems (few-atom molecules) relatively easily, but pushing the
same benchmark accuracy out to large systems -- when it matters – is an ongoing
challenge.
This talk focuses on two areas where high accuracy for relatively large structure
sizes (hundreds or thousands of atoms) is indeed desirable, if it can be had: (i) The
structure, stability and dynamics in biomolecular systems, where relatively weak
interactions (e.g., Hydrogen bonds, van der Waals) make a critical difference, and
(ii) surface reconstruction at a larger scale, where the physically relevant structure
of an interface is determined by relatively small (per atom) energy contributions.
We use all-electron electronic structure theory as implemented in the FHI-aims code
to meet these challenges. Numeric atom-centered basis sets provide essentially
converged numerical accuracy for the task at hand, and efficient parallelization up
to massively parallel architectures (hundreds or thousands of processors) allows us
to reach the relevant system sizes. In particular, we address the possible bottleneck
of an algebraic eigenvalue solver (Kohn-Sham equations) on massively parallel
machines. On the biomolecular side, we focus on polyalanine molecules (5-20
aminoacids) large enough to form secondary or tertiary structure, where accurate
vacuum experiments (IR spectroscopy) are available for quantitative comparisons to
experiment. Regarding surface reconstruction, we address the structure and
thermodynamic stability of commensurate, graphene-like
reconstructions that form on SiC(111).
