last update October 10, 2010

A major goal of my scientific efforts is to develop new, and hopefully efficient, algorithms for the large scale electronic structure calculation of large systems. Eventually, ab-inito methods need to become powerful enough to handle complex molecular structures of central biological interest. Although having gained some experience with Density Functional Theory in the course of my master's thesis, the most promising approach currently, is the development of a Direct SCF program operating at the Hartree Fock level with parallel features using MPI as well as PVM (GREMLIN).

Reasonable simulation of any biological system will always imply the need for consideration of the effect exerted by the environment (solvation). The aforementioned Direct SCF Hartree Fock program, GREMLIN, can meet this need via inclusion of a multipole moment expansion up to the hexadecapole (SCRF). Alternatively, the Polarizable Continuum Method can be employed where the actual definition of the solute/solvent boundary turns out to be self-evolving (IDAPCM). A major challenge is the systematic incorporation of non-polar contributions in some parameter-free way. Among these are contributions like dispersion and cavitation. The latter effect plays a role in the related phenomenon of hydrophobic hydration.

The success of theoretical scientific work is fundamentally dependent on the computational performance of the underlying program. Attention is paid to proper and comprehensive code design. Standard languages such as C and Fortran77 are used together with their corresponding parallel interfaces to MPI and PVM.

From all the empirical approaches of describing molecular behaviour by far the most fascinating are free energy calculations. This is because of the surprising high level of accuracy they can achieve. Fundamental insight may be gained into complex phenomena of molecular physics and very practical problems can be adressed in the fields of drug discovery and chemical compound design.

Most of today's knowledge of proteins is relevant only to the water-soluble form. Although comprehensible from a technical point of view, the second large domain, ie membrane bound receptors, need all our attention. It is largely this latter class that facilitates a great variety of very crucial functions in the cell. These include cell signaling, signal transduction, exo-transport, gradient- formation and maintainance and many more. While of general interest to almost all biological sciences, this research appears to require a major revision of our current view of the physics covering the membrane interior domain, particularly the hydrophobic environment and the abrupt change of this hydrophobic environment occurring at the interface.

An impressive example of how to use specially designed computer hardware for scientific purposes has been given with the development of MD-GRAPE-2/3. The performance this computer chip offers can hardly be topped by any conventional approach of high performance computing. However, it remains a challenge to get the standard methods, ie Ewald summation techniques, to profit efficiently from this special purpose hardware. One possible solution to circumvent this problem is to use continuum models. Recent investigation into this area has identified Poisson Boltzmann approaches to be particularly suited for MD-GRAPE-2/3. More recently the GPU has become an interesting alternative to ASICs. A first implementation is available under POLCH .