Classical molecular dynamics (MD) simulations have been performed to study atomic processes related to the reordering at interfaces and the relaxation of nanostructures (cf. ). However, to yield macroscopic relevant and predictive results, suitable interatomic potentials are necessary, developed on ab initio based approximations (cf, ). The bond order potential based on the tight binding model is used to enhance the MD, as it preserves the essential quantum nature of atomic bonding, yet abandons the electronic degree of freedom. Tight binding and ab initio methods require complete diagonalizations of the Hamiltonian, which scale with 3rd order of the number of particles included in the simulations and restrict MD to a few thousand atoms. The analytic BOP, however, achieves linear scaling and is recognized as a fast and accurate model for atomic interactions (cf. ).
In addition, the electronic structure, which is not included in simple empirical potentials, thus e.g. not applicable in image calculations for TEM and electron holography, is described in the bond order potentials to mimic the electronic interactions correctly. Thus it allows to contribute to solve the open problem in image matching related to the charge density and the frozen lattices (cf.  ). Furthermore, good scattering potentials and a reconstruction of electron exit waves in electron holography are the presuppositions to solve the inverse problem of electron diffraction. The retrieval of the local parameter variations enables a reliable interpretation of electron micrographs and is the basis for a general direct object reconstruction (cf.   ).
The Figure characterizes the principle of object retrieval starting from an electron hologram, using single reflex reconstruction out of the Fourier transform and solving the inverse problem of scattering /8/.
The Figure shows the spontaneous Si(100) surface dimerization and sigma-π bond exchange (indicated by different colors) with the BOP4+ after 1ns annealing at 600 K maximum temperature.