Structure
determination and prediction pose a major challenge to computational
material science, demanding efficient global structure search techniques
tailored to identify promising and relevant candidates. A major bottleneck
is the fact that due to the many combinatorial possibilities, there
are too many possible geometries to be sampled exhaustively. Here,
an innovative computational approach to overcome this problem is presented
that explores the potential energy landscape of commensurate organic/inorganic
interfaces where the orientation and conformation of the molecules
in the tightly packed layer is close to a favorable geometry adopted
by isolated molecules on the surface. It is specifically designed
to sample the energetically lowest lying structures, including the
thermodynamic minimum, in order to survey the particularly rich and
intricate polymorphism in such systems. The approach combines a systematic
discretization of the configuration space, which leads to a huge reduction
of the combinatorial possibilities with an efficient exploration of
the potential energy surface inspired by the Basin-Hopping method.
Interfacing the algorithm with first-principles calculations, the
power and efficiency of this approach is demonstrated for the example
of the organic molecule TCNE (tetracyanoethylene) on Au(111). For
the pristine metal surface, the global minimum structure is found
to be at variance with the geometry found by scanning tunneling microscopy.
Rather, our results suggest the presence of surface adatoms or vacancies
that are not imaged in the experiment.