We introduce a statistical correlation analysis method to obtain information
on the local geometry and orientation of the tip used in scanning tunneling
microscopy (STM) experiments based on large scale simulations. The key quantity
is the relative brightness correlation of constant-current topographs between
experimental and simulated data. This correlation can be analyzed statistically
for a large number of modeled tip orientations and geometries. Assuming a
stable tip during the STM scans and based on the correlation distribution, it
is possible to determine the tip orientations that are most likely present in
an STM experiment, and exclude other orientations. This is especially important
for substrates such as highly oriented pyrolytic graphite (HOPG) since its STM
contrast is strongly tip dependent, which makes interpretation and comparison
of STM images very challenging. We illustrate the applicability of our method
considering the HOPG surface in combination with tungsten tip models of two
different apex geometries and 18144 different orientations. We calculate
constant-current profiles along the $<1\bar{1}00>$ direction of the HOPG(0001)
surface in the $|V|\le 1$ V bias voltage range, and compare them with
experimental data. We find that a blunt tip model provides better correlation
with the experiment for a wider range of tip orientations and bias voltages
than a sharp tip model. Such a combination of experiments and large scale
simulations opens up the way for obtaining more detailed information on the
structure of the tip apex and more reliable interpretation of STM data in the
view of local tip geometry effects.Comment: Progress in Surface Science, accepted for publication, 25 pages
manuscript, 9 figures, abstract shortene