Semilocal DFT at the GGA level underestimates
long-range dispersion interactions. 
This problem causes a bad overestimation of 
graphite lattice in the stacking direction
(8.84 A (PBE) vs. 6.71 A (exp)).

In this example, dispersion correction method 
of Tchatchenko and Scheffler (PRL 102, 073005 
(2009)) is used to cope with this problem.

Optimal length of the lattice vector c normal to 
the stacking direction is determined in a series
of single point calculations with varied value 
of c (all other degrees of freedom are fixed at 
their experimental values). 

The computed c vs. energy dependence 
is written in the file results.dat and can be 
visualized e.g. using xmgrace. The optimal 
value can be obtained using the attached 
utility (python with numpy or Numeric is needed):

./utilities/fit.py results.dat 

200 iterations performed
Ch-square: 4.30305519481e-09
---------

        E0(eV):         -37.433456779
        d0(A):  6.65603352689

The computed value of 6.66 A agrees well with 
experiment (6.71 A).

Details of implementation of TS in VASP + a number 
of tests:
Bucko et al., PRB 87, 064110 (2013).

