Dear Roberto,
Maybe what you are seeing is the rippling of the energy/forces due to
the aliasing (wrap around errors) of the fft.
The orbitals (squared) and neutral atom potential are projected into
the grid so if you change them this error can introduce
"noise" in the convergence of the total energy/forces. The neutral
atom potential of carbon usually is very hard in reciprocal
space, so you have to increase a lot the mesh cutoff. You could try
with a mesh cutoff which is twice as big as the
one you are using right now.
Best,
Eduardo
On 15/05/2008, at 15:04, Roberto Sapiens wrote:
Hi:
I've been doing some calculations on graphene ribbons and got
interested in knowing how the total energy converges as the energy
shift varies. I had an assumption that the smaller the energy shift
the lower the total energy, which I confirmed doing such a
calculation for small molecules and isolated atoms. On the other
hand, I found, for the ribbon (a 1D periodic system), that the
minimum for the total energy occurs for an energy shift around 0.05
eV. Why such a difference?
With my best regards,
Roberto