[CPMD-list] "SURFACE" keyword
Salomon Billeter
ZRLSRB at ch.ibm.com
Tue May 18 09:17:31 CEST 2004
Hi Aloysius,
the default for isolated systems is that the system is centered in the cell
(just a parallel shift of all atoms). This is done to ensure that the
density at the borders of your cell is as small as possible (the Poisson
solvers decoupling the periodic images electrostatically from each other
require at least this). This applies also to 2D and 1D periodic systems. If
you don't want to have the atoms shifted, you can specify
CENTER MOLECULE OFF
in the &CPMD section. There are good reasons to use the SURFACE keyword for
systems which are periodic in two dimensions only because that way, they
don't "see" their periodic images along the Z-direction. There are however
a few points to consider:
- In the current version, the (default) Poisson solver for 2D and 1D
periodic systems is Hockney, which does not work in these cases. Use the
Mortensen PS instead (POISSON MORTENSEN in the &SYSTEM section).
- For the Mortensen Poisson solver, you need to give more space in the
Z-direction than for Hockney to make it decouple the periodic images
electrostatically. A good reference for box sizes of isolated systems (in
your case just along the Z-direction) is J. Hutter's posting of Aug. 17,
2002 in this list.
- The soon-to-be-released version of CPMD sets the default to Mortensen,
and a Poisson solver with the box requirements of Hockney is implemented,
too. This can save you CPU time and memory considerably.
Best regards,
Salomon
More information about the CPMD-list
mailing list