[CPMD-list] Discrepancy between MORTENSEN and TUCKERMAN

[Juerg Hutter]

Hi

some words on the Poisson solver for isolated systems.
There are three solvers for isolated systems implemented
in CPMD: Hockney, Tuckerman, Mortensen

The Hockney solver is the most thorough method, but it's
not working with all options (especially linear response).
Requirements are that the molecule is in the middle of the box
and that the electron density falls off to zero within the box.
The 1/r interaction is correctly included for the full box.

The Tuckerman solver works in Fourier space alone using a
numerical approximation to 1/r. The maximum extend is L/2
(L is the box length). To be exact this requires that
L is twice the extend of the charge distribution of the
molecule. With oder words the box should be twice the
size of the Hockney box. In real application this
requirement is relaxed and a smaller box can be used
with good accuracy.

Finally the Mortensen solver is just an analytic solver
in Fourier space for spherical boundary conditions.
The same restrictions as for the Tuckerman solver hold.
In addition, using smaller boxes is restricted by the
fact that the 1/r term works only within a sphere inscribed
within the box.

Comming back to the problem described in the original posting
I would guess that the box size was large enough for
the Tuckerman solver but not for the Mortensen solver.

regards

Juerg Hutter

----------------------------------------------------------
Juerg Hutter                   Phone : ++41 1 635 4491
Physical Chemistry Institute   FAX   : ++41 1 635 6838
University of Zurich           E-mail: hutter at pci.unizh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
----------------------------------------------------------