[CPMD-list] Charged Surface

[Salomon Billeter]

Dear Eung-Gun, dear Axel,

may I extend a little bit:

- 2-D periodic systems indeed better should have no net charges. If there
is a net charge, the Mortensen Poisson solver must be used (see below), and
the total energy should be corrected for the interactions with the
neutralizing background charge
- A 2-D periodic Poisson solver does not take much care of cancelling the
dipole moment resulting from two opposite charged defects on either
surface: it only does so between images along the (Z-) axis; the periodic
images within the (XY-) plane do see each other, and the two charges on
either surface definitely cause an electric field across the capacitor
formed

Here are the specifics about the Poisson solvers for 2-D periodic systems.
You have two options:

- Mortensen (the default): the normal conditions apply (i.e. the box size
along the Z-direction should be twice the range of the non-zero charge
distribution). A neutralizing background charge is added if there is a net
charge
- "Hockney": this one is not a real Hockney solver because the influence
function is prepared in reciprocal space (using expressions derived from
2-D Mortensen), but you get away with the smaller Hockney requirement for
your box along the Z-axis (i.e., the charge distribution must be zero at
the top and the bottom of your box). In case of a non-zero net charge, the
resulting total energy has no simple reference

Best regards,
Salomon