[gpaw-users] Parameters for a plane-wave calculation

[Jens Jørgen Mortensen]

Hi!

The plane-wave basis set that I have been working on for some time now 
is starting to work quite well.  You are welcome to give it a try, but 
there are probably still some bugs.  There is currently only one new 
parameter and that is the cutoff energy for the plane-wave expansion for 
the wave functions (ecut).  This can be set like this:

     from gpaw.wavefunctions.pw import PW
     calc = GPAW(..., mode=PW(250))

or just mode='pw' if you want the default 340 eV.

The density grid can currently be described by the good old "h" 
parameter (or "gpts" parameter), but here it would be more natural to 
use a cutoff energy: ecutdens (=0.5*h^2).  Default value for ecutdens 
could be 2*ecut which corresponds to h=0.235 Å.

With the lcao and finite-difference modes, we are interpolating the 
density to a fine grid with grid-spacing h/2 using tri-quintic 
interpolation.  This fine grid is used for evaluating the xc-energy and 
the compensation charges used for the Coulomb energy.  For the PW mode, 
I'm doing the interpolation in reciprocal space, so the fine grid 
doesn't have to be h/2 - here we could introduce another set of 
ecutdens/h/gpts parameters for those that want to do fine-tuning.

Questions:

* should we introduce an ecut parameter that triggers PW-mode?
* what should we call ecutdens?
* what should we call the parameters for the extra grid (ecutdens2, 
hdens2, and gptsdens2)?

And finally:  Say we want to do an LCAO calculation with interpolation 
and Coulomb energy done in reciprocal space (more efficient for small 
systems) - how do we specify that?  use_reciprocal_space=True?

If anyone out there has some ideas or comments, then I would like to 
hear them.

Jens Jørgen