R*Kmax for different RMTs [Re: [Wien] (no subject)]

[Steven Homolya]

On Tue, 2 Dec 2003, Dr.R.S.Rao wrote:

> Dear Wien users and Steven:
> 	We have tested by keeping RMT*KMAX constant also. There is almost
> same result regarding the relative stability of structures. Very little
> change in terms of energy differences per atom. The problem is that
> relative stability of orthorhombic and AlB_2 type depend on the chosen
> RMT and the energy differences are not negligible : about 5-10 mRy per
> atom. To reach hihger compressions we have to reduce RMT so that plane
> wave cut-off at different compressions remain same.
> 	So how to rely on the results? Generally it is preferable to make
> RMT as large as possible, but we want to go for higher compression and
> keep plane wave cut-off constant.

Dear Ashok,

Make sure calculation is well converged wrt all numerical parameters. I
think most people would be surprised how hard one has to push the
numerical model to get mRy accuracy.

Note that:

Smaller RMT is more accurate, see FAQ. Nevertheless, (my) common sense
says that fine enough spherical muffin tin and Cartesian plane wave grids
(i.e., large NPT + small R0 in case.struct and large R*KMAX in case.in1)
should yield same result. "Fine enough" may be difficult to realise in
practice though. Smoothly "squeezing" (interpolating) plane waves between
almost touching muffin tins is not easy with regular grids...

Energies computed with different RMTs are not directly comparable. As far
as I understand one should choose atomic RMTs that are suitable for all
calculations of interest. Then one needs to make sure about numerical
convergence, and then compare energies. Indirect comparison (i.e., after 
establishing reference levels) is possible but numerically demanding.

If still having problems, send more info (specifics) to list.

cheers,

Steve


-- 
Steven Homolya
School of Physics and Materials Engineering
Monash University, VIC 3800
Australia
Tel: +61 3 9905 3694
Fax: +61 3 9905 3637