[Wien] SCF-CONVERGENCE

[Stefaan Cottenier]

> My question is what can be deciphered from these values (CTEST, ETEST).
> Should i be looking for small values of these.

You shouldn't look too much at them. CTEST is the same as :DIS (apart from
the small value which is possible specified after -cc). It tells how much
the charge density changes from iteration to iteration. ETEST tells how much
:ENE changes from iteration to iteration. Ideally, at perfect convergence
neither the charge density nor the total energy should change any more, and
CTEST (:DIS) and ETEST should be zero. In practice, you will never reach
zero exactly, but should reach a 'reasonably small number'. What means
'reasonable' depends on your particular case, and that you have to
experience first by doing several tests as a function of basis set size
(Rkmax) and kmesh for a representative case. In those test, apart from
obtaining the best compromises for kmesh and Rkmax, you can observe how
small :DIS needs to be in order to have the physical quantities you are
interested in sufficiently converged (within one scf).

> My other question is if i want to have a charge convergence of 0.000001(
> microRy), should i be increasing the k points. Looks like for my system,
> looking for a very low convergence value makes the system painstakingly
> slow.

I hope you have a good reason to go to this (almost crazy) precision. For
sure there will be many systems where this is just not possible. Most
likely, you can do with less. Test, as described above.

Stefaan