[Pw_forum] newbie question: K-points, funtional choise and PP generation/use

[Axel Kohlmeyer]

On Thu, 2011-12-08 at 11:57 +0400, Пермяков Евгений wrote:
> Hi! I'd like to use quantum espresso in research of properties of MoS2
> nanoparticles, modified with Co and Ni. This basically means
> calculations with 20 to 60 atoms of modified MoS2 slab and 10 to 30
> atoms of interacting molecule, as well as 1- and 3-dimensional
> periodic calculations with same number of atoms. The atomic species of
> interest are: Mo, S, Co, Ni, H, C, N, O, K, Na, Li.
> 
> Before performing real calculations I'd like to reproduce geometry of
> several basic with reasonable agreement with expreimental data. The
> problem is choise of PP/functional pair. For some time I was stick
> with b3lyp and pbe functionals of gaussian-03 package with
> full-electron non-periodic calculations, wich resulted in good
> agreement with experimental data. However, I cannot find PPs for this
> functionals for QE and full-electron approach is a little too much for
> my new systems. With quick research I was able to generate simple PP
> for H atom, but got  0.78  A H-H bond length instead of 0.74, and
> runned into incompatibility of ultrasoft PP and Gamma algorithm.  So,
> I decided to stop and search for advise.
> 
> The questions are:
>      1. Can someone point a good review about PP approximation (both
>         math and use)?

try the introductory book by jorge kohanoff
http://www.amazon.com/Electronic-Structure-Calculations-Solids-Molecules/dp/0521815916

>      1. As well as planewave approach for basis sets (It is completly
>         new for me, I used only GTO basis sets before) ?

same recommendation as before.

>      1. And possibly a good library of PP (espetially interested in
>         desription of generation procedure) ?

Q-E has an extensive on-line library of pseudopotentials

>      1. What the hell are K-points, given in K_points sections and how
>         should I choose them for ultrasoft PP ?

some recommendation as for your first question.
i highly recommend to not start any serious calculations
before you understand what k-points are good for.
it will save you a lot of time.

>      1. Are geometry and energy errors in QE/DFT systematic, i.e. can
>         I use some corrections coefficient or I should work with them
>         as random values?

DFT is as good (or bad) in QE as it is in other applications.
however, with a plane wave basis set, you don't have to worry
about basis set superposition errors anymore and the size of
the basis set is effectively controlled by a single parameter
(the plane wave cutoff). with ultra-soft pseudopotentials, you
also need to worry about the density cutoff (for others anything
beyond 4x the wavefunction cutoff is a waste of CPU time and
memory), which depends on how much "softer" the potential has
been made relative to the original norm-conserving potential
that it was derived from.

> Of course, I'm going at least to try to find and read reference
> articles for QE, yet my math skills are not very hight.

that would only be the second step. from your questions it
is obvious, that you first need a general introduction to
solid state physics and electronic structure calculations
in that environment. the kohanoff book is a very good introduction
for people with the kind of background that you seem to
have and thus should serve you well to provide the background
knowledge before you look into the details of the implementation
and special methods.

cheers,
    axel.

> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum