[Wien] Convergence check

[Stefaan Cottenier]

The equilibrium lattice parameter is a *property* of the system you
study, while RKM and kmesh are *input* parameters that determine the
numerical accuracy of your calculation (only the numerical accuracy, the
'physical accuracy' is determined by the XC-functional that you use).

A procedure that is not unique, but quite safe and that does not involve
more work than strictly needed:

* For a reasonable guessed lattice constant and a default RKM (7.0),
check as many properties as you can as a function of kmesh (starting at
4-10 k-points, and always doubling the mesh). Properties are :ENE,
:MMIxxx, :FORxxx, :EFGxxx, ...

* For the lowest kmesh that gives you the 'same' values as for your
higest kmesh, check the properties again, but now as a function of RKM
(start at 6 and go in steps of 0.5 up to 8.5 or 9.0). Only for :ENE you
cannot do this, as increasing the basis set will lower the energy always.

* If you insist on having verified total energies, repeat the second
step for a lattice parameter that is 0.1 au different from your guessed
one, and plot the *difference* of these total energies as a function of
RKM.

* At this point, you know now the optimal k-mesh and RKM. Now you can
safely vary the lattice parameter to find an E(V) curve, elastic
properties, etc.

Stefaan

Hui Wang wrote:
> Dear wien2k master:
>     I have 2 problems about convergence check, please see below: 
>     (1) If there is crystal structure, we don't know its equilibrium 
> lattice constant, *I am wondering which quantity should be determined 
> first, such as: equilibrium lattice constant, RKM, kpoints ? *
> my understanding:
> ===================
> (1)use all the default value and a reasonable kmesh to determin the 
> equilibrium lattice constant first, then test kpoints in this 
> equilibrium lattice constant to get sufficient kpoints, and last use the 
> tested kpoints and this equilibrium lattice constant to test RKM, to 
> find a good RKM. Then to get the physical or mechanical properties.
> (2)use the experimental lattice constant first, test kpoints and then 
> test RKM. Last, use the tested kpoints and RKM to optimize this 
> structure to determin the theoritical equilibrium lattice constant. then 
> to get the physical or mechanical properties.
> ===================
> *Correct me if i am wrong? or show me the sequence of how to do these 
> convergence check ? Thanks.*
>  
>     (2)*Besides the RKM, kpoints, are there some other parameters which 
> are very important or very sensitive to the system during the process of 
> total energy convergence check, especially for magnetic system ?*
> ** 
> * *   Thanks in advance:)
>  
> yours sincerely:
> Hui
> 
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien


-- 
Stefaan Cottenier
Center for Molecular Modeling (CMM)
Ghent University
Technologiepark 903 (2nd floor)
BE-9052 Zwijnaarde
Belgium

http://molmod.Ugent.be
email: Stefaan . Cottenier /at/ UGent . be