[Pw_forum] optimizing pw input parameters

[mbaris at metu.edu.tr]

Dear Meisam Aghtar,

> which parameters should be optimized sequencly in a scf runing? the 
> parameters are celldm(1),ecutwfc, ecutrho,mixing-beta and K-points.

All of them.

> another problem is that for optimizing the ecutwfc the system 
> converges for some numbers and it doesn't for some more. I have to 
> mention that there is overlap in davidson diagonalization.
>

yes, they have such bad manners.




As previously mentioned in the forum, observables such as celldm(1) 
converge much more rapidly than total energy. I do not know about your 
specific problem, but what I do for in the project I am currently 
working on is this: I have written a small script that scans a given 
parameter range (ecutwfc, ecutrho, etc) and minimizes celldm(1) for the 
specific set using a golden search algorithm, then, plotting the 
results (for example celldm(1) vs ecutwfc for different k-point sets), 
I decide which set to use.

why do you want to alter mixing beta? Do you have convergence problems?

and for ecutrho, it is generally safe to use 10xecutwfc,  (remember, I 
do not know what you are working on, and assuming something 
straightforward). Please dig more about how ecutrho is used in the 
code.  and it is better to add your affiliation after a post.

Merry X-mas everyone!

Osman Baris Malcioglu
Ph.D. Candidate
METU Physics
Ankara, Turkey