ABINIT Discussion Forums

Dear colleagues,

I'm trying to relax the a and b dimensions of the NiAs type structure (Phys. Rev. B 33, 4221 (1986)). Atoms are fixed by symmetry at 2a and 2c positions. However, I just cannot make it through - the relaxation always stops because the a and b change too much, i.e. dilatmx is too small (I tried up to 1.5). This is strange for it has been done before with DFT (ex. Comp. Mater. Sci. 37, 349 (2006); or J Comput Aided Mater Des 8, 193 (2001)). Perhaps there is some simple error in the input file, which I'm missing.
I have tried different options for tolerances, ionmov, diemac, ecuts, meshes, pseudos (Teter LDA and TM LDA), iscf, nnsclo, nline, npulayit.

Any help and suggestions would be welcome.
Thank you in advance!

Yours

Igor Lukacevic

Are you sure the structure you input is correct and you calculation parameters (kpt and ecut) are converged enough ?

Hi!

Of course, I forgot to attach the input file. I'm attaching it now. I'm not 100% sure of the structure - I am led by the experimental paper on it (Phys. Rev. B 33, 4221 (1986)): hexagonal cell; NiAs type (spgroup 194, P6_3/mmc, c/a=1.75, N=2, Wyckoffs - oxygens (0,0,0) and (0,0,1/2), bariums (1/2,2/3,1/4) and (2/3,1/3,3/4); at 13.9 GPa - a=3.617 A, c=6.349 A).

Concerning ecut and ngkpt: I don't know...I never managed to finish the calculation with even the smallest ones (the highest ones I tried were ecut 40 and ngkpt 4 4 8).

Thanks for helping me out!

Igor

Hello,

I think that you inverted parameters c and a in your input file :
acell 2*11.998 6.835 has to be changed to acell 2*6.835 11.998
So you started from a structure very far from the equilibrium, that's normal abinit cannot manage to relax without complaining.

You also mistyped the reduced coordinates -- (1/2,2/3,1/4) instead of (1/3,2/3,1/4) for the first oxygen atom -- in the message below but it is correctly typed in your input file so this is no matter.

David

Dear David,

as Gandalf would say (at the gates of Khazad Dum):

"Gandalf, you old fool!"

thnx
Cheers!

Igor