Relaxed unit cell off by 6%

rx9000 · Post by **rx9000** » Wed Nov 10, 2010 9:22 pm

I am doing a ground state calc using "ionmov 2" to relax the geometry. When I do this, the forces vanish (fcart is 0.00000000d0) for acell about 6% different from what other LDA papers get/experimental value.

I do not think it is due to ecut/nkpt since the forces are zero. Could this be due to a bad choice of pseudopotentials?

Any thoughts are appreciated

rx9000 · Post by **rx9000** » Thu Nov 11, 2010 6:38 pm

I used prtkpt to choose a better k-grid and now it is off by 4%. I suppose I need to run the calculation longer.

blackburn · Post by **blackburn** » Thu Nov 11, 2010 7:26 pm

A zero force is often due to symmetries. Can you post your input so we can see what's happening? 4% is a lot (unless you deal with a fancy system).

rx9000 · Post by **rx9000** » Sat Nov 13, 2010 7:30 pm

The system is a cubic perovskite ... the following code takes about 10 minutes. The SCF iterations converge.

Code: Select all

udtset 1 1
ndtset 1

# Definition of the unit cell
acell 3*7.6707
rprim                           # a cubic unit cell
  1 0 0
  0 1 0
  0 0 1

# Definition of the atom types
ntypat 3
znucl 38 22 8                   # strontium (38), titanium (22), oxygen (8)

# Definition of the atoms
natom 5
typat 1 2 3 3 3
xred
  0.0 0.0 0.0                   # Sr sits at cubic site
  0.5 0.5 0.5                   # Ti sits at body-center site
  0.0 0.5 0.5                   # O sits at face-centered sites
  0.5 0.0 0.5
  0.5 0.5 0.0

ecut1? 14

kptopt 1
ngkpt 4 4 4
nshiftk 1
shiftk 0.5 0.5 0.5

nstep 50
tolvrs 1.0d-10
iscf 5
diemac 6.0

optcell 2
ionmov  2
ntime 40
dilatmx 1.15
ecutsm  0.5
tolmxf?1 5.0d-8

The result is very different from acell=7.37. I have not used DFT codes before, so I don't know if I can expect to do this ground state calculation with better than 5% acell accuracy under a few hours of computer time.

Code: Select all

 Cartesian components of stress tensor (hartree/bohr^3)
  sigma(1 1)=  2.06389860E-10  sigma(3 2)=  0.00000000E+00
  sigma(2 2)=  2.06389891E-10  sigma(3 1)=  0.00000000E+00
  sigma(3 3)=  2.06389888E-10  sigma(2 1)=  0.00000000E+00

            acell    7.8784678429E+00  7.8784678429E+00  7.8784678429E+00 Bohr
           etotal11 -7.3794800643E+01

david.waroquiers · Post by **david.waroquiers** » Mon Nov 15, 2010 10:57 am

Hello,

I would advice you to make a convergence study with respect to the energy cut off for the plane waves, ecut 14 is quite low for norm conserving pseudos (systems with oxygen usually needs at least ecut 30.

David

rx9000 · Post by **rx9000** » Mon Nov 15, 2010 9:49 pm

I found the problem: the pseudopotential I was using isn't good for these titanates, according to the Troullier-Martins PSP README:

http://www.abinit.org/downloads/psp-links/lda_tm_psp1_data

Thanks for the help everyone.

(And I did need to increase ecut to 30 Ha or so)

ABINIT Discussion Forums

Relaxed unit cell off by 6%

Relaxed unit cell off by 6%

Re: Relaxed unit cell off by 6%

Re: Relaxed unit cell off by 6%

Re: Relaxed unit cell off by 6%

Re: Relaxed unit cell off by 6%

Re: Relaxed unit cell off by 6%