ABINIT Discussion Forums

Hi everybody

I am trying to calculate elastic properties of CsPbBr3, I do relaxation to the cell using the following input:
ndtset 2 # There are 2 datasets in this calculation

# Set 1 : Internal coordinate optimization

ionmov1 2 # Use BFGS algorithm for structural optimization
ntime1 15 # Maximum number of optimization steps
tolmxf1 1.0e-6 # Optimization is converged when maximum force
# (Hartree/Bohr) is less than this maximum
#natfix1 3 # Fix the position of two symmetry-equivalent atoms
# in doing the structural optimization
#iatfix1 1 2 3
#iatfixx 3
#iatfixy 3
#iatfixz 3
# Set 2 : Lattice parameter relaxation (including re-optimization of
# internal coordinates)

dilatmx2 1.05 # Maximum scaling allowed for lattice parameters
getxred2 -1 # Start with relaxed coordinates from dataset 1
getwfk2 -1 # Start with wave functions from dataset 1
ionmov2 2 # Use BFGS algorithm
ntime2 30 # Maximum number of optimization steps
optcell2 2 # Fully optimize unit cell geometry, keeping symmetry
tolmxf2 1.0e-6 # Convergence limit for forces as above
strfact2 100
#natfix2 3
#iatfix2 1 2 3
#iatfixx2 3
#iatfixy2 3
#iatfixz2 3

#Common input data

#Starting approximation for the unit cell
acell 3*6.017 Angstr

rprim 0.0 0 1
1 0.0 0
0 1 0

#Definition of the atom types and atoms
ntypat 3
znucl 55 82 35
natom 5
typat 1 2 3 3 3

#Starting approximation for atomic positions in REDUCED coordinates
#based on ideal tetrahedral bond angles
xcart 0.0000000000E+00 0.0000000000E+00 0.0000000000E+00
5.6852410708E+00 5.6852410708E+00 5.6852410708E+00
5.6852410708E+00 0.0000000000E+00 5.6852410708E+00
5.6852410708E+00 5.6852410708E+00 0.0000000000E+00
0.0000000000E+00 5.6852410708E+00 5.6852410708E+00

#Gives the number of bands, explicitely (do not take the default)
nband 23

#Definition of the plane wave basis set
ecut 38
ecutsm 0.5
pawecutdg 40
#Definition of the k-point grid
ngkpt 6 6 6
nshiftk 4
shiftk 0.5 0.5 0.5
0.5 0.0 0.0
0.0 0.5 0.0
0.0 0.0 0.5

#Definition of the self-consistency procedure
diemac 10.0 # Model dielectric preconditioner
nstep 50 # Maxiumum number of SCF iterations
tolvrs 1.0d-18 # Strict tolerance on (squared) residual of the
# SCF potential needed for accurate forces and
# stresses in the structural optimization, and
# accurate wave functions in the RF calculations
# enforce calculation of forces at each SCF step
optforces 1



then I calculate ddk eigen values using the following input:

ndtset 3

# Set 1 : Initial self-consistent run

kptopt1 1
tolvrs1 1.0d-18 #need excellent convergence of GS quantities for RF runs

prtcif1 1
# Set 2 : Calculate the ddk wf's - needed for piezoelectric tensor and
# Born effective charges in dataset 3

getwfk2 -1
iscf2 -3 #this option is needed for ddk
kptopt2 2 #use time-reversal symmetry only for k points
nqpt2 1 #one wave vector will be specified
qpt2 0 0 0 #need to specify gamma point
rfelfd2 2 #set for ddk wf's only
rfdir2 1 1 1 #full set of directions needed
tolwfr2 1.0d-20 #only wf convergence can be monitored here

# Set 3 : response-function calculations for all needed perturbations

getddk3 -1
getwfk3 -2
kptopt3 2 #use time-reversal symmetry only for k points
nqpt3 1
qpt3 0 0 0
rfphon3 1 #do atomic displacement perturbation
rfatpol3 1 5 #do for all atoms
rfstrs3 3 #do strain perturbation
rfdir3 1 1 1 #the full set of directions is needed
tolvrs3 1.0d-10 #need reasonable convergence of 1st-order quantities

#Common input data


# acell COPY RELAXED RESULT FROM PREVIOUS CALCULATION
acell 3*6.0047171678 Angstr



rprim 0.0000000000E+00 0.0000000000E+00 1.0000000000E+00
1.0000000000E+00 0.0000000000E+00 0.0000000000E+00
0.0000000000E+00 1.0000000000E+00 0.0000000000E+00

#Definition of the atom types and atoms
ntypat 3
znucl 55 82 35
natom 5
typat 1 2 3 3 3

#Starting approximation for atomic positions in REDUCED coordinates
#based on ideal tetrahedral bond angles

# xred COPY RELAXED RESULT FROM PREVIOUS CALCULATION
# Here is a set of default values, for automatic testing : suppress it and fill the previous line
xcart 0.0000000000E+00 0.0000000000E+00 0.0000000000E+00
5.6736350428E+00 5.6736350428E+00 5.6736350428E+00
5.6736350428E+00 0.0000000000E+00 5.6736350428E+00
5.6736350428E+00 5.6736350428E+00 0.0000000000E+00
0.0000000000E+00 5.6736350428E+00 5.6736350428E+00


#Gives the number of bands, explicitely (do not take the default)
nband 23 # For an insulator (if described correctly as an
# insulator by DFT), conduction bands should not
# be included in response-function calculations

#Definition of the plane wave basis set
ecut 38.0 # Maximum kinetic energy cutoff (Hartree)
ecutsm 0.5 # Smoothing energy needed for lattice paramete
# optimization. This will be retained for
# consistency throughout.

#Definition of the k-point grid
kptopt 1 # Use symmetry and treat only inequivalent points
ngkpt 6 6 6 # 4x4x4 Monkhorst-Pack grid
nshiftk 4
shiftk 0.5 0.5 0.5
0.5 0.0 0.0
0.0 0.5 0.0
0.0 0.0 0.5

#Definition of the self-consistency procedure
diemac 10.0 # Model dielectric preconditioner
nstep 50 # Maxiumum number of SCF iterations
# enforce calculation of forces at each SCF step
optforces 1


The results are not correct and it tells me gamma point is not relaxed.

Any idea please?

Thnaks
Nawzad

What does "the results are not correct" mean? You show many steps and I don't know if you validated the first ones, and where things stop working.

First off, you should converge everything (k, ecut) in particular pawecutdg seems much too small.

nband might also be too small if there are unconverged empty bands which enter in the algorithm - with perturbations they can become important.

For the physics you might have to add van der Waals with Cs and Pb atoms.

Are you sure that the geometry relaxations finished properly? You limit the number of steps a lot.

For the ddk you should not have nqpt 1

Start with that and see if things improve, and you should start with a few more phrases about what you are trying to do...

Dear Matthieu
Thanks for the reply
My results ( Elastic Tensors) do not agree with other available results ...
I have did convergence to ecut, kpt and pawecutdg...
I will increase nband and above input variables...
but optimization finishes without any warning .
Thanks


Nawzad

Hi Nawzad,

what does "gamma point is not relaxed" mean? Do you get something absurd, 10% off, 20%?

Also, first tell us how close you are in lattice and structure.

Some datasets may have "finished" all of their iterations, but if the tolerance is not satisfied the results will be bad.

Are you actually using PAW? The pawecutdg is not present in your last input file.

mverstra wrote:Hi Nawzad,

what does "gamma point is not relaxed" mean? Do you get something absurd, 10% off, 20%?

Also, first tell us how close you are in lattice and structure.

Some datasets may have "finished" all of their iterations, but if the tolerance is not satisfied the results will be bad.

Are you actually using PAW? The pawecutdg is not present in your last input file.

Dear Matthieu
I get the following comment in my output file
"Unstable eigenvalue detected in force constant matrix at Gamma point.
The system under calculation is physically unstable"
my results are more than 20% off.
Literature lattice constant is 6.017 Angstrom, Mine is 6.0047171678

I am using PAW but with Norm-conserving pseudopotentials
Thanks

Nawzad

It is not possible to use PAW with NC potentials. In QE I think this is an option because all of their potentials are in the same format, but it's still unorthodox (evil and wrong) to mix the two. If you give abinit NC potentials in the files file, it will not run PAW. If you give it a mix of paw and NC datasets, it should stop.

If you have an unstable mode, the relaxed elastic tensor will be wrong, so we have to back up a bit. Did you check the convergence of all the DFPT runs as I suggested in the previous post? How much of an instability do you get?

Next, how well did you relax the initial structure (perhaps 30 steps was not enough, and you can check forces and stresses in your initial GS run, before the DFPT, using the copied relaxed coordinates)?

Next, check the eigenvector of the unstable mode - it will tell you which way the system wants to distort - if this breaks symmetry, and you are sure the phase you gave it is the correct GS, then perhaps your pseudopotential is not good enough for this phase, or you are not converged enough in ecut, k, etc... Converging the GS is not enough to guarantee the DFPT will be as well. In passing, your dilatmx and ecutsm should be the same _everywhere_ (relaxation, GS, DFPT) otherwise you can easily get unstable modes. I think this is not the case in your inputs, which will change the basis set of plane waves, and break the relaxation convergence etc... you did previously.

You should read the abinit forum nettiquette viewtopic.php?f=20&t=251 - we need more info in general, on platform, version, outputs etc... Most of my questions could be answered by reading your output directly...

hi
Have you found a solution to the problem?I have the same problem.

Best wishes
jlwindy

Dear all,
A small comment here. Elastic response is very very (I really insist ) sensitive. To be sure your relaxed structure is OK you have to perform a single point energy with the structure and parameters you will use for DFPT calculation and check if forces and stresses are as small as you expected after the relaxation. If you have such a problem of "instability" could you report the output of this "single point energy" calculation of the relaxed structure such that we can see if this is OK?
Best wishes,
Eric