ABINIT Discussion Forums

Hi forum,

I need help with the calculation of the formation enthalpy of hcp Zn.
With the Zn PSP from the USppPAW table "Zn-gpbe-n-campos-mod" the formation enthalpy of 2.23 eV is too high (1.36 experiental) [ETOT(hcpZn) - 2* ETOT(Zn)]
My question is: Is "occopt 7" the correct choice to calculate hcp zinc and the zinc atom in a big box?
Do I have to consider anything else when calculating the Zn atom in a big box?

Thanks in advance for any comments.

Regards,
Marc

Attached you find my two input files

################
# hcp Zn
################
ndtset 1 jdtset 1

# Set 1 : Internal coordinate optimization
ionmov1 2 # Use BFGS algorithm for structural optimization
ntime1 50 # Maximum number of optimization steps
tolmxf1 1.0e-6 # Optimization is converged when maximum force
tolvrs1 1.0e-15 # Strict tolerance on (squared) residual of the SCF

occopt 7
tsmear 0.005

kptopt 1
ngkpt 3*24
nshiftk 1
shiftk 0.0 0.0 0.0

# Set 2 : Lattice parameter relaxation (including re-optimization of
# internal coordinates)

dilatmx2 1.1 # Maximum scaling allowed for lattice parameters
getwfk2 1 # Start with wave functions from dataset 1
getxred2 1 # Start with reduced coordinates from dataset 1
ionmov2 2 # Use BFGS algorithm
ntime2 120 # 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 # Test convergence of stresses (Hartree/bohr^3) by
tolvrs2 1.0e-15 # Strict tolerance on (squared) residual of the SCF

acell 5.0441007830E+00 5.0441007830E+00 9.2546416431E+00 Bohr
angdeg 90 90 120

#Definition of the atom types and atoms
ntypat 1
znucl 30
natom 2
typat 1 1

xred 1/3 2/3 1/4
2/3 1/3 3/4

#Definition of the plane wave basis set
ecut 24 # Maximum kinetic energy cutoff (Hartree)
ecutsm 0.5 # Smoothing energy needed for lattice paramete
pawecutdg 48.0

#Definition of the self-consistency procedure
nstep 80 # Maxiumum number of SCF iterations
istwfk *1 # Do NOT take advantage of the time-reversal symmetry

#Definition of parallelization

paral_kgb 0 # Parallelize over k-points
wfoptalg 14 # Locally Optimal Block Preconditioned ConjugateGradient
fftalg 401 # FFT routines for MPI
fft_opt_lob 2 # Calls to getghc made parallel on set of bands nbdblock
npkpt 2 # Number of processors for k-point parallelization
npband 1 # Number of processors for band parallization
npfft 1 # Number of processors for fft parallelization
bandpp 1 # Bands per processor
iprcch 6 # Forces corrected with use of Harris functional formula
accesswff 1 # Make wave functions accessible via MPI I/O



For the Zn molecule I use a box with 3*24 Bohr and the same cut-off energy.

##################
# Zn Atom in big box
##################

ionmov 0 # Use BFGS algorithm for structural optimization
ntime 80 # Maximum number of optimization steps
tolmxf 1.0e-6 # Optimization is converged when maximum force
tolvrs 1.0d-16 # Strict tolerance on (squared) residual of the SCF

nband 8
occopt 7
tsmear 0.005

nsppol 2

kptopt 0
nkpt 1
kpt 0.0 0.0 0.0
kptnrm 1 wtk 1

diemac 1.5 diemix 0.2
nline 5 nsym 1

acell 3*24 Bohr
ntypat 1 znucl 30 natom 1 typat 1
xred 0 0 0

ecut 24 # Maximum kinetic energy cutoff (Hartree)
ecutsm 0.5 # Smoothing energy needed for lattice paramete
pawecutdg 48.0

nstep 80 # Maxiumum number of SCF iterations
istwfk *1 # Do NOT take advantage of the time-reversal symmetry

wfoptalg 14 # Locally Optimal Block Preconditioned ConjugateGradient
fftalg 401 # FFT routines for MPI
fft_opt_lob 2 # Calls to getghc made parallel on set of bands nbdblock

My question is: Is "occopt 7" the correct choice to calculate hcp zinc and the zinc atom in a big box?
Do I have to consider anything else when calculating the Zn atom in a big box?

Hello,
Just like most other "pure" calculation parameters (stuff that are not "physical") you have to check convergence of your calculation with respect to occopt choice and tsmear value. Some quantities are hardly sensitive to occopt/tsmear while other are tremendously. If you haven't already done that, you must do it.
For as much as I know, any value of occopt from 3 to 7 would fit your calculation; 3 being the most physical but likely the one with the larger affect on the final output.
If you'd like to see first why it could be important to check convergence with respect to tsmear, have a look at the following exercices and correction :
http://sites.google.com/site/anglade/anglade-abinit-hands-on.pdf.
On page 6, 7 and 8 (figure 3, 4 and 5) you'll find pictures displaying tsmear affect on various aluminum properties computed within Abinit.

Then to answer your two questions I would say : be certain to have extensively checked convergence with respect to non physical parameters :
- energy cutoff
- kpt sampling;
- tsmear;
- pseudopotential;
- XC functional;
- other if applicable.

I can also remember getting very weird results when using a pseudoptentials for Zn that doesn't include semi-core electrons.

regards

Thank you for your fast reply and your link.

I double-checked the convergence again with
- energy cutoff up to 24 Ha
- kpt sampling up to 28 x 28 x 28
- tsmear 0.001 - 0.04
- pseudopotential: USPP and AtomPAW

and for the single atom
- acell up to 30 x 30 x 30

and the cohesive energy was always in the range of E = 2.25 eV for small values of tsmear, or larger for values above tsmear 0.01.
The lattice constants for hcp Zn compare very well with experimental values (less than 1% difference). Thus, I think there might be an issue with my "Zn in a big box" calculation.

Kind regards,
Marc

The following migth not help but I notice that :
- Your second k-mesh is much less converged than the first one (have a look at the values of kptrlen in the output).
- You haven't tried any of the "classical" pseudopotentials. What about testing the HGH for instance.
- If none of the above woks, you could consider tuning the results by choosing an other exchange correlation functional.

Thanks for your hints. I will take a second look on kptrlen.
The last two days I was testing your proposed HGH PSP. The calculated cohesive energy with 2.18 eV is not much better (same input files, but ecut up to 120 Ha). Lattice constants a = 2.666 with c/a = 1.864 are very similar to the PAW USPP ones. (experimental a = 2.665, c/a = 1.856)
It seems to me the problem lies in the Zn single atom calculation.

Regards,
Marc

You were right about the semicore states.
In PRB56,7206:

The conventional LDA approach fails to describe properly the localization of d states which are fully occupied in Zn.

They suggest to downshift the d-states with LDA+U and U=2.0 eV, which solves the problem.

Thanks again,
Marc