Dear SIESTA users,
I'm trying to determine the cohesive energy
of different compounds and I'am faced with the problem
of overbinding using SIESTA.
Just for an example case I tested simple bcc Fe.
I get Ecohes < -10 eV/atom using standard DZP
basis and various pseudos available in the atm dir
(with and without semi core 3d74s1, pseudo core, normal
3d64s2 config). The acceptable value would be Ec > -5. eV/atom.
I have no idea how to improve the results.
It could be that further expansion of the basis set to TZP or sg.
could help a bit, although in published papers I found that
DZP is already good for Fe.
The calculation of Ec is as follows: Ec=E(a0)-2*E_atom
or Ec=E(a0)-E(inf), where E(inf) is sg. like a=30. Ang.
It could be I should play with the parameters
PAO.EnergyShift 50 meV
PAO.SplitNorm 0.15
MeshCutoff 200.0 Ry
Best regards, Peter Sule
----------------------------------------------------
Dr. Peter Sule
senior research associate
Institue for Technical Physics and Materials Science
(MTA-MFA), Konkoly-Thege M. 29-33, Budapest, Hungary
tel.: (36) 1 392 2222/3605
fax.: (36) 1 392 22 73
www.mfa.kfki.hu/~sule / [email protected]
-----------------------------------------------------
My typical fdf file looks like:
SystemName bcc fe
SystemLabel fe
NumberOfAtoms 2
NumberOfSpecies 1
%block ChemicalSpeciesLabel
1 26 Fe
%endblock ChemicalSpeciesLabel
#PAO.BasisType nodes
PAO.BasisSize DZP
#PAO.EnergyShift 50 meV
#PAO.SplitNorm 0.15
#MeshCutoff 200.0 Ry
#SpinPolarized T
MaxSCFIterations 550
DM.MixingWeight 0.01
DM.NumberPulay 4
DM.Tolerance 1.d-4
#DM.UseSaveDM
SolutionMethod diagon #orderN
ElectronicTemperature 3500 K
Diag.ParallelOverK
XC.functional GGA
XC.authors PBE
AtomicCoordinatesFormat Fractional
%block AtomicCoordinatesAndAtomicSpecies
ck AtomicCoordinatesAndAtomicSpecies
0.0 0.0 0.0 1 # Fe 1
0.5 0.5 0.5 1 # Fe 2
%endblock AtomicCoordinatesAndAtomicSpecies
LatticeConstant 2.87
