Page 1 of 1

Different results with the same input of energy calculation

Posted: Mon Nov 13, 2017 8:49 am
by mhl980
Hi VASP users,

I would like to calculate the adsorption energy of Na3PO4 on Fe2O3 surface.
However, when I calculate the total energy of Fe2O3 surface only, the results are different as different runs with the same input. My model contains 60 atoms (36 O, 24 Fe) with the vacuum of 15 Angstrom in z direction.
First of all, 1 model extracted from Material Studio has been conducted the ionic relaxation (K-point 111). Then an energy calculation has been applied for optimized CONTCAR configurations. Every calculation has considered spin polarization and has run on 64 CPUs of supercomputer.

The INCAR file for ionic relaxation:
System = Fe2O3_relax
PREC = Normal
ENCUT = 400
LREAL = Auto
ISMEAR = 1
SIGMA = 0.1
NELMIN = 5
#EDIFF = default 1E-4
#EDIFFG = default
NSW = 500
IBRION = 2
NPAR = 8
NELM = 500
ISPIN = 2
ICHARG = 2
LWAVE = F
LCHARG = F
#POTIM = default
LORBIT = 2
IDIPOL = 3
#MAGMOM = default

The INCAR file for energy calculation:
System = Fe2O3_energy
PREC = Normal
ENCUT = 400
LREAL = Auto
ISMEAR = -5
SIGMA = 0.1
NELMIN = 5
#EDIFF = default 1E-4
#EDIFFG = default
NSW = 0
IBRION = -1
NPAR = 8
NELM = 500
ISPIN = 2
ICHARG = 2
LWAVE = F
LCHARG = F
#POTIM = default
LORBIT = 2
IDIPOL = 3
#MAGMOM = default
ALGO = Fast

I ran these settings 4-5 times and got different total energy of Fe2O3 surface. I also changed K-point, ALGO, PREC, to have the comparison. The calculations have been shown below:
ALGO ENCUT PREC ALGO ENCUT PREC Maximum difference
Relax 111 Normal 400 Normal Energy 991 Fast 400 Normal -428.01 -429.14 -429.26 -429.24 1.25 eV
Energy 991 Normal 400 Normal -428.34 -428.42 0.08 eV
Relax 331 Normal 400 Normal Energy 991 Fast 400 Normal -428.27 -429.75 -429.48 -427.47 -429.63 2.28 eV
Energy 991 Fast 400 Accurate -429.46 -428.88 0.58 eV
Energy 991 Normal 400 Normal -429.46 -428.41 -428.40 1.06 eV

The unit is eV. The results in the same row have the same setting. 111, 331, 991 are K-Point mesh settings. The difference is very high, sometimes it reaches 2.28eV.

I would appreciate any help.
Regards,
Ha Le

Re: Different results with the same input of energy calculat

Posted: Mon Nov 13, 2017 2:56 pm
by mhl980
Sorry that the table cannot display clearly. I will post table again:

I ran these settings 4-5 times and got different total energy of Fe2O3 surface. I also changed K-point, ALGO, PREC, to have the comparison. The calculations have been shown below:

Relax 111, ALGO Normal, ENCUT400, PREC Normal --> Energy 991, ALGO Fast, ----ENCUT 400, PREC Normal: --result run 1 -428.01 eV, run 2 -429.14 eV, run 3 -429.26 eV, run 4 -429.24 eV; ---------------------maximum different 1.25 eV
Relax 111, ALGO Normal, ENCUT400, PREC Normal --> Energy 991, ALGO Normal, ENCUT 400, PREC Normal: --result run 1 -428.34 eV, run 2 -428.42 eV; -----------------------------------------------------------------maximum different 0.08 eV
Relax 331, ALGO Normal, ENCUT400, PREC Normal --> Energy 991, ALGO Fast, ----ENCUT 400, PREC Normal: --result run 1 -428.27 eV, run 2 -429.75 eV, run3 -429.48 eV, run 4 -427.47 eV, run 5 -429.63 eV; maximum different 2.28 eV
Relax 331, ALGO Normal, ENCUT400, PREC Normal --> Energy 991, ALGO Fast, ----ENCUT 400, PREC Accurate: result run 1 -429.46 eV, run 2 -428.88 eV; -----------------------------------------------------------------maximum different 0.58 eV
Relax 331, ALGO Normal, ENCUT400, PREC Normal --> Energy 991, ALGO Normal, ENCUT 400, PREC Normal: --result run 1 -429.46 eV, run 2 -428.41 eV, run 3 -428.40 eV;--------------------------------------------maximum different 1.06 eV