Bad convergance of 111 surface of Ni FCC for 16 and 20 atoms

Message

smanos · #1 Post by **smanos** » Thu Nov 29, 2012 7:23 pm

Hi everybody

I am running static calculations of Pt and Ni 100, 110, and 100 surfaces using the PBE pseudopotentials provided by VASP with lattice parameters between 3.4-3.6 Angstroms using slabs of 8,16 and 20 atoms in order to calculate the total energy.

The problem I am having is that the energy of the 111 surface of Ni for the slabs of 16 and 20 atoms do not converge around the expected lattice parameter but instead seem to take almost random values(noise?) of total energy. The results from Pt seem to be as expected as well as the 8 atoms slab of Ni.

Good result example 110 surface of Ni

Lat.Par 8 at. Ev/atom 16 at. eV/atom 20 at. eV/atom
3.40 -5.128783125 -5.311644188 -5.3479240
3.44 -5.150519125 -5.345969125 -5.3849910
3.48 -5.157258125 -5.362994313 -5.4041565
3.52 -5.149944625 -5.364136313 -5.4070085 MINIMUM
3.56 -5.129965750 -5.351039562 -5.3952765
3.60 -5.099693375 -5.326171438 -5.3714810
3.64 -5.060201000 -5.290791438 -5.3368875

Bad result from 111surface of Ni 16 and 20 atoms
111surf. 8 at. Ev/atom 16 at. eV/atom 20 at. eV/atom
3.4 -5.359081875 -5.381255813 -5.258384
3.44 -5.397001250 -5.321948813 -5.3276285
3.48 -5.416319125 -5.422114188 -5.4124735 MINIMUM
3.52 -5.418698500 -5.570956500 -5.298763
3.56 -5.407347375 -5.471981563 -5.3990275 MINIMUM
3.6 -5.384022250 -5.424294060 -5.3961860
3.64 -5.349900000 -5.340784250 -5.4047125

I have tried to change certain parameters like PREC =High Sigma =0.05, Increase ENCUT to 500 eV and Increase the number of NBADS but there was no improvement on the convergence.

Can anyone make a suggestion why this is happening?
All input files are the same for Ni and Pt except POTCAR and INCAR because for Ni I need to include magnetism.

Below are the INCAR and KPOINTS files that I used.

Thank you.
KPOINTS file

Automatic mesh generation
0 ! number of k-points = 0 ->automatic generation scheme
Gamma !Monkhorst
19 19 1

INCAR FILE

SYSTEM = [COMPUTATION-ID:4661; STRUCTURE-ID:29481; COMPOSITION:Au1]
#
############# THE BASIC VALUES DETERMINING THE ACCURACY
ENCUT = 300 #eV
ENAUG = 544.6 #eV
PREC = NORMAL
LREAL = .FALSE. # FALSE for unit cells with > ~20 atoms
##################TYPE OF CALCULATION
##################
IBRION = 2
IALGO = 68
ISPIN = 2
ISIF = 2
NSW = 0
ISYM = 2
################
SIGMA = 0.10
POTIM = 0.4
SYMPREC = 10E-5
########VARIABLES
MAGMOM = 20*0.822
NBANDS = 125
###################
###################CONVERGENCE AT
EDIFFG = 1.0000e-04
EDIFF = 1.0000e-06
#################
################# OUTPUT FILES TO BE WRITEN
LELF = .FALSE.
LCHARG = .FALSE.
LWAVE = .FALSE.
#################
################# Cluster-depending values
NPAR = 1

<span class='smallblacktext'>[ Edited ]</span>

#2 Post by **admin** » Mon Dec 03, 2012 1:27 pm

the lattice parameters have to be calculated for the BULK unit cell, this
parameter has to be used for the surface calculation (directions along
the surface normal) furtheron only the interlayer distances should be relaxed

smanos · #3 Post by **smanos** » Tue Dec 04, 2012 1:41 pm

I have made bulk calculations and I obtain reasonable results for the lattice parameters and the bulk modulus. I do not mind relaxing selectively the layers because I want to use these results to fit Embedded atom method potentials. I perform static calculations on the surfaces to increase the number of surface configurations that I can use for the fitting.

Also I repeated these calculations for Ir, Au, Pd, Cu and Co. I obtain reasonable results for 111 surface from all except Ni and Co so I observe at this point that this problem is related to magnetic materials ?

Thank you for replying to my post.

<span class='smallblacktext'>[ Edited Tue Dec 04 2012, 01:54PM ]</span>