ABINIT Discussion Forums

Hi,
I want to optimize the supercell organization and only relax the surface atoms. But it has a error, and this is my .in file :
occopt 4
tsmear 0.04
acell 4.0345 4.0345 18.0345 Angstrom
angdeg 90.0000 90.0000 90.0000
ntypat 3
znucl 56 22 8
natom 7
typat 1 1 2 3 3 3 3
xcart 0 0 0
0 0 0.2237
0.5 0.5 0.11185
0.5 0 0.11185
0 0.5 0.11185
0.5 0.5 0
0.5 0.5 0.2237
chksymbreak 0
natfix 5
iatfix 1 3 4 5 6
ecut 13
ixc 14
pawecutdg 50
optcell 0
ionmov 3
ntime 500
tolmxf 6.0d-4
nstep 100
tolvrs 1.0d-20
iscf 17
npulayit 20
ngkpt 6 6 1
nshiftk 2
shiftk 0.5 0.0 0.0
0.0 0.5 0.0
the error is :
fixsym: ERROR -
Atom number 2 is symmetrically equivalent to atom number 1,
but according to iatfix, iatfixx, iatfixy and iatfixz, they
are not fixed along the same directions, which is forbidden.
Action : modify either the symmetry or iatfix(x,y,z) and resubmit.

I donot think the atom 1 and 2 are in the same place, because the c of the cell is 18.03 A, the atom 2 is not in the cell boundary. Can you tell why ?

When I remove these two lines :
natfix 5
iatfix 1 3 4 5 6
and I restart the file ,it also has an error:
chkpawovlp : ERROR -
PAW SPHERES ARE OVERLAPPING !
Distance between atoms 1 and 3 is : 0.71590
PAW radius of the sphere around atom 1 is: 2.52098
PAW radius of the sphere around atom 3 is: 1.82908
This leads to a (voluminal) overlap ratio of 99.95 %

I use the paw pseudopotentials which are transformed by the ultrasoft pseudopotentials , used the sofeware of USpp2Abinit.
This is my first time to use the paw pseudopotentials , and I donot konw what I should notice. Can someone tell me where is wrong?
Thanks~

Hello,

I see two problems in your input file. The first one is related to the paw overlap error. You give coordinates of the atoms with xcart but it seems that its reduced coordinates (am I right ?). So instead of xcart (positions of atoms in cartesian coordinates), you should use xred (positions of atoms in reduced coordinates with respect to unit cell vectors). The second problem is your fixing error. What abinit is telling you is that atoms 1 and 2 are equivalent by symmetry (this does NOT mean that they are in the same place !). Indeed you the z=0.11185 plane which is a mirror plane so these two atoms should be moved together (and they will, even if you use chksymbreak 0, because by symmetry, their forces will be the same). Also, for a surface, you might need more layers.

David

Dear David,
Firstly, you are right, the positions of atoms are in reduced coordinates and it is my mistake. And I have modified it to xangst .
For the second error, I use 5 layers as a surface instead of the previous 3 layers. And the system is symmetric still. If I want to
relax the surface atoms and fix the others in the symmetric system, what should I set ? Thanks~

Hello,

Well increasing the number of layers will not change the symmetry of your system but in fact you have 2 surfaces and not just one and each surface is symmetry to the other one in your case (remember you are using periodic boundary conditions ...). So what you can do is fix the intern layers while letting the surface atoms relax. For your 3 layers example, this would be :

natfix 3
iatfix 3 4 5

If you really want to fix only one surface (but this might be more difficult to analyse though ... because even if you fix one of them, you still have a surface !), I think that setting the variable nsym to 1 will do the job (and moving very slightly your atoms so that they are not any more in a symmetric position), but I do not recommend you this option.

David

Dear David,
Thank you very much, and I will take your advice to fix the intern layers.

Regards.