Vibration calculation - Increased computer time

[forsdan]

Dear VASP-users,

I'm performing calculations with the IBRION=5 tag and Selective dynamics in order to obtain the force constants. I start with a relaxed system and set "T T T" after each atomic position in the POSCAR file. Then I start the calculation and everything seems to run smoothly.

Here's when the problem comes into the picture. Due to that I only have a finite amount of time on the cluster available only (say 10%) of the atoms will be displaced in the different directions. Fine, no problem there but when I wish to continue to displace the other atoms I thought it would be possible to set "F F F" after those atoms that had already been displaced and run another calculation. Well, it works but the analysis of the constrained symmetry for selective dynamics will result in a different space symmetry group and the number of irreducible k-points will be larger than in the first calculation. This results in a significantly increased computer time.

My question is now if the best solution to this problem is simply to keep "T T T" after each atom but to rearrange the POSCAR-file so that it instead will be begin with those atoms who haven't been displaced yet. In my opinion this will give the same space symmetry group as in the first calculation and the same time consumption. Do you agree ?

Best regards

Dan Fors

[tjf]

Nope. If the order of the atoms in POSCAR effects the determined spacegroup then there's a serious bug.

Read also what the manual says about symmetry under IBRION=5.

[forsdan]

Thank for your reply but please let me rephrase my question with a simple example. Let us say that the first POSCAR looks like

bcc:
2.831000000000000
1.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 1.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 1.0000000000000000
2
Selective dynamics
Direct
0.0 0.0 0.0 T T T
0.5 0.5 0.5 T T T

and that only the first atom had time to be displaced in the three directions. Then my first thought was to rewrite the POSCAR-file into

bcc:
2.831000000000000
1.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 1.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 1.0000000000000000
2
Selective dynamics
Direct
0.0 0.0 0.0 F F F
0.5 0.5 0.5 T T T

and run a second calculation. In my case this resulted in a different space group. So I wondered if I rewrite the POSCAR into

bcc:
2.831000000000000
1.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 1.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 1.0000000000000000
2
Selective dynamics
Direct
0.5 0.5 0.5 T T T
0.0 0.0 0.0 T T T

this would solve the problem and give the same space group as the first POSCAR. If I understand you correctly it will solve the problem unless there is a serious bug in the program ?

Best regards

Dan Fors
<span class='smallblacktext'>[ Edited Thu Aug 10 2006, 10:12AM ]</span>

[tjf]

OK, I see. You're running into a time limit, right?

In this case, I'd recommend a 'suck it and see' approach.

I'd guess that what's happening with the different space groups is that this is for the displaced geometry. Moving different atoms around for the finite differences breaks the symmetry of the base geometry in different ways. That's why the manual says you should switch off symmetry all together, or else you're "comparing" numbers from different parameter sets. I think you'll find that your second way of doing the second set of derivatives it will result in the same symmetry as the selective dynamics way.

So in other words: The symmetry you're seeing isn't the BCC symmetry of the basic material, but the lower symmetry of the lattice with a discplaced atom. This you cannot change.



<span class='smallblacktext'>[ Edited Thu Aug 10 2006, 07:10PM ]</span>