Can I trust this calculation?

[Tyx]

Hello all,

I thought that if EDIFF is set to something, (e.g. 10^-7), then
unless specified differently, EDIFFG is 10*EDIFF (e.g. 10^-6).
Which in an ionic relaxation would mean that it would stop
when dE < 10^-6.
I have a 16-atom calculation which says it has converged, with
IBRION=2, ISIF=4 and NSW=100, at only 4 ionic steps.
It is a BCC structure, and ISYM=2. My NBANDS is also
sufficiently high.
The thing is that dE=10^-1 only:
1 F= -.51345366E+02 E0= -.51347518E+02 d E =-.513454E+02 mag= 1.8481
2 F= -.50555593E+02 E0= -.50560989E+02 d E =0.789773E+00 mag= 0.0001
3 F= -.51462877E+02 E0= -.51474414E+02 d E =-.117511E+00 mag= -0.0011
4 F= -.51463011E+02 E0= -.51474619E+02 d E =-.117645E+00 mag= 0.0020

The forces are all practically zero, and the drifts too, and it
stopped without errors, saying that the convergeance has
been reached.
The cell has not changed shape, the atoms have moved
slightly. It also seems that changing from ISIF=4 to ISIF=2
makes no difference to the result whatsoever. Using
IALGO=38 (Normal).


Why did it stop? Shouldn't it move around until dE<10^-5?
Can I trust this result?

Thanks!


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

[tjf]

[quote author=4 to ISIF=2
makes no difference to the result whatsoever.
[/quote]


Does symmetry allow the shape to change?

[Tyx]

I don't know, but shouldn't it say something? At a larger volume, the same calculation gives dE=10^-2, in 3-4 ionic steps.
It is two species 3 atoms of one, 13 of the other, placed on a BCC lattice. And at the end, everything looks fantastic apart from dE...as with the smaller volume.
<span class='smallblacktext'>[ Edited Fri May 19 2006, 03:15PM ]</span>