Poor electronic convergence in MD study with VASP6.3

[yuke_li2]

Hi Sir,

Now I try to use molecular dynamics (MD) with VASP.6.3.0 to study the Cu/water interface.
At the 55th step in MD, the electronic step doesn’t converge. But if I use vasp.5.4.4, I don’t encounter this problem.
Enclosed please find the my Inputfiles and outfiles for VASP.6.3.0.

###PS
I always (around ten cases, with differnce initial structure) meet electronic convergence problem with VASP6 in MD. But It's fine for the VASP.5.4.4.
Now I have to use VASP.5.4.4 to study MD.
####

Best,
Li

[andreas.singraber]

Hello Li,

sorry for the late reply!! The underlying algorithm behind ALGO=VeryFast was changed in version 6. Hence the difference in the convergence for your system. I am not sure why the newer version is less robust (it was of course changed to do the opposite). However, you can turn on the "old" VASP 5 flavor of the algorithm even in VASP 6 by setting

Code: Select all

ALGO=Old VeryFast

in your INCAR file (see here).

Best,
Andreas Singraber

[andreas.singraber]

Hello again,

I had another look at your simulation. Although your VASP 5 run does seem to run smoother at first glance, the convergence is actually bad in both cases. There are a lot of remarks of this kind:

Code: Select all

Information: wavefunction orthogonal band   17  0.7966

which indicates that you should check again the electronic convergence. Also, if you check energy conservation during your MD run you will find very bad results! Please have another look at your INCAR setup, see e.g. these old posts:
forum/viewtopic.php?t=4966#p4975
forum/viewtopic.php?t=3561

Best,
Andreas Singraber

P.S. I moved this topic here from the "Bugreports" section because there is no indication of a bug.