My Community

Dear all,
In one of my recent MD simulations, the cell was found to drift all the time, even though at the beginning of each continued run the warning was given as "The initial velocities result in a center of mass drift but there must be no drift. The drift will be removed !'' Actually I have ever come to this problem before, only for some systems. I am using PWA-PBE potentials, with the INCAR file given as below. It seems the cell is driven drift due to some external force? But where does it come from? How to remove this cell drift? Any suggestions are welcomed. Thanks.

Lucy

INCAR:
PREC = medium
ALGO = fast
LREAL = Auto
ISTART = 1

#LWAVE =.F.
#LCHARG =.F.

IBRION = 0
NSW = 5000
ISIF = 0
POTIM = 2
SMASS = 0
#NBLOCK = 50
TEBEG = 900.0
TEEND = 900.0
ISYM =0

ISMEAR = 0

You may need to check timestep (POTIM) and also the convergence of atomic force. More accurate force during MD simulation can be achieved by increasing NELMIN (say, NELMIN=8).

Thanks, wolyim.
I have ever reduced timestep by setting POTIM to 2, 1, and 0.5 fs, but it won't help.
Maybe I should try to increase NELMIN. But could you tell me what is the critieron for the convergence of atomic force? The following is the part of the force information from one instant configuration.

Thanks.

POSITION TOTAL-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
6.99103 9.66573 7.43038 -1.148521 -3.622050 0.843642
6.59678 9.51917 11.38591 1.810247 0.353440 -0.410854
10.50864 1.88823 3.34410 -0.953976 -2.299889 1.564243
10.70807 1.78926 7.55391 0.185263 -1.511023 0.119872
10.58190 1.79821 11.23499 0.331939 -2.173235 -0.202865
10.97758 5.57323 3.38793 -1.176639 -0.439821 0.160908
10.72765 5.64620 7.54205 0.748391 -1.501335 0.275057
10.88056 5.60225 11.40676 -0.556360 0.250096 0.524774
11.11747 9.46349 3.66961 -1.352851 -0.187506 -2.103097
10.59968 9.58054 7.35285 1.581977 -1.542414 -0.102466
10.90169 9.43785 11.37441 0.339839 -0.134274 0.944831
1.28121 11.35778 2.22192 -0.527608 0.892877 -1.646577
1.06871 11.47784 5.54932 -0.480802 0.455663 0.432853
0.90840 11.32191 9.39851 0.332695 0.176562 -0.353672
0.63391 3.39283 1.31700 1.522101 0.393793 1.384042
1.14721 3.61947 5.42862 -0.457561 -0.248378 0.365340
1.17575 3.42826 9.30778 0.011935 0.480890 0.076570
1.21948 7.41374 1.67818 0.061300 0.822482 0.031637
1.09259 7.54345 5.39081 -0.183585 0.154566 0.591487
1.14322 7.57316 9.54556 -0.180668 0.173029 -0.496458
5.12510 11.39635 1.81626 -0.703063 0.458697 -0.375202
5.14955 11.33297 5.36467 -0.394405 1.117086 0.781789
5.05186 11.46743 9.60694 -1.324422 -0.241921 -1.128875
4.72472 3.63783 1.84153 0.880502 -0.055939 -0.686070
5.03895 3.51603 5.55214 -0.105100 -0.013110 1.047708
4.98316 3.77932 9.48414 -0.895875 0.495468 -0.562593
4.58497 7.51763 1.91938 0.146731 0.519342 -0.713156
4.75050 7.50297 5.38288 1.149150 0.520741 0.670980
5.01212 7.72406 9.60105 0.200593 -0.661263 -0.340675
8.76706 11.57983 1.30065 0.081384 0.181771 0.893147
8.99783 11.08043 5.47699 -0.207508 1.112651 -0.445684
9.01032 11.17822 9.58831 -0.496560 0.769249 -0.902653
8.71927 4.15548 1.74090 -0.035018 -1.681101 -0.849624
8.87295 3.96441 5.87782 0.028021 -0.954097 -1.162093
8.30447 3.91957 9.45356 0.585236 -0.046221 0.385023
8.87278 7.83380 1.77737 0.153508 -0.234537 -0.289855
8.95644 7.57569 5.44218 0.335861 0.486264 -0.111746
9.00458 7.47562 9.63610 -0.838967 0.357001 -0.147792
-----------------------------------------------------------------------------------
total drift: -0.140560 -0.024625 -0.037210
d Force =-0.1695157E+00[-0.294E+00,-0.454E-01] d Energy =-0.1684081E+00-0.111E-02
d Force =-0.1968216E+01[-0.271E+01,-0.122E+01] d Ewald =-0.1968147E+01-0.692E-04

There are numerical errors in computing the total energy and the force. If the errors between the energy and force are inconsistent, a drift in the center of mass will be observed. Typically, the quality of the total energy is better, because the total energy converges faster both in terms of the energy cut-off and the self-consistent iterations. There are some discussions on molecular dynamics in the VASP manual.

You can also judge the quality of your simulation by examining the energy conservation. I don't think it is necessary to eliminate the drift if it doesn't affect the result of what you want to study.

Hello dear,

From your INCAR I can see that you are using the ALGO = Fast. Thus, you are required to set the IALGO = 48 too.

Hope it will be helpful!