Single Molecule Molecular Dynamics

[Lucas Lodeiro]

Hi all,

I want to perform single molecule molecular dynamics.

First I do a convergence test for cutoff, cell size and poisson solver. I read in old mails, that MT is better than WAVELET for MD, due to the first is translational invariant.

After that I start a MD from the optimized molecular geometry at the same level of theory. I want to obtain statistical information for a given temperature (300K for example), I think an NVT is not well behaved for a single molecule, then I ran NVE MD. As the geometry is at the minimum, I start with 2*T (where T is the target temperature) thinking in virial theorem (at equilibrium equal potential and kinetic energy).

But the system is not converging (the average temperature) at the desired temperature. It shows a weird temperature oscillation (different as expected due to bond oscillation), it present in some moment higher temperature than the initial and a potential energy lower than the initial... Also, the molecular rotation exhibits variations along the MD and translation in the box. For that I start a new MD with the same INPUT and geometry, adding ANGVEL_ZERO. The MD starts without rotation but in 300 fs it starts to rotate.

I attach the files for both MD.

I read about the geometrical constraints, it would be useful for diatomic molecules, but for H2O or complexer molecules is not a solution.

Has somebody faced these problems? some advice?

regards

[Lucas Lodeiro]

[Thomas Kühne]

Dear Lucas, 

have you conducted the geometry optimization with exactly the same settings, 

i.e. basis set, PW cutoff, EPS_SCF, etc.? Since your system has so few DOFs, 

fluctuations will be large and conservation by cancellation inherently poor. 

Hence, with such a tiny system I would rather stick with NVT instead of NVE + 

viral theorem and not use ANGVEL_ZERO … 

Greetings, 

Thomas

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4P%2BMnqg3JM5dGm-%3DQzSD912dx1A1qo8shYaMgdNLgFzTqA%40mail.gmail.com.

[Lucas Lodeiro]

Thanks Thomas for the advice about the MD.

Yes the geometry optimization was with the same settings, using the default convergence criteria. I attach the input and output.

There is a little energy difference (1 E-5) between the geo_opt run (optimized configuration) and the zero step in the MD, due to the centering positions keyword, I understand that is due to non-continuous fourier grid. But the potential energy minima along the MD is 1 E-4 lower than the geo_opt geometry minima... and I cannot understand why.

Regards

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/404EE40E-BC28-4A32-8B53-6513F7D47F8C%40gmail.com.