[gmx-users] Energy drift

[Mark Abraham]

On Mon, Aug 25, 2014 at 6:16 PM, Xiaobin Cao <xcao at lsu.edu> wrote:

> Thank you, Mark. Belows are what I found:
>
> 1, I test the double precision version. There is no drift with NVE, while
> the single one has.
>

OK, that suggests that the range of the magnitudes of the forces might be
too large for single precision. I'd still like to know how small the time
step needs to be to work with single precision (because if it's the range
that is too large, then a really small time step won't help).

2, Howerver, the bond energy drift still exists with NVT. I don't know why.
> 3, For NVE simulation, the temperature is lower than what I generate. For
> example, the average temperature is aroud 273K if I generate the
> temperature at 298.15K.
>

Is that with single or double? Is this model known to produce a good
temperature? You should also be asking the original authors about how they
generated their results.


> 4, the diffusion coefficient calculated by NVE with 1 fs time step is ten
> times smaller than the experimental result at 298.15K, but the result is
> improved with 0.5 fs time step.
> 5, I prefer to use NVT, because I can control the temperature. However, if
> the bond energy still drifts to high overtime, this will reduce rotation
> and translation energy because the kinetic energy and temperature are
> stable. If so, the calculated diffusion coefficient will be reduced, and
> this is consistent with my calculated results. Do yo have any idea to fix
> the bond energy drift in NVT ensemble?
>

Not without more data :-) The model may not work at all!

Mark


> By the way, my studied system is gas phase, which has a longer correlation
> time (about 2 ns).
>
> Thank you.
>
> -----------------------------------------------------------------------------------------
> >Hi,
>
> >Repeating: how small a step is small enough to suppress the drift?
>
> >Mark
>
>
> _____________________________________________
> Xiaobin Cao
> Dept of Geology & Geophysics
> E235, Howe-Russell Geosciences Complex,
> Louisiana State University, Baton Rouge, LA 70803, U.S.A.
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>