Re: Simulation at different time-steps

From: Aron Broom (broomsday_at_gmail.com)
Date: Tue Jul 16 2013 - 23:47:17 CDT

In theory there isn't anything wrong with having such a mixed trajectory,
in fact in some simulations one may intentionally use a variable timestep
to accommodate changing temperatures for instance. But even if you restart
from long before the error, the fact that such an instability occurred
after such a long time (rather than very soon as a result of some initial
bad contacts for instance) makes one wonder if there could have been an
accumulation of problems due to the timestep, in which case it would be
hard to say where exactly things went wrong.

In terms of your subsequent simulations, you may not need to go all the way
down to 1.0fs, something like 1.5fs is probably more than enough to save
you from future errors.

As far as the trajectory in VMD goes, it's probably simplest if you just
resave the 1fs part with a 1 frame skip. You can do this in VMD itself or
using catDCD which has been mentioned a lot recently.

On Tue, Jul 16, 2013 at 11:02 PM, Mare Libero <mareliber27_at_yahoo.com> wrote:

> Hello everybody,
>
> I have a 2fs time-step simulation that crashed after 14.2ns with the
> error message:
>
> ERROR: Constraint failure in RATTLE algorithm for atom 4944!
> ERROR: Constraint failure; simulation has become unstable.
>
> The simulation instability and the RATTLE algorithm error are
> extensively documented on this mailing list.
>
> In my case, repeating the simulation at 1fs fixed the problem. I was
> wondering if there is anything theoretically wrong in restarting the
> simulation from the point it crashed with a shorter time-step. I understand
> that continuing the same simulation with different time-steps will create a
> heterogeneous trajectory with different time increments per frame. But I
> cannot think of anything theoretically wrong in doing so. Am I missing
> anything?
>
> Also, since I am now trying to compare trajectories collected with
> different time-steps, is there a way to visualize trajectories in VMD as a
> function of the time rater than frame number (I know this question may be
> more appropriate for the VMD mailing list).
>
> Thanks in advance for any input,
>
> Mar
>
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:26 CST