Re: Atoms moving too fast

From: LEWYN LI (
Date: Tue Dec 06 2005 - 13:55:24 CST

Dear Eric,

         How quickly are you cooling down the system? Could you perhaps
cool it down a bit more slowly i.e. using smaller temperature increments?

         I noticed that you are using tcouple to reduce the temperature.
You could perhaps try the other temperature control methods, such as
reassignTemp. I usually use reassignTemp to heat my system up. I don't
know how the command would behave if you used it to cool the system down.

         I recall from your previous e-mail that, before the longer
equilibration, a lot of atoms were moving too fast, with velocities of
-inf. Do you still have a lot of atoms moving too fast after the 1 ns
equilibration? From your e-mail, it seems that, now, the too-fast atoms
are at least moving at some finite velocities. This could be an
encouraging sign.

         Good luck!


On Tue, 6 Dec 2005, Eric Perim wrote:

> Hello,
> I posted a message about having problems with my simulation, and
> followed the advice given by Lewyn of doing a longer simulation. I
> just finished doing an 1000000 timestep (1ns) equilibration (constant
> temperature and pressure) of my system, but still I get an error
> message saying the atoms are moving too fast when I try to start
> another simulation, this one to cool the system (using tcouple). The
> end of the log file is as follows:
> ENERGY: 980000 5007.0883 3041.1113 146.1419
> 14.6181 -59797.6788 6167.0269 9451988.1543
> 0.0000 13539.6654 9420106.1274 310.1190
> 9420679.0896 9420679.0896 310.1190 -3082206.2226
> -3081720.6492 142397.7388 -3082206.2226 -3081720.6492
> ERROR: Atom 6478 velocity is -401.956 -4511.72 16669.9 (limit is 10000)
> ERROR: Atoms moving too fast; simulation has become unstable.
> ERROR: Exiting prematurely.
> ==========================================
> WallClock: 16.802896 CPUTime: 10.117187 Memory: 7665 kB
> I'd like to know if anyone know how to solve this, since I can't
> carry on my research without doing these simulations.
> By the way does anyone know where can I get a routine for
> calculating the pair distribution of the system after simulation?
> Thanks in advance,
> Eric

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:24 CST