[gmx-users] (no subject)

Xavier Periole x.periole at chem.rug.nl
Tue Nov 9 12:52:54 CET 2004

Hector Mrz-Seara Monne wrote:

> hello, the system I'm working on seems to have the problem of 
> presenting strong forces during the dynamics that finish in the 
> explode of the system. Just to clarify  what Xavier 
> (x.periole at chem.rug.nl) wrote, how is possible to have a different 
> behavior, even a slightly different route, if you start for o point of 
> the dynamics that goes thougth the same path that does that has 
> explode out your system. Do you regenerate the velocities or something 
> similar? or is more like playing with intrinsic round errors of the 
> computer?
You got different solutions:
1) you generate new velocities to give to a conf before the exploxion 
point. It that case
the simulation can not be considered as the continuation of previous 
one. It is a different
simulation. But it will be correlated to the previous one because you 
used a conformation
taken out fron the trajectory.
2) you can use trjconv -f traj.trr -dump XXXX to extract the 
conformation and velocities
of the simulation at the time XXXX. create a new topol.tpr using those 
velovities. The
loss precision and the caotic behaviour of the propagation of your 
system will give you
a slightly different route and the system may not explode. You can check 
the deviation of
the two simulation by choosing a time long enought before the explotion 
and plot any
value (rmsd, energies, ...) and see that the two simulations diverge.
3) you can use the tpbconv -f traj.trr to extract the last frame of the 
traj.trr and resend your
run. You'd observe the same behavior of the point 2).

May be other solutions are possible ...



   Xavier Periole - Ph.D.

   Dept. of Biophysical Chemistry / MD Group 
   Univ. of Groningen
   Nijenborgh 4
   9747 AG Groningen
   The Netherlands

   Tel: +31-503634329
   Fax: +31-503634800
   email: x.periole at chem.rug.nl
   web-page: http://md.chem.rug.nl/~periole

More information about the gromacs.org_gmx-users mailing list