[gmx-users] Re: timestep in CG simulation

Xavier Periole x.periole at chem.rug.nl
Fri Oct 29 10:38:53 CEST 2004

Ashish Gupta wrote:

>>>  The suggested value of the timestep is 0.05 ps. With this value of 
>>> timestep the simulation
>>> seem to be running fine for some steps (some thousands of steps) and 
>>> then I get segmentation
>>> fault error. Why is this so ?  Also if I reduce the timestep to 
>>> 0.005 ps then the simulations run fine.
>> Your starting structure is probably allright (since you can run 
>> 1000's of
>> steps). Were there any error/warning messages before the crash? Have you
>> looked at changes in the energies and/or the structure? You could try to
>> re-start short before the crash, write out more frames (lower nstxtcout)
>> and look at them.
> Thanks a ton Anton for the reply .To remind others again, I am running 
> Coarse Grain
> simulation ( http://md.chem.rug.nl/~marrink/coarsegrain.html) and am 
> trying to figure out
> why a timestep of .05 ps gives me Segmentation Fault after about 48000 
> steps while
> a timestep of .005 ps does not. 

That is normal ... You can try also with full atomic description. A too 
large time step
does not allow you to describe the evolution of the system correctly and 
it results in
generating bad conformations, which most likely ends by bad contacts ...

>   I tried with different timesteps and found that the system crashes 
> sooner if I increase my
> timestep to 0.055 ps or 0.06 ps      I checked the energies in the 
> system . What I found is the energy of my system shoots
> up in one or two steps.  It is the kinetic energy which shoots up 
> first and then the system
> crashes. Any idea why this fast change in energy occurs ( in one or 
> two steps when the
> system has been happily running for some initial steps. )
>     I checked the velocities in the last saved frame and found that 
> the velocity of some beads
> shoots very high . But I have no idea why the system does so.
>      I am using a rlist of  1.3 nm and rvdw is 1.2 nm . The system 
> crashed even when I
> update my neighbour list every step. I am using Berendsen temperature 
> coupling with
> time constant 1 ps.

We generaly use 0.04 ps time step. Some times you have to equilibrate 
with lower time steps
and increase it gradualy. This depends on the nature of the system you 



   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