[gmx-users] 2 fs time step??

[Erik Lindahl]

It depends on the temperature of your system, not the number of steps. 
Normally the fastest motions are bond involving hydrogens, and after 
that bonds between non-hydrogen atoms.

If you constrain bonds 2fs should be perfectly safe. There's no reason 
to go to 1fs (and you gain almost a factor 2 in speed).

A lot of people use 2fs even when they only constrain bonds involving 
hydrogens, and that also seems ok. I've done some informal tests, and 
when using lincs you can actually often go to 3fs steps, but I cannot 
recommend other people to do that - yet.

Neighborsearching should ideally be done at the same _time_ intervals 
(not step intervals), so you should decrease nstlist for longer 
timesteps.

Cheers,

Erik


On Mar 31, 2004, at 6:35 PM, Vivek Raut wrote:

> Hi,
>
> I am running a simulation for 10 ns.
>
> How much of a difference does it make to keep the  time step (dt) = 2 
> fs and
> keeping it = 1 fs??
>
> Is there a compromise for accuracy against speed of MD simulations?
>
> Will my results vary drastically?>
>
> Ideally ( & practically) how much should be the value of dt??
>
> ( no. of atoms in my system= 9079)
>
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.