AMBER Archive (2003)

Subject: Re: AMBER: 1 fs timestep for GB?

From: Andreas Svrcek-Seiler (svrci_at_tbi.univie.ac.at)
Date: Tue Nov 25 2003 - 12:32:15 CST

Hi,
 
> I noticed that people tend to use 1 fs time-step for integration in GB/SA
> simulations. I understand that the scheme implemented does not involve
> friction with the environment and this can give rise to faster
> motions in some degrees of freedom of the solute. I am not sure how much
> faster bonds fluctuate in this case, but a 2 fs integration step
> did not produce anything strange for my nucleic acid simulation:
...I believe a 2 fs timestep won't hurt (at least not more than in other
simulations). The fast motions frozen
by SHAKE or RATTLE are bound to be similar in GB/SA and
other types of simulations.

> no drift in total energy over 10 ns - although the simulation was
> run at constant temperature (300 K) with tautp=5.0.
...to check whether a large timestep/finite cutoff/error in the gradient
leads to drifts, tautp= *say* 1.e9
(or -better- no heat bath at all) would be more appropriate.
The thermostat masks drifts quite efficiently.

Hope it helps,
good luck
Andreas

             
            )))))
            (((((
           ( O O )
-------oOOO--(_)--OOOo-----------------------------------------------------
              o Wolfgang Andreas Svrcek-Seiler
              o (godzilla)
                       svrci_at_tbi.univie.ac.at
      .oooO Tel.:01-4277-52733
      ( ) Oooo.
-------\ (----( )--------------------------------------------------------
        \_) ) /
              (_/

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu