> I'm running a dynamic and some water molecules from the periodic box
> became about 1000 Angstroms far away from mi solute molecule.
> In the restart file appears the following data as coordinates :
> [....]
> 58.9133223************ -13.5743458 59.3766289************ -13.2441555
> 58.8037298************ -14.5100250-247.8082567 -68.2815635 316.0647351
> [....]
> After generating this restart file with asterisks the dynamic crashs.
> I've measured the distances of some water molecules and are about 1000
> Angtroms in the previous restart file (which did not give any problem).
> To solve this I used ptraj to center and image the water box and I added
> the velocities from the input restart file used as input for ptraj to
> get the new one centered. Although the dynamics runs with no problem
> now, the info from the outmd shows diferences:
You can also set IWRAP=1 to image the water's, however your approach is
sound, i.e. center/image, write out a new restart file and copy by hand in
the old velocities. Future versions of ptraj will handle the velocities
automagically, that is sometime in the future...
In spite of this, small differences may result due to finite precision of
the coordinates and box leading to small shifts as it is imaged. The more
box images you are away from the center, the larger the potential
difference. However, it should be small; moreover, from your output it is
not possible to determine where the difference is coming from as you are
showing NSTEP=2000. Given a parallel run and non-determinism in the
communications algorithms, it is not expected that after 2000 steps the
energies will be the same or that the same phase space is sampled.
What you want to look at is the NSTEP=1 energies and see that these are
close.
Note that an additional issue can be that ptraj may incorrectly read
restrt or mdcrd files when the numbers munch together as in your example.
A fix is in the works and in the meantime, set IWRAP=1 in the mdin file.
tec3 . utah.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Mar 04 2007 - 06:07:40 PST
