Hi Logan,
I'm guessing that the more experienced hands on this list can give you
more sagely advice, but the smallest box dimension must be at least
2*(cut + skinnb) otherwise the simulation will not run. [Note that
skinnb is 2 Å by default in AMBER, though this can be changed.] This
would be the absolute minimum size system you could have, and I (and
probably others) would suggest having at least several Å of additional
solvent "buffer".
Just to give an example, I commonly solvate my systems using
"solvateoct solute solvent 9" in leap, thus guaranteeing at least ~16
Å between neighboring protein (image) molecules after NPT
equilibration (my real space cut-off is usually 9 Å). This amount of
buffer may be overkill for some systems and not enough for others in
terms of avoiding periodicity-induced artifacts in your simulations.
I would suggest reading an "oldie, but goodie" from the McCammon
group: http://pubs.acs.org/doi/abs/10.1021/jp9937757. More recently,
I know that Robert Best and Gerhard Hummer have examined this issue in
their force field optimization efforts: http://pubs.acs.org/doi/abs/10.1021/jp901540t
.
Hope that helps,
Paul
On Jul 13, 2010, at 4:23 PM, Logan Ahlstrom wrote:
> Hello,
>
> Is there an accepted distance to have between neighbors in a
> periodic MD
> simulation? i.e. is it enough for images to simply be beyond the
> electrostatic and van der Waals cut-off distances, or should they be
> "much"
> farther apart? Any input would be helpful.
>
> Sincerely,
>
> Logan
>
> --
> Logan S. Ahlstrom
> Ph.D. Candidate
> Department of Chemistry and Biochemistry
> University of Arizona
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jul 13 2010 - 17:00:05 PDT
