David -
I have not been following what you are up to here, but two perhaps useful
comments:
1) The unit cell size must be at least twice cut + skinnb; a check is done
at startup to insure this condition is met. If you have a constant pressure
run, however, you could end up with a unit cell "minimum sphere radius"
below cut + skinnb, and this could cause problems with missed interactions
in pmemd due to how the pairlist gridding algorithm works. The code
probably should be better protected against this possibility, but I don't
think it is, based on a quick look. For pmemd at least, and possibly sander
(I have my own pairlist algorithm so I am not sure), the critical factor
really does include skinnb because of the way the unit cell is gridded in
the pairlist algorithm.
2) However, I would not recommend going anywhere near this lower limit in a
simulation, in order to reduce periodicity artifacts. You can see impacts
on rdf's for ions at small box sizes, for example. I guess I have not
worried a huge amount myself about folks running really small boxes with
largish cutoffs, or folks running NPT at ridiculously low starting
densities - say below 0.5 g/cc or thereabouts without building a box that is
big enough to shrink without going below the limit.
I should probably slip a check into the pressure adjustment for pmemd 10,
just to be sure this sort of thing does not happen unnoticed.
Regards - Bob
----- Original Message -----
From: "David Cerutti" <dcerutti.mccammon.ucsd.edu>
To: <amber.scripps.edu>
Sent: Friday, October 19, 2007 11:31 PM
Subject: AMBER: Addendum: Heat capacity calculations not coming out right
>I wanted to add that I was not at first taking into account the -2.22
>cal/mol-K correction for quantum effects that Yujiw Wu did. This means
>that ONE of my answers is closer to the value he reported, enough that I
>would be satisfied with the agreement. I'm now repeating his experiment as
>best I can in AMBER, but I have to use 256 molecules rather than 216 for
>fear that the box will get too tight for the 9A cutoff.
>
> Out of curiosity, what would happen if the nonbonded pair list cutoff (as
> opposed to the direct interaction cutoff) were larger than half the box
> length?
>
> Thanks!
>
> Dave
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.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 Oct 21 2007 - 06:08:02 PDT
