On Tue, Apr 13, 2010, Jianing wrote:
> but maybe I have found the reason for this surprising phenomenon, and I
> need you to give a juge on this.
> you know, writing every 4fs for 3.5ns is a very large mdcrd file,
> especially when the trajectory is not a binary file, so I just cut
> it into hunderds of pieces. That is, when sander finishes every 10ps
> interval,it will output md1.rst, then, when do the next 10ps, the
> initial coordinates is from md1.rst. and this cycle lasts until 3.5ns is
> reached. Although this is proceeded by computer automatically, this may
> also bright in errors, because the data accuracy is different between
> calculating and reading data in sander, which I didn't know before.
OK -- my bet is that you are seeing the Langevin problem, since you are using
ntt=3. You *must* use a different random number each restart (or set ig=-1).
See:
.Article{Sindhikara09,
author = {Sindhikara, D.J. and Kim, S. and Voter, A.F. and Roitberg, A.E.},
title = {{Bad seeds sprout perilous dynamics: Stochastic thermostat induced
trajectory synchronization in biomolecules}},
journal = {J. Chem. Theory Comput.},
volume = {5},
pages = {1624-1631},
year = {2009}
}
and earlier articles referred to there.
[developers: is there any good reason we should make the default value be
-1 for ig?]
...good luck.....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 13 2010 - 05:00:04 PDT
