From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Sat Aug 11 2007 - 15:05:20 CDT
Hi Joshua,
I don't do nanotube simulations, so I can only offer general remarks,
but it sounds like some of these issues are shared by people who want to
do biopolymer simulations in which the polymer is bound to its periodic
image -- my thoughts are included below.
Joshua D. Moore wrote:
>
> When I visualize the nanotube in VMD and apply my axial direction unit
> cell vector to see the images, they seem to align properly. There is
> a gap between the images, but I thought that this was correct as this
> gap accounts for the bond length between the carbon atoms in the
> images. If I choose the periodic boundary so that the end atoms
> overlap each other, I still get problems, because I am guessing the
> atoms are overlapping!
You're probably correct in your alignment if the spacing looks right,
but if the atoms aren't actually bonded in your simulation, it is not
surprising that the ends would crumple slightly instead of doing what
you expect -- having atoms within bonding distance that are not actually
bonded to each other tends to cause this ;-)
>
> Now that I have experimented quite a bit, I don't see how the bonds
> between the images can be included at all. I could add bonds between
> the end atoms on each side of the nanotube to my *psf file manually,
> but I am not sure if this is the solution. I am not sure, but I
> really think that this is the problem..... the end atoms don't know
> that they are bonded to anything so they are free to move.
This is probably exactly what you should do. If you add bonds to the psf
file, either manually (not recommended, since the formatting of psfs is
very tricky) or through a patch added during psf generation, things
should work as you want. When namd is calculating bonds, it always
assumes that the covalent bond from one atom to another is to the CLOSER
of the corresponding partner in the periodic cell or its periodic image,
which usually behaves as you want. Here, for example, if you add the
bonds that you suggested, they will be assumed to go to the nearby atoms
of the periodic image, and thus behave as you'd expect.
Peter
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:04 CST