[gmx-users] pbc=full in 3.3.2

[Jay McAliley]

Sorry for the double post but I forgot my descriptive title...



Hi,

I am having a problem using pbc=full for crystalline polymers on Gromacs
3.3.2.
I have input files that run in 3.3.1 with no problem, but the system
explodes in 3.3.2.

Here's a simplified cartoon of my system, dotted lines being box boundaries:

      (CH3)  (CH3)
   ....|......|....
   .   CH     CH  .
   .  / \    / \  .
(C)--O   C--O   C--(O)
   .    ||     || .
   .     O      O .
   .              .
   .   CH3    CH3 .
   .   |      |   .
   .   CH     CH  .
   .  / \    / \  .
(C)--O   C--O   C--(O)
   .    ||     || .
   .     O      O .
   .              .
   .   CH3    CH3 .
   ....|......|....
      (CH)   (CH)

The system blows up, beginning with the carbons and oxygens in the periodic
images (shown in parentheses). Hydrogens initially stay put. All bonds that
cross the box boundary seem to lengthen; it's as if those bonds were defined
between the in-box atom positions, and they are pulling the atoms together
(each toward the center of the cell). It happens for the side group bonds as
well as the main chain bonds that define the infinite chains.

I won't say I'm positive I have my topology right.. but has anyone been
using pbc=full in 3.3.2?

I made a tarball of my input files if anyone is willing to take a look:
http://people.clemson.edu/~mcalilj/pbcfull.tar.gz
On our hardware these run ok for 3.3.1 but crash in 3.3.2.

Thanks for your help,

Jay McAliley