[gmx-developers] Linear bond angles= exploding energy

[Paul Whitford]

I wasn't sure if I should post this to the gmx-users or gmx-developers list.

I am constructing coarse-grained CA models for proteins.  Adjacent CA atoms
are connected by harmonic bonds and sets of three CA atoms are connected by
harmonic angles.  I have noticed that when a bond angle gets linear (pi) ,
there is a large spike in the potential energy and that part of the system
goes  a bit wild (unrealistic high accelerations).   I suspect this has to
do with the dihedral term, but I'm not sure.  In Amber, there was a comment
about a disconinuity when a dihedral has a linear angle (I can't find it at
the moment).  Was this considered when writing gromacs?   In all-atom
simulations, this is not be a problem, but in these coarse grained methods
linear angles are ok, since they are not "real" covalent bonds.   Any
thoughts/fixes would be appreciated.

Thanks

-Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20090401/35510915/attachment.html>