Re: COLVAR metadynamics: center of mass of atoms crossing simulation box

From: Giacomo Fiorin (
Date: Tue May 20 2014 - 11:41:56 CDT

Hello Lorenz, the centers of mass of the groups are always calculated
without PBCs (hence the need to disable wrapAll in many cases), but mutual
distances between the centers are instead calculated with PBCs.

This hasn't happened frequently before, but if you need it we can probably
add the forceNoPBC feature used in the distance-based variables to angle
and dihedrals as well in the coming release.


On Mon, May 19, 2014 at 8:17 AM, Lorenz Ripka

> We are performing metadynamics calculations using as collective
> variables (CV) an angle and a dihedral angle, defined using the centers
> of mass of several atoms of our protein complex (i.e. some of the groups
> of atoms span different molecules).
> The simulations proceed as expected until some of the atoms defining the
> CVs cross the periodic boundaries of the simulation box. When this
> happens large jumps are observed in our CVs.
> Our guess is that the centers of mass of the sets of atoms, used to
> define the angles, are computed AFTER wrapping their coordinates in the
> simulation box, thus resulting in sudden changes of their coordinates as
> the atoms cross the box boundary.
> This occurs even if we set wrapAll off. Indeed in the dcd files the
> coordinates of the complex are not wrapped.
> Can someone suggest a method to prevent this from happening or to solve
> this situation?
> Kind regards

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:20:48 CST