Re: Why wrapped and unwrapped trajectories show different energies over the time?

From: Brian Radak (
Date: Sun Apr 24 2016 - 09:03:07 CDT

Are you looking at the total potential energy or the actual total =
potential + kinetic energy? The kinetic energy can jump substantially
(~1 kcal/mol) upon restart, even if velocities are not written (try
multiple run commands in the same script).

If the potential is changing, then the kinetic energy is almost
certainly not the problem. I just wanted to bring up the kinetic energy
again because no one has yet been able to explain that to me.


On 04/24/2016 12:23 AM, Jason Swails wrote:
> How different are the energies? And how long was your simulation run for?
> Ideally the energies *should* be identical. However, trajectories are
> typically stored in single precision rather than full double precision
> (and that's true for the DCD format). This means that coordinates
> lose precision a lot faster when read from a trajectory as opposed to
> a data source that stores double precision data. If you store the
> trajectory in a crappy format, like simple truncated ASCII or PDB,
> then this scenario becomes even more pronounced.
> Of course there are many other ways that energies can differ by small
> amounts as well:
> 1. The PME grid is set up differently (possible particularly when
> doing NpT calculations where the unit cell size changes)
> 2. If the unit cell dimensions change (i.e., because you are running
> NpT calculations), then you have to make sure that the unit cell
> vectors applied to every frame is the same as the unit cell vectors
> for that particular frame.
> #1 is a likely concern (check for electrostatic energy differences).
> I would expect #2 to be handled automatically (since DCD files store
> unit cell information for each frame), but it's worth checking to make
> sure.
> The trick here is to find some frames where the energy differs, and
> see where the energy differs and by how much.
> HTH,
> Jason
> On Sat, Apr 23, 2016 at 10:22 PM, zeynab hoseyni <
> <>> wrote:
> Dear all,
> I ran NVT simulation of a nano-tubes in addition to a
> dimer solvated in water. In the namd conf file
> I turned on the wrapAll and wrapWater options. I saw that the
> nano-tube and dimer were so
> much moving around and going out of the simulation box frequently
> during the simulation. Since I already turned
> on the wrap options then sometimes either nano-tube or dimer jump
> from one side to the other side
> of simulation box, depending on which one catching the side wall,
> not suitable for doing some
> calculations. I then decided to unwrap the nano-tube (resname
> CNT) and dimer (resname ARG)
> by following command in vmd:
> pbc unwrap -sel "resname CNT or resname ARG" -all
> Doing so I calculated the total energy of the wrapped and
> unwrapped trajectories using namd energy
> plugin. I expected they show the same plot of total energy versus
> time, but they show slightly different plots.
> Can anyone help on this? In addition to this I don't have a
> clear understanding of what wrapping option
> does in namd or vmd as I see sometimes nano-tube or dimer
> traverses the cell side (one part pass the side
> and one part still in the box) but it isn't wrapped and sometimes
> it is in the similar situations and it would be wrapped..
> Don't know exactly what it does.. Can anyone introduce me some
> reference or tutorial to understand more?
> Thanks,
> Zeynab
> --
> Jason M. Swails

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:22:03 CST