# [gmx-users] Direction-periodic confusion

John Whittaker johnwhittake at zedat.fu-berlin.de
Tue Sep 17 12:49:03 CEST 2019

```Hi all,

I performed some simple umbrella sampling simulations involving two
peptides, after having increased the distance between them to about 6.5
nm, with the goal of deriving a PMF as a function of the absolute distance
between them.

In a stupid attempt to save resources, I created a long rectangular box
(about 15 x 6.5 x 6.5) and used the x direction as my reaction coordinate,
thinking it would give me something pretty close to what I would get if I
had used "r" as the coordinate. I used geometry = distance and dim = Y N N

Now I'm realizing that this is most probably not the case and I think I
tried every trick in the book in order to somehow "translate" my results
to those I would get with the absolute distance fixed, but no luck (unless
someone has a clever suggestion?).

Anyway, now I plan to carry out the same simulation in either: a) a much
bigger cubic or dodecahedron box or b) the same box, but using pull
geometry = direction-periodic.

The problem is, I originally planned to go with option b from the
beginning, but I got a bit confused with how direction-periodic works.

Essentially, I'd like to do what

geometry = distance
dim      = Y Y Y

does, but not have to worry about using a big box that is filled with a
bunch of water I don't really need (so at larger separation distances,
there are no issues/errors like "Distance between pull groups 1 and 2
(6.274720 nm) is larger than 0.49 times the box size (3.207410)" in the Y
and Z direction).

Is this possible using Direction-periodic?