[gmx-users] Umbrella Sampling/SMD pulling distance

[Shrinath Kumar]

Hi,

I'm trying to do an SMD simulation to generate configurations for Umbrella
sampling. I want to pull two groups together.

I'm having trouble understanding why the pulled distance must smaller than
the half the *shortest* box-vector length.

I know that distance is calculated taking periodicity into account and that
distances/vectors between objects should obey the minimum image criterion.
Hence. if the restraint distance becomes larger than the box length you'd
get discontinuities in the pulling. But I thought this is checked
component-wise. i.e
dx   =   x_i - x_j  <  Lx*0.5
dy   =   y_i - y_j  <  Ly*0.5
dz   =   z_i - z_j  <  Lz*0.5

but instead what seems to be necessary is that
| r | = sqrt(dx^2 + dy^2 + dz^2) < min(0.5Lx, 0.5Ly, 0.5Lz)

Could someone explain why it is not sufficient to check whether dx,dy and
dz are smaller than half their respective box vector lengths to avoid
discontinuities in the pulling? But instead, we must check if | r | is
smaller than half the shortest box vector lengths.

I ask because in my system the two groups that I want to pull together are
separated a large distance in the x-direction but not that much in the y
and z-direction. So accordingly, I made a box that is long in the
x-direction but small in y and z but this doesn't work out and I get the
error the distance between pull groups is larger than half the box size.

Sincerely,
Shrinath