[gmx-users] gmx traj

[Alex]

Hi all,

I have an aqueous system featuring a membrane and a bunch of ions driven by
a constant field in the z-direction (PBC). In order to look at current
fluctuations in time, for every frame, I represent current as proportional
to <v_z> for the whole ion group (K, CL, for instance), so I run gmx traj
with -ov and -com options. This does yield the velocities.

Here is the fun part: there is a long-standing simple formula for getting
the actual average current, which goes something like the time-average of
per-frame q*<v_z>/L_z, which is nothing but the ion charge, times the
effective frequency it crosses the entire box. This produces a number. In
order to cross-check this number, we use a much more reliable current
calculation based on the ion fluxes -- this is completely independent and
is a script that processes the ion trajectories. That also produces a
number.

The two numbers are vastly different -- the current from gmx traj is 20
times smaller than the flux-based estimate. The average is over 250 ns,
which I believe is 2500 velocity frames. Is there anything PBC-related I am
missing, or is the <v_z>-based method is simply unreliable, given the noise
and the velocity output frequency (which I am obviously happy to increase)?

Thank you,

Alex