[gmx-users] Strange acceleration behaviour after continuation

[Kamps, M.]

Dear GMX users,

I am trying to replicate a molecular flow of small molecules between
two solid slabs of FCC bound atoms. The slabs are bound via LJ
interactions.The flow is created by using periodicity, where the fluid
is accelerated continuously between the two slabs. The periodic box is
created as a long small 'channel', where periodicity can occur within
the X and Y directions. The Z direction is restricted due to the fixed
'walls'.

I am now trying to control the fluid flow, by altering the
accelerations. To do this, I first accelerate the fluid with a
constant acceleration 'a1'. This will result in a movement of the
fluid over time. At t=0 the velocity is zero, while at t=1 the
velocity equals 0.06 nm/ps2. The velocity development is linear,
meaning it has a linear increase from 0 to this value of 0.06 nm/ps2.

After this 'first stage', I want a less rigorous increase within the
velocity, therefore I perform a second simulation, continued from the
previous simulation. In this case the acceleration is set to a value
that is 20% that of 'a1' (the previous acceleration). Within the MDP
file it is set to genvel=no and continuation=yes. The .cpt and .gro
files are provided as the input for this second simulation.

My expectation is that the velocities at the beginning of this
simulation are equal to that of previous simulation, hence the
continuation. However, this is not the case. The velocities at the
beginning of this simulation start at roughly 0.02 nm/ps2 and increase
(slowly) to 0.04 nm/ps.

My question: why is there a mismatch in the velocities? Am I doing
something wrong? Is my logic flawed?

Mark