[gmx-users] Pull code, Velocity distribution

[Samadashvili Nino]

Hello,

       I have been using the pull code for friction calculations. I am not doing umbrella sampling but just pulling one slab of crystal on top of another with the constant velocity. I would like to know how  Gromacs is calculating the pulling force (pullf.xvg) during sliding.  Is the pulling force obtained through U = 1/2K(vt-(x-x0))^2 harmonic potential? 
I am using following parameters:

pull                           = umbrella
pull_geometry            = direction_periodic
pull_dim                    = Y Y Y
pull_start                   = yes
pull_ngroups              = 1
pull_group0                = ICE_A
pull_group1                = ICE_B
pull_pbcatom0           = 0
pull_pbcatom1           = 0
pull_vec1                  = 1 0 0 
pull_rate1                  = 0.004 
pull_k1                      = 10000

Since I have pull_start=yes, does it mean that the initial spring length is the COM distance between ICE_A and ICE_B? 
Another question is regarded to the velocity distribution. I used g_traj to plot the velocity distribution of my system which I compared with the  distribution plot calculated from the coordinates and they dont match. 
The distribution calculated by Gromacs has a tail while my calculations don't show any. Could you please tell me how is it calculated in Gromacs?
Thank you for your attention to this matter.

Cheers,

Nino