[gmx-users] Restart from .gro and tip4p water

[Francesco Mercuri]

Hello,
I'm running restart jobs for a single TIP4P water molecule with
the settle constraint as a test case.
However, when restarting from a previous configuration as a
.gro file, the velocities associated to the massless point of TIP4P water
(4th atom of the sequence) are different from their starting values.
As an example, this is part of the output I get when
restarting from an old .gro file (1st.gro) and try to run a longer
simulation:
                                                                                                                                   
trjconv_d -f 1st.gro -o 1st.trr
gmxdump_d -f 1st.trr
[...]
   v (4x3):
      v[    0]={-2.15989e-01, -4.22379e-02,  3.49524e-03}
      v[    1]={ 1.59082e-01, -2.82498e-01,  1.91192e-01}
      v[    2]={ 3.26919e+00,  9.52916e-01, -2.46670e-01}
      v[    3]={ 2.78138e-01,  5.42943e-02, -4.60771e-03}
[...]
                                                                                                                                   
grompp_d -f 2_in.mdp -po 2nd.mdp -o 2nd.p.tpr
gmxdump_d -s 2nd.p.tpr
[...]
   v (4x3):
      v[    0]={ 0.00000e+00,  0.00000e+00,  0.00000e+00}
      v[    1]={ 0.00000e+00,  0.00000e+00,  0.00000e+00}
      v[    2]={ 0.00000e+00,  0.00000e+00,  0.00000e+00}
      v[    3]={ 0.00000e+00,  0.00000e+00,  0.00000e+00}
[...]
                                                                                                                                   
tboconv_d -s 2nd.p.tpr -f 1st.trr -o 2nd.tpr -extend 0.01
gmxdump_d -s 2nd.tpr
[...]
      v[    0]={-2.15989e-01, -4.22379e-02,  3.49524e-03}
      v[    1]={ 1.59082e-01, -2.82498e-01,  1.91192e-01}
      v[    2]={ 3.26919e+00,  9.52916e-01, -2.46670e-01}
      v[    3]={ 2.78138e-01,  5.42943e-02, -4.60771e-03}
[...]
                                                                                                                                   
mdrun_d -v -s 2nd.tpr -o 2nd.trr -x 2nd.xtc -e 2nd.edr -g 2nd.log -c 2nd.gro
gmxdump_d -f 2nd.trr
[...]
   v (4x3):
      v[    0]={-2.15989e-01, -4.22379e-02,  3.49524e-03}
      v[    1]={ 1.59082e-01, -2.82498e-01,  1.91192e-01}
      v[    2]={ 3.26919e+00,  9.52916e-01, -2.46670e-01}
      v[    3]={-8.99518e-05,  1.09747e-03,  6.92733e-04}
[...]
                                                                                                                                   
Thus, even if the velocities are correctly passed to the input
file for the restart run (2nd.tpr), the velocity associated to the
massless point for the first step of the restart run is different
from its original value.
                                                                                                                                   
However, this happens when the first trajectory file (1st.trr)
is created starting from an ascii .gro file (e.g. with
trjconv_d -f 1st.gro -o 1st.trr).
On the other hand, if the trajectory file is just the result
of a previous simulation, then the velocity components of
the massless point are set to zero in the first step of the
restart run.
In both cases, the gen_vel option was set to "no".
                                                                                                                                   
Do you have an explanation for this behavior?
Changing the velocities of the massless points would result
in different dynamics, in a way that, let's say, a first
run for t ps followed by a restart run for another t ps
gives different results from a single run of 2t ps.
How would it be possible to get "correct" velocities for
the massless atom starting from a .gro file?
Best regards,
                                                                                                                                   
        Francesco

-- 
___________________________________________________________
Sign-up for Ads Free at Mail.com
http://promo.mail.com/adsfreejump.htm