[gmx-users] how to set "gen_vel" in mdp file and continue a crashed run

[Jian Zou]

Thank you very much for your kindly help, Xavier, David,

Having read your comments on gen_vel and tpbconv, I want to put my 
understanding here.
Please check if they are correct

normally,

1. "gen_vel=yes" with "gen_temp=ref_t" (ref_t for T-coupling) will speed up 
equilibration slightly.

2. for a new simulation, "gen_vel=yes" should be set for PR-MD, and 
"gen_vel=no" for the final MD, velocities at t=0 will be read from the gro 
file;
for a continuation md simulation, "gen_vel=no" should be set, otherwise you 
start a new simulation from the structure or conformation of the previous 
simulation.

3. to continue a complete run, just using the final gro and "gen_vel=no"; or 
use tpbconv with -until or -extend option.

4. to continue a crashed run with tpbconv, gen_vel does not influence 
because velocity will be read from trr file.


to continue a crashed run, with T, P coupling
set in mdp file, eg. "nstxout=10000" and "nstvout=10000"

# check the last complete frame containing
gmxcheck_s -f md.trr

# dump the last frame with higher precision
trjconv -f md.trr -s b4md.tpr -n index.ndx -o frame.gro -ndec 6 -dump 1234

#  convert the dumped frame back to trr format
trjconv -f frame.gro -n index.ndx -o frame.trr

#  generate new tpr file with restart coordinates & velocities
tpbconv -f frame.trr -s b4md.tpr -n index -e md_ener.edr -o restart.tpr

#  Re-start with the new tpr file
mdrun -s restart.tpr -o md2.trr ...


Thank you so much again.


Regards,

Jian Zou