[Pw_forum] problem changing the time step with pwscf

Eduardo Ariel Menendez Proupin eariel99 at gmail.com
Fri May 23 18:33:53 CEST 2008


Hi,
I noticed that running molecular dynamic calculations , the time step (dt)
cannot be changed if one restarts from a previous run
(restart_mode='restart'). This happens at least in versions 4.0 and 3.2.
If it is done, pw.x runs without complain (see note below), but the
velocities and the kinetic energies are not correctly calculated. Also, in
the prefix.msd.dat file, the times are calculated as dt*istep, producing a
jump in time if dt is changed. The time should be something like
previous_final_time+(istep-previous_nstep)*dt. Note that the time is
correctly reported in the standard output.

This can be tested with example04, running it mannually to avoid deletion of
the temporary save, and restarting wiht a modified input file
si.md8.in  using
    restart_mode='restart',
    dt=10,
    nstep=150,

Note: There is also a warning message when restarting in example 4: RECOVER
from restart file failed: file not found.

Here is my bash script (just one more) to extract the temperature and the
energies from the output file. The jumps are seen plotting the dat files.

INFILE=si.md8.out
SUFFIX=si8.dat
grep kbar $INFILE |awk '{print $6}' >>p_${SUFFIX}
grep 'temperature           =' $INFILE |awk '{print $3}' >>t_${SUFFIX}
grep 'Ekin + Etot (const)' $INFILE |awk '{print $6}' >>etotal_${SUFFIX}
grep 'kinetic energy (Ekin) =' $INFILE |awk '{print $5}' >>ekin_${SUFFIX}
paste etotal_${SUFFIX} ekin_${SUFFIX}|awk '{printf("%7.7f \n", $1-$2)}'
>epot_${SUFFIX}

Why change the time step? Maybe just to check the stability of the
simulations. Or maybe a large dt is enough for driving the system to a
certain temperature near equilibrium, but a smaller one is need after the
thermostat is switched off.

-- 
Eduardo Menendez
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20080523/1334e33c/attachment-0001.htm 


More information about the Pw_forum mailing list