AW: fixed atom/Constraint failure in RATTLE algorithm/

From: Norman Geist (
Date: Tue Jun 04 2013 - 02:01:10 CDT



you are right, constant pressure simulations with fixed atoms run into
problem often, but usually if these atoms are near the box boundaries, as
the box can't shrink then and will come with a message like "Periodic cell
has become too small for original patch grid". The problem you face may
happen due the coordinate rescaling which comes with changing the box size.
Here the fixed atoms can't be rescaled and so atoms will get moved into each
other what causes high repulsion forces which unload as huge velocities and
blow apart your molecular system. It's a good idea to equilibrate your box
density/pressure first, before starting the pulling stuff. Maybe modifying
the barostats "period" and "decay" times to a smoother timescale could also
work, so that the cell fluctuation amplitude is lower and a higher decay
time would let the system relax to the new pressure conditions.


Good luck

Norman Geist.


Von: [] Im Auftrag von MEHRAN
Gesendet: Montag, 3. Juni 2013 21:55
An: Norman Geist
Betreff: Re: namd-l: fixed atom/Constraint failure in RATTLE algorithm/



Hi Norman,

First of all let me thank you for the reply, I really appreciate your help.
I totally agree with you, it must be a bad initial structure and it must
have some thing to do with pressure. when I turn off my " Constant Pressure
Control " it works perfectly. I increased number of minimization steps, as
you suggest, to 5000 and the pressure went from -6250 bar to -3475 bar. I
think problem is happen at the beginning of dynamic when langevinPiston
starts to increase the pressure (pressure set to be 1 bar), those atoms
bonded directly or by the others to the fixed atom cannot obey
langevinPiston. Although boundary is at least 30A away from fix residue,
still it does not work. In NAMD tutorial for pulling protein, they do not
control pressure at all. So I believe that it is not easy to fix residue
near boundary and also fix pressure.

Now, I am trying to do Equilibration without controlling pressure for few
thousands step then if the pressure is dropped enough, use its output for
new MD simulation. if it does not work I will try other pressure control
method "Berendsen pressure bath coupling". I will keep this post updated if
my solutions works.

please let me know if you any suggestion or comment,





This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:17 CST