Re: Re: pairlist i_upper mismatch!

From: Schreiner Eduard (eschrein_at_ks.uiuc.edu)
Date: Sun Nov 23 2008 - 02:08:40 CST

so, apparently the structure of your molecule goes crazy.
This can indeed happen, if the volume of the cell changes
too rapidly, because the coordinate of the atoms are rescaled
accordingly.
Maybe someone wants to correct me, but I think, it is not a good
idea to increase the number of steps between the evaluation
of full electrostatics and nonbonded interactions during a critical
phase of the simulation. These keywords
"nonbondedFreq"
"fullElectFrequency"
are a little confusing, so check the manual.
I would evaluate everything every step, use a tiny time step and write
all
the infos (including all the restarts) to see what is really going on.
Did you save the cell size, coordinates ..... every time step?
It may give you a clue what really happens: cell shrinks/explodes,
which atoms are moving where ....
When does the error you report occur, after the first time the
pressure is adjusted?
I still think the slow Langevin parameters in conjunction with a
small time step may help. Maybe someone wants to correct me if I'm
wrong.

It may also be worth to equilibrate for some time in at constant volume,
and switch then to constant pressure using slow Langevin piston
parameters. However, if your cell size is too bad, I do not know if it
will
help you.

eddi

On Nov 22, 2008, at 10:20 PM, Diego Alejandro Vargas wrote:

> Floris and Eduard,
> Thank you for your suggestions. I am still encountering a problem
> though. After reinitializing the velocities to the same temperature
> as langevinTemp and langevinPistonTemp I no longer get the i_upper
> pairlist error. Also, I was able to run it with a margin as little
> as 2.0 (0 or 1 give me the error that the cell has become too small
> for patch grid). Changing the langevinPistonPeriod and decay did not
> change much; with this a margin of around 10 was needed to prevent
> the error.
>
> Now when I try to run the simulation I usually get either one or
> both of the following errors:
> ERROR: Atoms moving too fast; simulation has become unstable.
> ERROR: Constraint failure in RATTLE algorithm for atom...
>
> From checking the molecule in VMD, these are atoms belonging to the
> protein.
>
> I changed the timestep (shorter), nonbondedFreq (larger),
> fullElectFrequency (larger) to try and solve this, but the error
> persists. Along with these errors I got ERROR: Stray PME grid
> charges detected: 3 sending to 0 for planes 0. And when reducing the
> timestep even more (0.5 fs) I get only the error FATAL ERROR: Bad
> global exclusion count! Which leads me to believe the error may
> still be in the cell size.
>
> Eduard, non of my cell's dimensions are smaller than 50A, the three
> dimensions are: 98, 120, 97. And the errors occur on the very first
> step of equilibration. I restarted the velocities just for
> continuity from minimization (10,000 steps): You are right it
> shouldn't make the difference but any simulation needs either
> velocity or temperature specification and I did not define
> temperature in the configuration file.
>
> Do you have any insight into this errors, and if so any suggestions?
>
> Diego

=====================================================
Eduard Schreiner
Theoretical and Computational Biophysics Group
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801
Phone: 217-244-4361
Fax: 217-244-6078
http://www.ks.uiuc.edu/~eschrein/
=============================================

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:50:08 CST