Hey, Gyorgy -
The problem is definitely with the incorrectly defined constraints as
I mentioned yesterday. You may find Chapter 5 of the gromacs manual
useful to repair this.
I hope to be able to look through your attached system on the weekend,
if you don't locate the problem before yourself...
Regards and no runtime errors in your simulations,
Vitaly
On Thu, Feb 17, 2011 at 11:48 AM, gyorgy.han...@fc.up.pt wrote:
Hi Vitaly,
Thanks for your answer.
Actually, we tried in a lot of ways to run the simulations. This was also
one way, and in fact, if you use a much smaller time step, the simulation
doesn't crash (I'm not saying it gives a physical result).
If I remove the [bonds] part from the topology, the simulation runs and the
geometry is conserved, but produces positive total energy. Maybe this is
related to the unphysically high force constants of the angle bendings, but
I think, this doesn't matter if the angles are contrained.
(Maybe the problem is somewhere here, but I don't know how to constrain
properly the angles if not like this (and nobody can tell me in my
proximity))
If I remove the [angles] section, de system's total energy is negative but
the geometry of the anion is completely wrong.
I also have problems when running on more than 1 processor, even if I use
-pd for mdrun (to switch to particle decomposition instead of domain
decomposition), the simulation crashes immediately:
Range checking error:
Variable n has value 7. It should have been within [ 0 .. 7 ]
If you want to test the topology on one single ion pair, you can find a
configuration in the attachement.
Thanks in advance,
Best,
Gyorgy
Quoting Vitaly Chaban vvcha...@gmail.com:
Hello,
I got your files. However, the error is completely another than you
mention and it occurs directly at the first time-step of the
simulation.
Inner product between old and new vector = 0.0!
constraint #1 atoms 1377 and 1381
Wrote pdb files with previous and current coordinates
step 0Inner product between old and new vector = 0.0!
constraint #1 atoms 2177 and 2181
Wrote pdb files with previous and current coordinates
Inner product between old and new vector = 0.0!
constraint #1 atoms 2 and 6
Wrote pdb files with previous and current coordinates
Segmentation fault
There are a number of strange things in your configuration with the
unhealthy defined constraints in topology as probably the most severe
one. If you have constraints, why do you define bonds force constants
at the same time?
Good luck with your study,
Vitaly
On Wed, Feb 16, 2011 at 1:58 PM, gyorgy.han...@fc.up.pt wrote:
Hi Vitaly,
Thanks for your answer. I was busy with other stuff these last two weeks,
that's why I have time to reply only now.
So the system is the BMIM PF6 ionic liquid.
We chose two potentials to describe the system, and both treat the anion
as
rigid. We cannot use LINCS because it shouldn't be used if one constrains
not only bonds but angles too, so we have to use SHAKE.
But with SHAKE I encountered some strange things as I wrote in my first
thread. In the attachement I am sending you all the files needed for the
simulation, maybe you'll find the problem.
I havent mentioned but we try to constrain the angles by adding fictious
bonds. Maybe there is something wrong with this...
Thank you in advance for your help.
All the best,
Gyorgy
Quoting Vitaly Chaban vvcha...@gmail.com:
Hey Gyorgy,
Your current topology file(s) is(are) also important to analyze the
situation and fix the problem.
Pass my kind regards to Julia...
Dr. Vitaly V. Chaban, Ph.D.
Department of Chemistry
University of Rochester
Rochester, New York 14627-0216
The United States of America
=
The mdp file is attached.
Best,
Gyorgy
Quoting Justin A. Lemkul jalem...@vt.edu:
gyorgy.han...@fc.up.pt wrote:
Dear all,
I am setting up a simulation of ionic liquids with the PF6 anion.
According to the potential, the anion should be kept rigid, wich
obviously means that bond lengths and angles have to be
constrained. LINCS doesn't work with angle constraints (i.e.
constraing a triangle), so we decided to use SHAKE. However, SHAKE
seems to work a bit strangely: I know SHAKE mustn't be used with
domain decomposition, but even if I set the corresponding variable
to NO in the mdp file, the simulation crashes on 8 procs and gives
the following error message:
Fatal error:
1 particles communicated to PME node 7 are more than a cell length
out of the domain decomposition cell of their charge group.
If I try to run mdrun with -pd (to 'really' switch off domain
decomposition), the simulation doesn't chrash but gives nonsense
(the energy seems to increase constantly).
I am not an expert user so maybe I do something wrong but, anyway,
does anyone have an