[gmx-users] PBC, lincs and bonds breaking

[Paul Whitford]

>>* I am trying to simulate my own structure based simulation using 3.3.2.
*
>What sort of structure?
RNA with a ligand (the ligand is the trouble)

>>* I
*>>* have developed the forcefield and it works fine without periodic boundary

*>*> conditions.  When I use pbc=xyz atoms that are connected in the [bonds]
*>*> section of the topology get broken.  What happens is 1 of the two atoms

*>*> will
*>*> cross the boundary and that single atom will get placed on the opposite**
>> side
*>*> of the box.
*
>This won't be the source of the problem, unless they broke PBC in 3.3.2
>and you're the first to notice :-)



I should clarify.  At 1 step, the atom gets moved from 1 side of the
box to the other.
In the following step, the entire molecule gets "ripped to pieces".
It appears that lincs
is trying to satisfy the constraints, so the whole molecule gets move
"near" the center.  It


is near in that it is no longer close to the side of the box, but the
individual atoms are still
quite a distance apart (~20 Angstroms).


>>*  In the next step, the system will explode.  If I use lincs, I

*>*> get error messages about angles getting changed and that lincs can not
*>*>
 satisfy the constraint.  I get the same issue with shake.  when i run


*>*> lincs,
*>*> the structure is output before the errors and after, so I can see that 1

*>*
> atom is placed on the opposite side of the box and that is the same atom
*>*> that I get lincs errors about.   Below is my .mdp file.

*

>You don't mention your EM or equilibration regime. Please use one, or

>describe yours :-)

http://wiki.gromacs.org/index.php/Steps_to_Perform_a_Simulation


I run minimization and that works, IF none of the atoms cross the
border.  If I make my box a little smaller,
such that atoms are over the boundaries initially, then I get the same
errors with lincs.  If I use a larger box, and

minimize, I can start the run juts fine.  The simulations runs ok for
about 16,000 time steps, and then the
problem surfaces.  If I use the cvs version, I get the message

There were 6 inconsistent shifts. Check your topology

   Energies (kJ/mol)
           Bond          Angle    Proper Dih.  Improper Dih.          LJ-14
    6.62576e+06    1.09907e+03    2.71644e+02    2.49075e+02   -1.13891e+03
     Coulomb-14        LJ (SR)   Coulomb (SR)      Potential    Kinetic En.

    0.00000e+00    3.58079e+00    0.00000e+00    6.62624e+06    4.66262e+04
   Total Energy    Temperature Pressure (bar)
    6.67287e+06    1.10673e+03   -2.18054e+04

There were 2 inconsistent shifts. Check your topology

           Step           Time         Lambda
          11400        5.70000        0.00000

There were 22 inconsistent shifts. Check your topology
-------------------------------------------------------

Program mdrun, VERSION 3.3.1
Source code file: stat.c, line: 257

Fatal error:
XTC error
-------------------------------------------------------

"You Hear Footsteps Coming From Behind" (Colossal Cave)

If I use 3.3.1 I get the following message (immediately) when I try to
run the system

-------------------------------------------------------
Program mdrun, VERSION 3.3.1
Source code file: mshift.c, line: 91

Fatal error:
More than 8 graph edges per atom (atom 432)

-------------------------------------------------------

"It Costs Too Much If It Costs a Lot" (Magnapop)



>>* I saw the same errors have been posted when people try to run infinite
*>*> polymers, but I didn't see any resolution to this.  Any help would be
*>*> appreciated.  thanks

*
>Does the same problem occur for a different part of your system if you
>change the seed for generating velocities?

I haven't tried different seeds, but I have done runs at different temperatures
and, I get the same errors, once the molecule hits the boundaries.

Perhaps these issues with other versions can make it obvious what is going on.

>Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20071026/9df4b934/attachment.html>