[gmx-users] Shake block error:

[Ben Roberts]

Hello all,

I'm trying to run a simulation and short dynamics run to get a
relatively equilibrated structure for a molecule.

The simulation is unusual because it is a highly branched molecule, and
I have modified the .itp file to repel the ends of the branches both
from the centre and from each other (in order to generate a completely
extended structure). This was done via a Type 6 bond, with a very long
length (I'm aware that a Type 6 bond is a simple harmonic restraint).

The first stage, running Editconf, Grompp and Mdrun (Energy
Minimisation) works fine. However, when I try and run Grompp a second
time in preparation for the dynamics run, I get the following message:

	Walking down the molecule graph to make shake-blocks
	There are 132 charge group borders and 1 shake borders
	There are 132 total borders
	Division over nodes in atoms:
     		187     188
	Fatal error: Shake block crossing node boundaries
	constraint between atoms (174,203)
	NNODES=2, MYRANK=0, HOSTNAME=og00.vcp.local
	NNODES=2, MYRANK=1, HOSTNAME=og00.vcp.local
	NODEID=0 argc=13
	NODEID=1 argc=13

(At this point I am using Grompp to set the run up for 2 CPUs).

Atoms 174 and 203, which are indeed on different nodes, are in fact in a
1,4 relationship with each other. Accordingly, I can't very well remove
the dihedral parameters or 1,4-pair interaction.

It is worth noting that neither atom 174 nor atom 203 is involved in any
of the repulsive bonding interactions that I have added manually
(described above).

I note that the following email has been sent before:

[gmx-users] How to circumvent shake-block errors ?

David van der Spoel spoel at xray.bmc.uu.se 
Fri Dec 14 09:17:03 CET 2001 

     Previous message: [gmx-users] How to circumvent shake-block errors
? 
     Next message: [gmx-users] RH 6.2 and standard kernel 
     Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] 



On Fri, 14 Dec 2001, Marc Baaden wrote:

>just came across an error I hadn't seen before ...
>looks like this:
>
>Walking down the molecule graph to make shake-blocks
>There are 4228 charge group borders and 1 shake borders
>There are 4228 total borders
>Division over nodes in atoms:
>  3633  3634
>Fatal error: Shake block crossing node boundaries
>constraint between atoms (3632,3637)
>
>I can see that there might be a problem splitting a constrained
>bond over 2 processors .. is there a way to cricumvent this ?
>(I have only 1 molecule in my simulation).
That's corrected. It simply isn't implemented. One option you have is to
use constraints only for the hydrogens and use flexible bonds for the
heavy atoms.

Groeten, David.



If (as I suspect) it is the fact that atoms 174 and 203 are on opposite
sides of the splitting that is causing me grief, how can I go about
fixing this? (There are no "constraints" that I have put in, only
"bonds", "angles", "dihedrals" and "pairs").

Thanks in advance for any help.

Regards,

Ben

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

Benjamin P. Roberts, B.Sc.(Hons)
Ph.D. Student

Department of Medicinal Chemistry
Victorian College of Pharmacy
381 Royal Parade
Parkville VIC   3052
AUSTRALIA

Phone	+61 3 9903 9674
Fax	+61 3 9903 9582
Mobile	+61 401 784 940
E-mail	Ben.P.Roberts at vcp.monash.edu.au

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