[gmx-users] Infinite polymer, pbc=full and LINCS

[Berk Hess]

>From: Malin Bergenstråhle <malbe at kth.se>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: <gmx-users at gromacs.org>
>Subject: [gmx-users] Infinite polymer, pbc=full and LINCS
>Date: Thu, 17 Nov 2005 10:41:05 +0100
>
>Hi,
>
>I am simulating a system of infinite polymer chains using pbc=full and 
>LINCS bond constraints.
>My system is unreliable and sometimes the simulations crash due to LINCS 
>complaining about too large bond rotations.
>I have now noted that it seems like LINCS gives warnings of the kind:
>
>Step 22525, time 67.575 (ps)  LINCS WARNING
>relative constraint deviation after LINCS:
>max 0.000029 (between atoms 22 and 23) rms 0.000004
>bonds that rotated more than 30 degrees:
>  atom 1 atom 2  angle  previous, current, constraint length
>     347    348   30.6    0.1000   0.1000      0.1000
>
>and the atoms are always situated on different sides of a periodic bound. A 
>question that now arises is whether this could be a bug with pbc=full or 
>whether there is a problem with my system. I am using  Gromacs 3.3.
>
>Any suggestions about how I could check this or whatever about whatever the 
>problem could be I would be grateful.

I have implemented pbc=full for Lincs for 3.3 and have never observed such 
warnings myself.
But it is quite strange that you bond-length is prefect before constraining
while the bond rotates 30 degrees.

It would help if you could restart your system just before a lincs warning
and write to xtc every step, then you can check what happens to a bond.

But how do you get such incredibly small constraint deviations?
Are you using an extremely small timestep?

Berk.