[gmx-users] LINCS warnings for a single molecule in otherwise stable system

[Mark Abraham]

Hi,

While we can't exclude the possibility of a code bug, your analysis and
results suggest the problem lies elsewhere. Is the EG model stable if you
e.g. put one in a box of decane? It's a stab in the dark, but if EG was
parameterized only for aqueous-phase properties, then all bets are off once
it gets into the membrane...

Mark

On Thu, Jul 2, 2015 at 12:35 PM Cara Kreck <cara.kreck at student.curtin.edu.au>
wrote:

> Hi everyone,
>
> I am trying to run a simulation of a lipid bilayer with a multi-component
> solution. One of the components is ethylene glycol. Some of the components
> have a tendency to wriggle their way into the lipids. When an EG molecule
> does this, it tends to get uncomfortable and trigger some LINCS warnings.
> It only happens with a single EG molecule at a time. It never happens with
> any of the other components or the EG molecules in the solution. It never
> leads to anything else becoming unstable or the system blowing up, it just
> causes a crash when the warning limit is reached. The strangest thing is
> when the simulation is continued from the checkpoint, the problem goes away
> for a while but eventually reoccurs within a few hours with a completely
> different EG molecule wedged in the membrane.
>
> I have tried increasing both lincs-order and lincs-iter, but it still
> happens. I have tried increasing lincs-warnangle, but that just lets the
> angle fluctuations get bigger before the job is killed. I tried using a
> slightly different topology for EG and I thought that had helped for a
> while. The first time the angle fluctuations started, they stayed around
> the 30 degrees mark and went away by themselves after around 600 warnings,
> so the job was then able to run for the full 24 hours allowed. The last 2
> jobs I have submitted have been killed prematurely by LINCS warnings that
> didn't go away by themselves though. This is very frustrating as I am
> running on a very busy cluster and it can take 24-48 hours for a job to
> progress through the queue, only to then be killed within a couple of hours
> of running. The only thing I can think of now is to increase
> lincs-warnangle by a small amount so that the smaller fluctuations are
> given a chance to sort themselves out before being killed.
>
> If anyone else could give me any suggestions, I would really appreciate
> it. I am using GROMACS 4.6.7 and the GROMOS53A6 FF. Here is a copy of my
> mdp file:
>
> integrator               = md
> dt                       = 0.002 ; 2fs
> nsteps                   = 500000000 ; 1000 ns
> comm_grps                = DOPC !DOPC
> nstxout                  = 1000
> nstvout                  = 0
> nstlog                   = 1000
> nstenergy                = 1000
> energygrps               = DOPC !DOPC
> nstcalcenergy            = 5
> nstlist                  = 5
> ns_type                  = grid
> pbc                      = xyz
> rlist                    = 0.8
> coulombtype              = Reaction-Field
> rcoulomb                 = 1.4
> epsilon_rf               = 62
> vdwtype                  = Cut-off
> rvdw                     = 1.4
> tcoupl                   = berendsen
> tc-grps                  = DOPC !DOPC
> tau_t                    = 0.1 0.1
> ref_t                    = 303 303
> Pcoupl                   = berendsen
> pcoupltype               = semiisotropic
> tau_p                    = 1.0 1.0
> compressibility          = 4.6e-5 4.6e-5
> ref_p                    = 1.0 1.0
> gen_vel                  = no
> gen_temp                 = 150
> gen_seed                 = 173529
> constraints              = all-bonds
> constraint_algorithm     = lincs
> lincs-order              = 6
> lincs-iter               = 2
> continuation             = yes
>
> Thanks,
>
> Cara
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>