Fatal Domain Decomposition Cell Error during NVT Equilibration

User discussions
1

GROMACS version: 2019.5
GROMACS modification: To my knowledge No

Hello everyone,

I am trying to prepare a system of 125 PEC6 (pentaerythritol tetrahexanoate) molecules for viscosity and free energy calculations.
But every time I am running a NVT simulation over 10 ns, which I believe to be necessary to calculate the viscosity via EMD-methods, I get the following error message with varying PME ranks:

Fatal error:

"1 particles communicated to PME rank X are more than 2/3 times the cut-off out of the domain decomposition cell of their charge group in dimension y.

This usually means that your system is not well equilibrated."

The initial configuration was produced by packmol, followed by an energy minimization. Thereafter I alternated between NpT and NVT to adjust the box size with 10 simulations รก 1 ns. After this I try to run the 10 ns equilibration, but everytime the fatal error occurs.
I tried diffrent variations (e.g. with higher temperatures, lower density, diffrent values for tau_t) but the fatal error persists.

Now I am hoping to find some insight through the help of someone more experienced than me.
Are there common mistakes which are leading to this kind of error?

Thank you in advance!
Jan

edit: Strangely, NpT Simulations seem to run through the 10 ns without error.

Following are the mdp options I am using:
;############################## Run control ##############################
integrator = sd
dt = 0.0005
nsteps = 20000000
;############################## Minimization #############################
emtol = 10
emstep = 0.01
niter = 20
;############################# Output Control ############################
nstcalcenergy = 100
nstenergy = 1000
nstlog = 10000
nstxout = 0
nstvout = 0
nstfout = 0

;########################### Neighborsearching ###########################
cutoff-scheme = verlet
nstlist = 10
ns_type = grid
pbc = xyz
rlist = 1.2
verlet-buffer-tolerance = 0.005
;############################# Electrostatics #############################
coulombtype = PME
coulomb-modifier = none
rcoulomb = 1.2
pme_order = 4
ewald_rtol = 1e-05
epsilon_surface = 0
fourierspacing = 0.12
;############################## van der Waals ##############################
vdwtype = cutoff
vdw-modifier = none
rvdw = 1.2
DispCorr = EnerPres
;########################### Temperature coupling ###########################
;tcoupl = is implicitly handled by the sd integrator
tc_grps = system
tau_t = 2.5
ref_t = 373.15
;############################ Velocity generation ############################
gen_vel = yes
gen_temp = 373.15
gen_seed = -1
;################################### Bonds ###################################
constraints = none
continuation = no
constraint-algorithm = lincs
lincs-order = 4

2

Hi,
a question during equilibration, did you check that pressure and temperature are equilibrated during the 10 1ns-simulation?

Best regards
Alessandra

3

Hello Alessandra,

the system seemed somewhat equilibrated but there are some high spikes in temperature and a drift right at the end. So I started another 10 ns alternating simulation from the last step, but the error occured after 2 ns. Again during the NVT run. The pressure seems to be fine, but the temperature shows some very distinct spikes every now and then, also during the NpT runs. I analyzed the energies in the edr during one of those spikes and it seems to be mostly related to a high bond energy.

Is my understanding correct, that this error usually occurs with instable systems and not so much because I chose the wrong decomposition cell distribution? Or a too high degree of parallelization?

My guess right now would be that there is a mistake somewhere in the parametrization of the molecules, even though I double and triple checked, or the thermostat is overregulating the system. I also tried a Nose-Hoover thermostat but the error occurs there as well.

Thank you very much for your reply
Jan

4

Hi,

Is my understanding correct, that this error usually occurs with instable systems and not so much because I chose the wrong decomposition cell distribution? Or a too high degree of parallelization?

Diffucult to generalize, but I think that in your case the source of error is due to the system not yet well equilibrated.
The reasons why a system has difficulty to equilibrate may be different. You wrote that you observed something strange in the bond energies. I can suggest (for example) to check that

  1. the time step you use is compatible with the force field. Force field can be parameterized using contraints for h-bond (or all bonds). Then you have to active the option in the mdp file
  2. if you have implement the force field, you can check that the bond parameter are corrected implemented (with correct units)

Best regards
Alessandra