normal Yet another thread about LINCS warnings in Polarized Martini Water

  • mboy@seas.upenn.edu
  • mboy@seas.upenn.edu's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
1 year 4 months ago - 1 year 4 months ago #8632 by mboy@seas.upenn.edu
Hello all,

Adding to the number of threads on some of the lincs errors I have seen in this thread. Below are some details about my simulation, a list of things I've tried, and a link to my input files. Any help would be much appreciated!!!!! Hopefully this consolidates all the recommendations and will lead to a one stop shop for folks seeing this issue!

I'm currently working toward martini coarse graining a strong polycation in a salt solution from an atomistic model. To best capture electrostatic interactions, I would like to use polarizable martini water and ions. I downloaded the force fields and mdp files from the tutorials and downloads sections on the MARTINI web page. However, when running the coarse grained simulation, I cannot get past lincs warnings and eventually the simulation crashes in my production run.

I've searched the MARTINI forum and tried the suggested solutions below to no avail:
- increasing vdw radius to 0.21 for PW insertion
- changing constraints to stiff bonds for energy minimization and reverting for equilibration
- turning off lincs warnings during equilibration
- Equilibrating using the Berendsen thermostat/barostat
- Slowing increasing timestep size to finish equilibration (10 fs to 15 fs to 20 fs)

Unfortunately, none of the above seems to work. When I increase to a 15 ro 20 fs step size, the polarized waters experience a slew of lincs errors like the ones below (full error list in the error.err file at the drive link below).

Program: gmx mdrun, version 2020.2
Source file: src/gromacs/mdlib/constr.cpp (line 224)

Fatal error:
Too many LINCS warnings (9807)
If you know what you are doing you can adjust the lincs warning threshold in
your mdp file
or set the environment variable GMX_MAXCONSTRWARN to -1,
but normally it is better to fix the problem

For more information and tips for troubleshooting, please check the GROMACS
website at www.gromacs.org/Documentation/Errors

Can anyone take a look to help us? I am running gromacs 2020.2 on a cluster with multiple gpus. I usually use 1-2 gpus per run, and 6-8 cpus per gpu, depending on resource availability. The files needed to recreate issues are available at the link below. The list of gromacs commands is in the mart-run bash script.

Thank you in advance for the help. Please let me know if you need anything more from my end.

Cheers,
Mike
Last edit: 1 year 4 months ago by mboy@seas.upenn.edu.

Please Log in or Create an account to join the conversation.

More
1 year 4 months ago #8646 by riccardo
For future reference: we exchanged a couple of emails with Mike and the issue turned out to be related with numerical instabilities caused by dihedral potentials, the one explained in the paper of Bulacu et al. ( doi.org/10.1021/ct400219n ).

In short, given a proper (e.g., function 1 of Gromacs) dihedral i-j-k-l, their derivative -- which needs to be compute when computing the force -- contains singularities when the angles i-j-k and j-k-l go collinear, i.e., they equal to 0 or 180. Hence, one has to prevent all such angles from becoming collinear. One of the most straightforward ways to do so is by using the "Restricted Bending Potentials" which have been devised and implemented in Gromacs by Bulacu et al. They can be used by defining an angle with function number 10 (see this Table manual.gromacs.org/documentation/2019/re...ts.html#tab-topfile2 from the GMX manual).

Please Log in or Create an account to join the conversation.

More
2 weeks 6 days ago #9224 by a786pati
Some time inserting water molecules in the well-packed system leads to overlapping. Also one can increase box dimension to avoid such warnings. Try recommended setting using PME with increasing the time step from 1fs to 20fs.

Please Log in or Create an account to join the conversation.

Time to create page: 0.091 seconds