MPI_Abort error when using Plumed with Lammps

LAMMPS LAMMPS Mailing List Mirror
1

Dear Experts,

I’m sorry to trouble you. I tried to make metadynamics calculation with covlars package, but it end up with the error message:

colvars: User-controlled feature “lower boundary” may not be enabled as a dependency in colvar dist.
colvars: …required by “grid” in colvar dist
colvars: Error: Failed dependency in colvar dist.
ERROR on proc 0: Fatal error in the collective variables module.
(…/colvarproxy_lammps.cpp:320)
Last command: run 2000000
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0

My operating system is Ubuntu 16.04 and my lammps version is “lammps-16Mar18”, installed with mpich-3.2.1 and fftw-3.3.8. I think I have already complied LAMMPS with MPI, and LAMMPS calculation run if I do not call COLVARS. But I don’t find where I make mistake. Would you mind point out it?

Thank you very much, and I look forward to hearing from you.

colvars.dat (526 Bytes)

2

changed the subject to reflect, that this question is about COLVARS
and not Plumed.

dear zhong wen, if you want the right people to respond to your posts,
you should use a correct subject line.

axel.

3

Thanks, Axel, for the correction.

Zhongwen, it looks like you adapted a PLUMED input file, try looking at the documentation, starting from the first example:
http://colvars.github.io/colvars-refman-lammps/colvars-refman-lammps.html#sec:colvars_crash_course

The most important fixes are:

  • please use multiple calls to atomNumbersRange begin-end (i.e. not begin:end);

  • please provide upper and lower boundary for the metadynamics PMF;

  • please check carefully the meaning of the keywords, for example you don’t want 10 kcal/mol-high Gaussians (assuming you had real units);

  • lastly, please do not put in a keyword explicitly, if you agree with the default value: the more you type, you just increase your changes of making mistakes. Ideally, you can reduce the metadynamics configuration should be just:

metadynamics {
name metad
colvars dist
hillWeight 0.001 # Or what you want
writeFreeEnergyFile on
outputEnergy on
writeHillsTrajectory on
}

Note, in particular, that Colvars and Plumed use slightly different meanings:

  • Plumed’s SIGMA is the parameter of the Gaussian, and is in the unit of the variable, thus you may end up with large and small numbers in a multidimensional calculation;
  • Colvars’ hillWidth equals 2 sigma’s and is in the units of the “width” parameter of the variable (which is 1 by default), thus you can leave it at the default, or increase it e.g. to 2 or 3 for smoother Gaussians. The code will print the value of sigma in the standard output for comparison with the Plumed convention, though.