PLUMED

RGPOT is a collective variable whose value is the energy of an rgpot potential on the atoms it lists. BIASVALUE adds that energy, its forces and its virial to the MD engine.

LOAD FILE=libRgpotPlumed.so
ene: RGPOT ATOMS=1-256 POTENTIAL=SuttonChen OPTIONS={preset platinum cutoff 5.0} TYPES=Pt
b: BIASVALUE ARG=ene
  • ATOMS are the atoms the potential sees.

  • POTENTIAL and OPTIONS are a name and its key value options from :doc:`../reference/potentialnames\`, the same words as LAMMPS pair_coeff. Option values are Angstrom and eV.

  • TYPES is one element symbol or atomic number for every atom, or one per atom in ATOMS order, separated by commas: TYPES=Cu,Ni,Ni,Cu or TYPES=29,28,28,29.

  • NOPBC evaluates the atoms as an isolated cluster.

Units

Positions, the cell, the value and its derivatives are in the PLUMED units: nm and kJ/mol by default, as in GROMACS, or whatever UNITS sets. They are converted to Angstrom and eV through the PLUMED Units object. The rgpot eV is 96.48533212331002 kJ/mol. The PLUMED UNITS ENERGY=eV is 96.48530749925792 kJ/mol, so a value printed in that unit is 2.6e-7 larger, relative, than the rgpot energy in eV. The forces the MD engine receives do not depend on UNITS. Natural units stop with an error.

plumed driver reads lengths in its --length-units, nm by default. UNITS only sets PLUMED’s own units. A file in Angstrom needs --length-units A.

Cells

With a periodic box, RGPOT passes PLUMED’s cell to the potential, orthogonal or triclinic, with the positions wrapped into it. The potential takes the minimum image, so its cutoff must be shorter than half of every cell width. The box derivatives are \(-V\sigma\) from the potential’s stress, so the pressure under BIASVALUE includes the rgpot virial and NPT runs are correct. A potential without a stress stops on a periodic cell with returns no stress instead of returning a wrong virial.

With NOPBC, or with no cell, the atoms sit in a cube wider than twice their extent plus the cutoff, so no periodic image interacts. The box derivatives then come from the positions alone.

PLUMED 2.9 reports an error raised during a step only as An error happened while calculating; RGPOT writes the reason to the PLUMED log first.

Build and tests

The shared library is libRgpotPlumed.so, built when pkg-config finds PLUMED and linked with libplumedKernel. -Dwith_plumed=enabled is an error when those headers are missing. Without PLUMED the tests are not defined and meson prints why.

engines/plumed/run_plumed.py runs these meson tests through plumed driver:

  • plumed_driver: the LJ dimer against rgpot-lj.

  • plumed_debug_forces: --debug-forces finite differences of the forces for LJ, Sutton-Chen, Stillinger-Weber and EMT, on clusters and triclinic cells, and for EAMTable with -Dwith_eam_table=true. EMT on a periodic cell has to stop with the stress error.

  • plumed_box_derivatives: the nine box derivatives on triclinic cells.

  • plumed_units: one cell in nm and kJ/mol, and in Angstrom and eV.

  • plumed_api: energy, forces and virial against rgpot itself. The C function is rgpot_sutton_chen_energy_forces. The Python classes are LJPot and StillingerWeberSi. The Python part needs -Dwith_python=true and otherwise prints a skip.

The numerical column of --debug-forces is noisy at a few 1e-5 of the largest derivative on cells of a few hundred atoms, so the tests allow 5e-5. A central difference with a chosen step agrees with the analytic forces to about 1e-8.