10.5. Functional verification of Winmostar

Winmostar maintains its high quality by calling on open source software used by researchers around the world. On the other hand, there are some features that strongly depend on Winmostar’s own implementation, and here we show the validation results of such features. The verifications shown here have been performed prior to release.

10.5.1. Gromacs to LAMMPS force field file conversion

Winmostar generates a data file for LAMMPS using a generic force field by converting the top file for Gromacs. Here are the results of calculating the energy for the same system using Gromacs and LAMMPS with this conversion mechanism.

Relative energy difference between Gromacs and LAMMPS for each energy component

Force Field

Bond

Angle

Dihedral

Improper

Coulomb

Vdw

Total

UFF

2.57E-8

6.965E-9

1.17E-8

2.11E-8

7.21E-8

3.71E-8

7.55E-9

Dreiding

6.32E-9

4.25E-8

7.20E-9

5.37E-8

7.21E-8

3.62E-8

2.47E-8

GAFF

2.82E-8

3.06E-8

1.10E-8

2.97E-9

7.61E-8

9.51E-10

3.62E-9

OPLS-AA/L+GAFF

2.82E-8

3.06E-8

1.10E-8

2.97E-9

7.22E-8

1.67E-8

6.31E-8

OPLS-AA

1.64E-9

2.60E-8

1.10E-8

2.97E-9

7.22E-8

1.67E-8

5.93E-9

The detailed calculation conditions are as follows

  • One molecule of benzene, calculated with AM1-BCC charge

  • Free boundary condition, cutoff radius 20 Å, no constraints on bond length or angle

  • Gromacs uses the 5.0.7 double precision version, LAMMPS uses the March 9, 2016 version

  • Coulomb, Vdw includes 1-4 interactions

  • Dreiding’s Improper is calculated with Improper Harmonic because Gromacs does not have a function equivalent to LAMMPS’ Improper_style umbrella.

  • Since the 1-4 interaction scaling coefficients in Gromacs are processed internally in single precision, the single precision equivalent coefficients are also set in LAMMPS.

  • The files generated by mktop is used for OPLS-AA

From the above table, we can see that the energies of Gromacs and LAMMPS are in agreement by 8-10 orders of magnitude for all components. It is difficult to get an exact match due to minor implementation differences between Gromacs and LAMMPS (e.g., approximate calculations of various functions, parallel calculations, etc.).

Similar tests have also been conducted on systems focusing on each component of energy, multi-molecular systems, etc.