Forcefields as model chemistries: retiring the Forcefield step

Notes for the campaign that makes the LAMMPS step take its interatomic potential from the Model Chemistry step – classical forcefields, ReaxFF, OpenKIM potentials and machine-learned forcefields alike – and retires the Forcefield step as a step in flowcharts.

It started from a question about why production MLFF MD and labelling ran from hand-written scripts rather than through SEAMM (2026-10-11). Part of the answer was that a fresh MLFF checkpoint, with its per-run engine options, is awkward to reach through the Forcefield step. Paul’s conclusion: the Forcefield step does nothing a model chemistry cannot do.

Contents:

• The plan
  • Where things stand
  • The design
  • Type names
  • Atom types
  • Retiring the step
  • Phases
  • Tests
  • Decisions
  • Related
• Phase A: classical forcefields and ReaxFF (2026-10-12)
  • What was built
  • Verification
  • Notes for later phases
• Phase B: machine-learned forcefields (2026-10-12)
  • What was built
  • Validation
  • Left for later