Skip to content

Force Fields

AutoPoly supports six force fields. Pass one as the force_field string to generate:

Force Field Value Best For Limitations
OPLS-AA "oplsaa" General organic polymers, biomolecules May lack parameters for exotic chemistries
LOPLS "lopls" Liquid-phase, better densities Less extensive than OPLS-AA
GAFF "gaff" Small molecules, drug-like compounds Charges need care (see below)
GAFF2 "gaff2" Updated GAFF, improved parameters Newer, less extensively tested
DREIDING "dreiding" Generic systems, metals, inorganics Generic parameters, less accurate for organics
COMPASS "compass" Commercial polymers, condensed phases Class II — needs the LAMMPS CLASS2 package

The force fields in detail

OPLS-AA — "oplsaa"

The safe default for organic systems. Extensively parameterized and validated, with wide coverage of organic functional groups and a large literature base. Best for commodity polymers (PE, PP, PS), organic liquids, and biomolecules.

Citation: Jorgensen et al., J. Am. Chem. Soc. 1996, 118, 11225–11236.

LOPLS — "lopls"

OPLS re-optimized for the liquid phase: better densities and enthalpies of vaporization for condensed systems. A good choice for polymer melts and solutions where density accuracy matters.

Citation: Siu et al., J. Chem. Theory Comput. 2012, 8, 1459–1470.

GAFF — "gaff"

The general AMBER force field — comprehensive coverage of organic and drug-like molecules, and the standard choice for small-molecule solvents and mixed systems. AutoPoly assigns Gasteiger charges automatically; for production runs replace them with AM1-BCC or RESP charges in system.in.charges.

Citation: Wang et al., J. Comput. Chem. 2004, 25, 1157–1174.

GAFF2 — "gaff2"

GAFF with improved bond, angle, and torsional parameters. Prefer it over GAFF for new projects when no legacy compatibility is needed.

DREIDING — "dreiding"

A generic force field with near-universal element coverage, including metals and inorganics. Ideal for exploratory structure generation and systems lacking specific parameters — but too approximate for quantitative production properties of organics.

Citation: Mayo et al., J. Phys. Chem. 1990, 94, 8897–8909.

COMPASS — "compass"

A Class II force field (with cross-terms) validated extensively against commercial polymers — excellent densities and mechanical properties for polyolefins, polyesters, and polyamides. Requires LAMMPS built with the CLASS2 package.

Citation: Sun, J. Phys. Chem. B 1998, 102, 7338–7364.

Choosing: a decision tree

System type?
│
├─ Polymer only
│  ├─ Commodity (PE, PP, PS) → OPLS-AA or LOPLS
│  ├─ Commercial/industrial  → COMPASS
│  └─ Novel/exploratory      → OPLS-AA, or DREIDING for a first look
│
├─ Small molecules only
│  ├─ Organic solvents  → GAFF or GAFF2
│  └─ Simple molecules  → OPLS-AA or GAFF
│
├─ Mixed (polymer + solvent)
│  ├─ Organic solvent → GAFF / GAFF2 (handles both)
│  └─ Aqueous         → GAFF (with TIP3P-style water) or OPLS-AA
│
└─ Special systems
   ├─ Metals, MOFs, inorganics → DREIDING
   └─ High-accuracy polymer properties → COMPASS

Common pitfalls

Don't mix force fields across components

One generate call uses one force field for everything in models. That is the correct behavior — never combine, say, an OPLS-AA polymer with a GAFF solvent in one box without validated combination rules:

# GOOD: one force field covers all components
generate(
    system,
    "compatible_system",
    [polymer, solvent],
    force_field="gaff",
)

Treat Gasteiger charges as a starting point

AutoPoly assigns Gasteiger (partial equalization) charges automatically for GAFF/GAFF2. They are fine for setup and testing, but for production runs compute AM1-BCC charges with Antechamber (or RESP with your QM package) and edit them into system.in.charges.

Don't use DREIDING for final numbers

DREIDING's generic parameters are for exploration. Generate and equilibrate the structure with DREIDING if nothing else covers your chemistry, then rebuild with a specific force field for production.

Validate before production

  1. Compare with literature for similar systems.
  2. Benchmark basic properties (density, energy) against experiment — if off by more than ~10%, reconsider the force field.
  3. Start small (10–20 chains, DOP ≈ 50) before scaling up.

See also