For researchers
How force-field data is organised
Fragments, reference systems, fits and parameters — why a parameter is only meaningful with its fit.
Force-field data on this site is not a flat table of numbers. Four things sit between "I have a structure" and "I have parameters", and each exists for a reason.
Atom types
A force field does not have parameters for carbon. It has parameters for a
particular carbon in a particular environment. The atom type encodes that:
c3_c2h1@ph is an sp²-ish carbon with two carbon and one hydrogen neighbour,
inside a phenyl fragment.
The @fragment half is not decoration. The same local environment inside a
different fragment is a different atom type with different parameters, because
the surrounding chemistry is different.
Fragments
A fragment is the unit assignment works on: a chemically meaningful piece — a phenyl ring, a carboxylate, a paddle-wheel — that carries the atom types for its atoms.
Assignment matches fragments against your structure in priority order, highest first. Priority defaults to the number of non-hydrogen atoms, so larger and more specific fragments claim their atoms before smaller and more generic ones can. Without that ordering, a benzene ring would be typed as six unrelated carbons.
Reference systems
Parameters are fitted, not measured. A reference system is a small model system with reference calculations against which fitting was done — a molecule small enough to compute accurately and representative enough to be worth fitting to.
Every parameter can be traced back to the reference it came from. This is the part that makes the data auditable: "where does this bond parameter come from" has an answer.
Fits
A fit is one parameterisation of one force field against a set of reference systems. A force field can have several.
This is the piece that surprises people. Two fits of the same force field will give different numbers for the same atom types, and neither is wrong — they were derived from different references, possibly with different settings, and possibly at different times as the reference set improved.
So a parameter is addressed as (fit, ric, atom types), never as
(force field, atom types). Dropping the fit from the address is how you get
numbers nobody can reproduce.
Parameters
The leaves. Each carries the atom types it applies to, the fragments those atom
types live in, a potential form (mm3 and friends), and the coefficients that
potential expects.
They are grouped by ric — the redundant internal coordinate family:
ric |
Term |
|---|---|
bnd |
bond stretch |
ang |
angle bend |
dih |
dihedral torsion |
oop |
out-of-plane |
cha |
charge |
vdw |
van der Waals |
Within a family there are refinements: bnd5 is a bond in a five-membered
ring, ang5 the corresponding angle, chargemod a charge correction spanning
two atoms, vdwpr a pair-specific van der Waals term. These live in the
type field beside the family, which is why a cha parameter can name two
atom types even though charges are usually per-atom.
What this means when you use it
- Record which fit you used, not just which force field.
- If a term comes back empty, the fit does not cover that combination. Choose a fit that does, or fit it yourself and deposit it.
- If you deposit parameters, deposit the reference system too. A parameter whose provenance is missing is a number, not a result.
See Assign a force field to a structure for the mechanics.