Single-Parameter Scaling Strategy for Force Field Optimization: A Case Study on Alkane Melting-Point Prediction

Classical
force fields critically determine the reliability of
molecular simulations, yet simultaneous optimization of their high-dimensional
parameters remains inefficient and risks disrupting cross-system compatibility.
Using alkane melting points as the target property, this work systematically
investigates how scaling individual parameters in multiscale force
fields affects prediction accuracy. Three linear alkanes (octane (C8), hexadecane (C16), and tetracosane (C24)) were evaluated with two all-atom (AA) models (L-OPLS, CHARMM36),
three united-atom (UA) models (TraPPE-UA, PYS, and OPLS-UA), and one
coarse-grained (CG) model (Martini 3). Unmodified force field tests
showed that L-OPLS, CHARMM36, PYS, and TraPPE-UA predicted melting
points within 7% of experimental values, whereas OPLS-UA consistently
overpredicted all alkane melting points, and Martini 3 exhibited significant
overprediction for C8. Parameter scaling revealed that,
for the UA models, bond force constant (kb) and angle force constant (ka) negligibly
impact melting points, whereas melting points positively correlate
with dihedral force constant (kn) and
Lennard-Jones (LJ) parameters (ε, σ).
Since LJ parameter scaling substantially perturbs liquid densities
and self-diffusion coefficientsunlike dihedral adjustmentsscaling kn emerges as the optimal strategy. Melting points
were corrected to experimental values with ± 10% kn scaling for TraPPE-UA and PYS, whereas
OPLS-UA required ≈50% reduction of kn. For AA models, partial charge scaling effectively tuned melting
points with minimal effects on the liquid properties. In Martini 3,
angle force constant scaling improved predictions (except for angle-lacking
C8). This systematic single-parameter scaling (SPS) delineates
structure–property relationships between force field parameters
and alkane thermophysical behavior, providing an efficient strategy
for the rapid refinement of established force fields.