CONTINUUM ELECTROSTATICS IN A COMPUTATIONAL UNIVERSE WITH FINITE CUTOFF RADII AND PERIODIC BOUNDARY-CONDITIONS - CORRECTION TO COMPUTED FREE-ENERGIES OF IONIC SOLVATION
Rh. Wood, CONTINUUM ELECTROSTATICS IN A COMPUTATIONAL UNIVERSE WITH FINITE CUTOFF RADII AND PERIODIC BOUNDARY-CONDITIONS - CORRECTION TO COMPUTED FREE-ENERGIES OF IONIC SOLVATION, The Journal of chemical physics, 103(14), 1995, pp. 6177-6187
Many simulations of the free energy of hydration of an ion in a polar
solvent are performed in truncated periodic systems in which electrost
atic forces are truncated and periodic boundary conditions are used to
eliminate surface effects. Simulations allow accurate calculation of
the properties of the truncated Hamiltonian because there are no long-
range forces present. However, in order to compare with the real unive
rse, it is necessary to correct for the effects of both truncated pote
ntials and periodic boundary conditions. A method of calculating the c
ontinuum dielectric properties of a truncated periodic system is deriv
ed and applied to the case of a single ion in a polar solvent. If the
continuum model is accurate at distances where the effects of the trun
cated potentials and periodic boundary conditions are significant, the
n the correction will be accurate and it will include all the effects
of the truncated ion-solvent and solvent-solvent potentials, as well a
s the effects of the periodic boundary conditions. When the simulation
s of Straatsma and Berendsen [J. Chem. Phys. 89, 5876 (1988)] for the
Ne(aq) to Na+(aq) transformation at 298 K and 0.1 MPa are corrected in
this manner, the calculations with different cutoff radii between 0.9
and 1.2 nm and different periodic cell size give the same corrected r
esult within experimental error. Similarly, Aqvist's [J. Phys. Chem. 9
8, 8253 (1994)] calculations for a different model of the Ne(aq) to Na
+(aq) transformation with periodic boundary conditions and varying cut
off schemes give self-consistent results, but these results are differ
ent from his results for the same model using the surface constrained
all atom solvent method. Possible reasons for this discrepancy are dis
cussed. (C) 1995 American Institute of Physics.