Effect of solvent polarity on the potential of mean force between two molecular ions: MD simulation

Molecular Dynamics Simulations of Bimolecular Electron Transfer: Testing the Coulomb Term in the Weller Equation

Reliable estimation of the driving force for photoinduced electron transfer between neutral reactants is of utmost importance for most practical applications of these reactions. The driving force is usually calculated from the Weller equation, which contains a Coulomb term, C, whose magnitude in polar solvents is debated. We have performed umbrella sampling molecular dynamics simulations to determine C from the potentials of mean force between neutral and ionic donor/acceptor pairs of different sizes in solvents of varying polarity. According to the simulations, C in polar solvents is a factor of 2 more negative than typically calculated according to the Weller equation. Use of the π-stack contact distance in the Weller equation instead of the van der Waals radius recovers the correct value of C, but this is mostly fortuitous due to the compensating effects of overestimating the dielectric screening at contact and neglecting both charge dilution and desolvation.

Systematic and statistical error in histogram-based free energy calculations

A common technique for the numerical calculation of free energies involves estimation of the probability density along a given coordinate from a set of configurations generated via simulation. The process requires discretization of one or more reaction coordinates to generate a histogram from which the continuous probability density is inferred. We show that the finite size of the intervals used to construct the histogram leads to quantifiable systematic error. The width of these intervals also determines the statistical error in the free energy, and the choice of the appropriate interval is therefore driven by the need to balance the two sources of error. We present a method for the construction of the optimal histogram for a given system, and show that the use of this technique requires little additional computational expense. We demonstrate the efficacy of the technique for a model system, and discuss how the principles governing the choice of discretization interval could be used to improve extended sampling techniques.