AMBER electrostatic potential energy

Our initial, small models of an isolated cellulose chain ranged from the dimer (cellobiose) to the octamer. The dynamics of these fragments have thus far been simulated only in vacuum, using different potential energy functions such as those of MM2(85) (9) and AMBER (10), with and without contributions from electrostatic terms and hydrogen bonds, etc. (The program DISCOVER, customized for carbohydrates and for operation on the Alliant FX/80 computer, has been used (12).) Generally, the time span for the simulations has been of the order of several hundred picoseconds to 1 nanosecond. 

The last term in the formula (1-196) describes electrostatic and Van der Waals interactions between atoms. In the Amber force field the Van der Waals interactions are approximated by the Lennard-Jones potential with appropriate Atj and force field parameters parametrized for monoatomic systems, i.e. i = j. Mixing rules are applied to obtain parameters for pairs of different atom types. Cornell et al.300 determined the parameters of various Lenard-Jones potentials by extensive Monte Carlo simulations for a number of simple liquids containing all necessary atom types in order to reproduce densities and enthalpies of vaporization of these liquids. Finally, the energy of electrostatic interactions between non-bonded atoms is calculated using a simple classical Coulomb potential with the partial atomic charges qt and q, obtained, e.g. by fitting them to reproduce the electrostatic potential around the molecule. 

Direct comparison of force fields to benchmark-quality CCSD(T) energies is complicated by the fact that most standard, workhorse force fields do not include polarization terms. This leads to errors, but these errors can be partially compensated by other errors. Hence, a force field that compares poorly to CCSD(T) benchmarks for a set of van der Waals dimers may still perform fairly well for condensed-phase properties, due to error cancellation. This is the rationale for obtaining atomic charges in the AMBER force field using restrained electrostatic potential (RESP) fitting (Bayly, 1993) to modest-quality Hartree-Fock/6-31G quantum chemical computations this method tends to overestimate dipole moments, but this is considered beneficial for simulations in water, to approximately cancel errors from neglecting polarization effects... 

---