A major complication in hybrid QM/MM methods is the treatment of the frontier between the quantum part, describing the reactive region, and the classical part, describing the environment. Two approaches to this problem, the “link atom” method and the “local self-consistent field” (LSCF) formalism, are compared in this paper. For this purpose, the LSCF formalism has been introduced into the CHARMM program. A detailed description of the two approaches is presented. The results of semiempirical calculations of deprotonation enthalpies and proton affinities of propanol and a tripeptide with different treatments of the frontier bond are compared. Particular emphasis is placed on the effect of an external charge. It is shown that the choice of the QM/MM electronic interactions included in the frontier region is of considerable importance in determining the electron distribution of the QM region and the overall energy. The link atom and LSCF methods are generally of similar accuracy if care is taken in the choice of the frontier between the QM and MM regions. QM and QM/MM geometry optimizations of ethane and butane are also compared. The introduction of a link atom in the frontier bond is shown to lead to distortions of the internal coordinates unless the frontier bond is treated in a special way. A number of practical points concerning the choice of the frontier between the QM and MM regions are presented. It is not advisable to remove classical charges from the interactions with a subset of the quantum atoms, as this can introduce significant errors in the energy computations. The presence of a large charge on the classical atom involved in the QM/MM frontier also adversely influences the energy, especially with the LSCF method, and it is therefore advised to select classical frontier atoms with small charges. Charged atoms which are not directly bound to the QM frontier but which are in its proximity are also shown to be a source of errors, and it is advised to introduce warning messages in QM-MM codes when such a situation arises.

We have investigated the relative orientational preference of two benzene and two toluene molecules in a vacuum and in water, by means of free energy calculations. The gas-phase simulations reveal that, whereas the T-shaped benzene dimer is 0.78 kcal/mol lower in free energy than its stacked homologue, the sandwich arrangement of the toluene dimer is preferred over the T-shaped structure by 0.18 kcal/mol. MP2/TZP ab initio binding energies, evaluated for both dimers, were found to be consistent with the molecular mechanical estimates, hence suggesting that the van der Waals and the electrostatic contributions to the macromolecular force field employed herein are well balanced. We further note that our results agree quite nicely with the experimental binding energies of Neusser and Krause, obtained from breakdown measurements. The tendency witnessed in the gas phase is magnified in an aqueous solution, with differences in free energy between the T-shaped and the sandwich arrangements of the benzene and the toluene dimers equal to −1.47 and 1.12 kcal/mol, respectively. The calculated association constants and osmotic second virial coefficients also correlate reasonably well with the experimental data of Tucker and Christian. The conflict between the orientational preferences of the benzene and the toluene dimers is suggestive that trends in “π−π” interactions in proteins should be rationalized by other factors than simple electrostatic/dispersion considerations. The analysis of Phe−Phe pairs in protein crystallographic structures sheds light on the influence of both sterical hindrances and ancillary interactions between the aromatic moities and neighboring functional groups on the orientational preference of the phenyl rings.