“…This issue has been addressed by including special hydrogen bond corrections in MNDO-type methods. ,− , In contrast, the OM x methods treat hydrogen-bonding interactions even without such corrections reasonably well, ,,, while inclusion of dispersion corrections generally further improves the accuracy. , One should note, however, that the addition of empirical attractive dispersion corrections to any semiempirical Hamiltonian parametrized without such corrections will inevitably deteriorate the accuracy of the computed heats of formation (which will become too small), while the computed relative energies may become more or less accurate. , Hence, it is more consistent to reparametrize the Hamiltonian with inclusion of dispersion corrections. This has so far been done only in PM7, which however suffers from error accumulation in very large noncovalent complexes, , and in the proof-of-principle MNDO-F method, which still has large errors in heats of formation.…”