Some difficulties encountered with AM1 and PM3 calculations

AM1 and PM3 underestimate frontier interactions with respect to steric repulsions. Therefore, if two structures differ by ~1 kcal/mol, their calculated ordering is unreliable. Activation energies tend to increase with substitution, regardless of electronic effects. Atomic charges are sometimes unrealistic (in enolates, the negative charge is larger on C than on O). At van der Waals distances, acid-base and coulombic interactions can prevail over steric repulsions. At all distances, basicities are overestimated and nucleophilicities underestimated. This may lead to anomalous ion-molecule and transition structures in gas phase reactions. Transition structures are tighter than in ab initio calculations. Optimisations may give chemically unreasonable structures. Minimum energy paths are then difficult to obtain. Usually, but not systematically, PM3 gives more reliable structures and AM[ more realistic energies.