The vibration frequencies predicted by the AM1 model

We analyse vibrational frequencies of 168 compounds with the AM1 model concerning its experimentally observed gaseous frequencies. Stretching of CH, NH, OH and CO bonds, its related bending frequencies, and the CC frame movements are the studied vibrations. The results show problems with the AM1 vibrational splittings. Often symmetric stretching frequencies, like in CH3, CH2 and NH3, appear switched with the corresponding antisymmetrical ones. Among the studied vibrations many stretchings are overestimated, while bendings oscillate around experimental values. Fluorine stretchings, NN, OO, CH, double and triples CC bonds and cyclic hydrocarbon breathing modes are always overestimated while torsions, umbrella modes and OH/SH stretching are, in average, underestimated. Graphical analysis show that compounds with the lowest molecular masses are the ones with the largest difference to the experimental values. From our results it is not possible to fit confortably the calculated frequencies by a simple linear relationship of the type, n(obs)=a*n(AM1). Better aggreement is obtained when different curves are adjusted for the stretching and bending modes, and when a complete linear function is used. Among our studies the best obtained statistical results are for CH, NH and OH. The conclusions obtained in this work will improve the AM1 calculated frequencies leading to accurate results for these properties