Selection of an MO Method

Regarding choice of a program for semiempirical calculations, it should be pointed out that AMPAC 1.0 is very similar to MOPAC 3.0, which is now out of date and has known bugs. Confusion has arisen in the mind of some users that one must use AMPAC to do AMI calculations. This is not true. MOPAC 4.0 is capable of MINDO/3, MNDO, and AMI calculations, whereas MOPAC 5.0 can be used for MINDO/3, MNDO, AMI, and PM3 calculations. 

Computed Properties Relative total energies Orbital energies Ionization potentials Net atomic charges Dipole moments Optimized geometries 

It has been discovered that certain irregularities can occur when computing conformational energies of flexible molecules. One would hope and expect that for any calculation, if you made several runs with nominally the same program on the same data, you would get the same results. However, 

Spurious results are more likely to occur when the hypersurface being traversed is flat. Because it is difficult to know ahead of time when such a flat surface exists for a complex molecule, the only safe course is to always try out and/or use the tightest available SCF convergence and minimization criteria offered by the programs. Papers reporting MO results from a geometry optimization procedure should be explicit about how the calculations were done. 

Quality of Results from Quantum Mechanical Methods 

---