Useful Tips for Running MacSpartan Plus Computations

• Always enter the keyword PRINTLEV=3 in the basic computational menu.  You get a lot of useful output this way.  See the red colored notes in the example run for nitrous oxide.
• Always use AM1 or PM3 for test computations, especially geometry optimizations.  Ab Initio 3-21G or 6-31G* are much slower.
• PM3 should be useful for semiempirical computations involving numerous inorganic atoms, but not for all.
• AM1 and PM3 semiempirical MO computations yield heats of formation, but do not trust these over much.  These are fairly good for a variety of molecules based upon first-row elements (unless they contain small rings or large non-bonding interactions), but are
• In general, FREQUENCY computations will yield vibrational frequencies that are 15-20% too large.  Intensities also are tyipcally difficult to get correctly, unless density functional or MP2 computations (not available on MacSpartan Plus) are used.  It is customary to scale computed vibrational frequencies by a stated amount, for example, by multiplying by 0.88.
• FREQUENCY computations should only be done at a geometry that is optimized at the same level of theory for which the FREQUENCY computation is run.  If this is not done, you will be obtaining a second-derivatives analysis at a geometry that is not a saddle point.  Result = GIGO.
• FREQUENCY computations are the best proof that a geometry optimization gives an energy minimum.  If you have optimized to a minimum, all computed frequencies will be greater than zero.  If you have optimized to a saddle point (transition state) one and only one frequency will be less than zero (imaginary).  Higher order stationary points are known, but do not correspond to minima or transitions states.
• Beware of starting computations that you expect to be low symmetry with a high symmetry geometry.  MacSpartan will not break symmetry unless you use the keyword NOSYMTRY in the basic computational menu.  NOSYMTRY makes the computation slower in many cases, but is a good way to detect Jahn-Teller (Pierls) distortion where it occurs.  For example, if you start a computation on cyclobutane as a rectangle, it will give an (incorrect) planar geometry.  If you break its symmetry to deplanarize, it will give a reasonable puckered geometry.  How do you know?  Often, you do not.  If a computation is small enough, it often is good to try both a NOSYMTRY run and a standard run.  If the computation is large, it is typical to do a standard run without the NOSYMTRY keyword.