MOPAC calculation on radicals - summary



 In reply to my question:
 > I'm interested in trying to model free radical alkene polymerisation
 reactions
 > using MOPAC (AM1, PM3).  There appear to be two possibilities for
 calculations
 > on radicals - the Dewar half-electron method (which is invoked by the
 keyword
 > DOUBLET) and UHF.   The results differ by a few kcal/mol.
 >
 > Can anyone advise me which option is likely to give the best results for
 > reaction thermodynamics and activation energies of a radical adding to a
 > double bond?  I'd also be interested in any comments on bond dissciation
 > energies estimated by these methods.  I'll summarise any replies.
 I received one answer - thanks to Andy Holder at Semichem (andy -8 at 8-
 semichem.com):
 "The third alternative is the best for your situation:  configuration
 interaction (CI)>  The reason is that both UHF and DHE suffer from serious
 inadequacies.  The DHE method simply isn't correct theoretically, although
 it works marginally for doublets.  In the case of UHF in the current
 semiempirical scheme, the results are not "spin pure" and have higher
 multiplicities mixed in.  (You can tell this by looking at the S**2 value,
 which is usually significantly different from the expected value.)  If the
 two points you are trying to compare are not the same amount of "spin
 impure" they are are not strictly comparable energetically.
 "In contrast, CI results are "spin pure" and they also allow
 orbtal switching
 during the course of a calculation on a reaction, as well as change in spin
 state.
 "MOPAC does have a CI module, but it is pretty limited and painfully slow.
 I
 suggest that AMPAC is a far better choice.  One of our collaborators is a CI
 expert and our CI module is the fastest, most complete, and easiest to use of
 anyone's."
 Semichem will supply a demo of AMPAC on request.  Subsequent CI calculations
 have
 shown similar results to the half-electron heats of formation for my radicals.
 Steve Rogers
 (Steve_Rogers -8 at 8- ici.com)