CCL:G: transition state searches with CASSCF ???

 Sent to CCL by: Wai-To Chan [chan]-[curl.gkcl.yorku.ca]
 <<<<<<<
 I wish to know whether it is possible to perform transition state searches at
 the CASSCF level of theory using Gaussian 03 rev C02. I asked the same question
 to the helpdesk of Gaussian but I gave up, after waiting more than 5 weeks.
 I would like to study a bi-molecular reaction in which two (neutral) molecules
 are transformed into two radicals (reversed dismutation reaction).
 Neither B3LYP nor X3LYP are capable to locate the TS, while unrestricted MP2
 gives the correct TS but unusual high barrier (delta H activation) and delta H
 are obtained.
 Thanks for sharing your thoughts how to handle this problem.
 Kind regards,
 Theo de Bruin
 theodorus.de-bruin#,#ifp.fr
 >>>>>>>>>>>>>>>
     If your TS is one linking between a pair of netural reactant
 molecules and two radicals and if it can be located with UHF-MP2
 then I presume it is a diradicaloid structure. If the spin-contamination
 of the UHF solution of the diradical is moderately higher than zero (  > 0.5)
 I will be surprised you can't locate the TS with DFT. Try using
 BHandHLYP starting from your UMP2 TS structure. Use a combination of
 guess=mix and stable=opt to obtain a spin-contaminated UDFT solution
 and restart the transition state search. This DFT functional has a larger mix
 of Hartree-Fock exchange and should stand a better chance of
 sucess than B3LYP.
   To run CASSCF you may start from the UHF solution from the UMP2
 results. Run a pop=naturalorbitals job to get the unrestricted
 orbital occupancy. Then run a UNO-CASSCF job (keyword: cas(uno,n,m))
 with guess=read specified to read off the natural orbitals from the
 checkpoint file.  For selection of the CASSCF active space
 pick the orbitals with occupancy in the range of 0.02--1.98.
   CASSCF optimization is a real pain. I will only use it as a last resort.
 Wai-To Chan