Summary : Spin Contamination



 Dear all,
 I have asked a question about the spin contamination. I thank those 4
 people who replied to my question for helping me.
 Phil
 ------------------------------------------------------------------------
 APLINCOURT Philippe
 Laboratoire de Chimie Theorique                         ___
 UMR 7565 S2RMC                                      __n/   \__
 Faculte des Sciences                               \___     * )
 Domaine Scientifique Victor Grignard B.P. 239         /      /
 54506 VANDOEUVRE-LES-NANCY Cedex                     /    _  |
 Tel    : 03 83 91 20 00   Poste 35 52                \___/ \_\
 Fax    : 03 83 91 25 30
 E-mail : Philippe.Aplincourt ( ( at ) ) lctn.u-nancy.fr
 ------------------------------------------------------------------------
 QUESTION :
 ----------
 > Dear CCL'ers
 >
 > I am studying the OO bond cleavage in the secondary ozonide (in order to
 > form a biradical .OCH2OCH2O. ) at the MP2 and DFT level (singlet state).
 > The barrier heights computed with Gaussian94 are :
 >
 > UMP2/6-31G** = 43.4 kcal/mol
 > Annihilation of the first spin contaminant:
 > S**2 before annihilation      .8871,   after      .0785
 >
 > PMP2/6-31G** = 20.9 kcal/mol
 >
 > UB3LYP/6-31G** = 20.8 kcal/mol
 > Annihilation of the first spin contaminant:
 > S**2 before annihilation      .7684,   after      .0234
 >
 > I would know your opinion about the reliability of this PMP correction in
 > which the largest spin contaminant is annihilated. What about the spin
 > contamination in the UB3LYP method (PMP2 and UB3LYP are very close...) ?
 > References are welcome.
 >
 > I will summarize.
 > Thanks a lot.
 ANSWERS :
 ---------
 *******************
 Dear Phil,
 	the problem you described may very well require a multi-reference
 wavefunction. Whenever one deals with biradical a theoretical description
 is much more complicated than for ordinary closed shell species. My
 experience with diradicals is that neither MP2 (unrestricted or projected)
 nor DFT can properly describe these species. Instead one needs some
 multi-reference method. To say this in very simplistic way, the failure
 of single-reference based method (like MP2 or DFT) depends on the
 energy gap between frontier orbitals. To see how big is the problem
 in the particular case it is desirable to run a single-point MCSCF
 calculations and look at the weights of individual references. If
 the wave function is not dominated by just a single reference the DFT
 and MP2 are not of much value. There is not a unique prescription telling
 you what is the limiting weight for use of DFT or MP2. It depends on
 other factors as well. However, as a hint you can take that if the
 largest coefficient in MCSCF wavefunction is less than 0.9 DFT and MP2
 could be in trouble.
 	The system you are trying to describe can very well require
 a multi-reference description
 	As for the choice of MCSCF active space, the best thing is
 to do all-valence CAS, however, usually a much smaller active space
 which includes only the orbitals close in energy to frontier orbitals
 is sufficient to see this problem.
 	If MCSCF shows that a multi-reference wave function is necessary,
 a method like MR-ACPF, MR-AQCC, etc. (which are modification of MR-CI)
 could be useful.
 Best regards,
 =================================================================
 Petr Nachtigall
 J. Heyrovsky Institute of Physical Chemistry
 Academy of Sciences of the Czech Republic
 Dolejskova 3
 182 23 Prague 8			phone: (+420-2)-6605-2015
 Czech Republic			fax:   (+420-2)-858-2307
 			e-mail: Petr.Nachtigall ( ( at ) ) jh-inst.cas.cz
 =================================================================
 *********
 Hi,
 I don't know much about this topics. However as far as I know, the
 biradical (say singlet state) need to be treated using multiconfiguration
 wavefunction (MC... or CAS...) or Valence-bond theory.  Probably using DFT
 or MP2 with single determinant wouldn't be a clever move. I think
 one could begin with CASSCF and add dynamical correlation on top of it
 (CASPT2 by Prof. Roos as in Molcas or MCQDPT by Dr. Nakano as in
 Gamess-US). That might be a better thing to try in my openion.
 Oh, there's no such thing as wavefunction for DFT, i.e. no point to talk
 about Spin contamination in UB3LYP (ref. Pople, Gill, Handy, Int. J.
 Quant. Chem., 1995, vol.56., 303. )
                                         Bye,
                                         Y. Tantirungrotechai
 *********
 Hy,
  About spin contamination, you can take a look at
 http://www.auburn.edu/~youngd2/topics/spin_cont.html
 you will find some references.
 Regards.
 Olivier
         PPPPPPPP    II     II   MM     MM
         PP      P   II     II   MMM   MMM
         PPPPPPPP    II     II   MM M M MM
  _____  PP          II     II   MM  M  MM
 |       PP          Ii     II   MM     MM
 |
 | MARESCA Olivier                         |
 | PIIM - CNRS UMR 6633                    |
 | Spectrometrie et Dynamique Moleculaire  |
 | Campus Universitaire Saint Jerome       |
 | Service 542                             |
 | 13397 Marseille Cedex 20 - France       |
 |                                         |
 | Tel : +33  4.91.28.85.80                |
 | Fax : +33  4.91.63.65.10                |
 |                                         |
 | lougras ( ( at ) ) france-mail.com                 |
 |                                         /
 | http://crrmn2.univ-mrs.fr/~maurice     /
 |_______________________________________/
 **********
 Dear Phil,
 no guesses, but you might do a CASSCF and a CASSCF MP2 single point
 calculation on the TS structure(s) to check your results. Before you do,
 check the handbook for hints (e.g. CIS).
 Stefan
 ___________________________________
 Dr. Stefan Fau
 Fachbereich Chemie, AK Frenking
 Philipps-Universität Marburg
 35032 Marburg, Germany
 fau ( ( at ) ) chemie.uni-marburg.de