Summary: DFT and Transition States

Dear Fellow Computational Chemists,
 	Here is the summary of replies I got from the list regarding
 DFT and Transition State description and a few thoughts for what they
 are worth. (I have assumed that everyone is happy for me to post their
 contributions, apologies if not).
 Tim Dransfield wrote:
 I would be really interested in hearing the summary of this one!  I've
 been doing DFT calculations on nitrate surfaces for several years, and
 B3LYP is far and away the best "bang-for-the-buck" for the stable
 species.
 I gave up on "pure" DFT for that reason. But, despite being the best
 of
 the "DFT" methods, B3LYP doesn't do all that well at barrier heights.
 Indeed, about five years ago it was standard practice to use BHandHLYP
 to
 calculate barrier heights, using the B3LYP geometries.  That's currently
 frowned upon, but I've yet to hear what one is supposed to do instead.
 I
 realize that your question is more geared towards the pure methods, but
 I'm very interested in what people have to say, all the same.
 Jean-Francois Truchon wrote:
 I'm a Ph.D. student in the group of Dennis Salahub. In his group,
 Emil Proynov has developped XC functionnals including the
 laplacian of the density and a correction for the kinetic
 energy. The name of the functionnals are BLAP and PLAP. In my study
 cases ( mainly hydrogen bonds ), those functionnals are very good
 ( comparable to G2 calculations ). Since I'm not in my office I
 don't have some references under my hands, but an article that
 might interest you has been written by Salahub, Chretien, Proynov
 about DFT and transition states calculation. I think it will be
 easy for you to find it.
 (Found it as -
 "Performance of density functionals for transition states"
 Salahub_DR, Chretien_S, Milet_A, Proynov_E, ABSTRACTS OF PAPERS
 OF THE AMERICAN CHEMICAL SOCIETY, 1998,Vol.215, No.Pt1, pp.186-COMP)
 Oystein Espelid wrote:
 lease send me a summary of you're replies, especially references to
 articles who covers this topic would be interesting. Personally it
 would surprise me if pure DFT (i.e. not including HF exchange) should
 generally perform bad, but certainly I expect one may find examples
 were this is the case. I have checked some reaction barriers and found
 that B3LYP results compares well with BP86 for organometallic reactions.
 Which is in line with the result of this reference:  V.R. Jensen and
 K.J. Børve J. Comput. Chem. 19, 917 (1998).
 Hans Martin Senn wrote:
 I would be interested in a summary of answers on this subject.
 As a small contribution, these are the few topical references I have:
 [1] R. V. Stanton, K. M. Merz, Jr., "Density functional transition
 states of organic and organometallic reactions", J. Chem. Phys. 1994,
 100, 434-443.
 [2] J. L. Durant, "Evaluation of transition state properties by density
 functional theory", Chem. Phys. Lett. 1996, 256, 595-602.
 [3] J. L. Durant, "Computational Thermochemistry and Transition
 States"
 in Computational Thermochemistry, K.K. Irikura, D.J. Frurip, Eds., ACS,
 Washington, DC, 1998, ACS Symposium Series, Vol. 677, pp. 267-284.
 Frank Jensen wrote:
 I havn't heard any arguments for why, but it is quite common.
 I would be interested if you get any responce on this.
 I think it is more a question of parameter fitting, pure
 dft gets barriers too low, and HF consistenly gets them too high.
 Thus a proper mixture of the two will get barriers in the right
 ballpark. And B3LYP has three fitting parameters....
 Me:
 I dug up a few references, but haven't had time to scour them.
 Brako S. Jursic has published many papers (particularly in '98 and '99)
 with PE surfaces for a number of reactions calculated at various levels
 >from HF up and with pure DFT with many functionals and hybrid DFT.
 One example where pure DFT works- Thomas Dargl and Wolfram Koch showed
 that pure DFT does OK for Transition states (in this case the
 Simmons-Smith reaction) where the barrier height is relatively high
 (40-60kJmol^-1). JCS Perkin Trans 2, 877 (1996).
 One further thing to consider when using DFT with PE surfaces is
 the inability to include attractive vdW forces which may mean that
 a pre-complex state 'before' the transition state may be missed.
 On the root causes of this under-estimation of TS energies, I don't
 personally believe that it is purely an exchange problem and the reason
 that adding exchange to give the hybrid functionals gives better results
 is just mixing something in that over-estimates TS energies to
 DFT that under-estimates TS energies - hence a cancellation of errors.
 If anyone who didn't reply to the original question has any more
 comments
 then I would welcome them (and of course people who replied to me
 earlier).
 Thanks to all who contributed.
 	Cheers,
 		Dave
 --
 ------------------------------------------------------------------------
 Dr. David W. Price,       Tel: +44 (0)118 9875123  extn 7415
 Department of Chemistry,  Fax: +44 (0)118 9316331
 University of Reading,    mailto:d.w.price %! at !% reading.ac.uk
 Whiteknights,
 READING                http://www.chem.rdg.ac.uk/g50/mmrg/dave/dave.html
 RG6 6AD
 U.K.
 ------------------------------------------------------------------------