SUMMARY: Second order TS's crossroads



    Hi!
 In summary, the following references were suggested:
 1) F. Jensen, "Transition Structure Optimization Techniques",
 in The Encyclopedia of Computational Chemistry,
 (1998) 3114-3123.
 2) J. Mol. Struct. 1995, 338, 117-130
 3) J. Chem. Phys. 104, 1774-8 (1996)
 4) J. Chem. Phys. 104, 8507-15 (1996)
 Below are my original request and full replies I have got.
 I thank to Frank Jensen, Per-Ola Norrby, Trevor Power,
 Stefan Fau, David Woon and George Vacek for the replies !
 best regards, Valentine.
 ----------------------------------------------------------
   Dear CCL'ers,
 There are two questions concerning second order TS's, which I
 found being rather contradictory in the literature. So:
 1) in some articles one may read something like that: "... we
 found that reaction proceeds through the second order transition
 state ... and gives desired products ..."
 while the other authors state: "... second order saddle points
 are of no interest for chemical reactions ..."
 Where is the truth? May a chemical transformation proceed through
 a second order transition state?
 2) Next,
 If there is a second order TS, does it mean that there are
 also two independent first order TS's representing each of the
 imaginary vectors of the former? Is the statement always correct?
 Any suggestions and references are very welcome.
 best regards,
 Valentine.
               -------------------------------
 ---------------------------------------------reply1------------
 from: Frank Jensen, <frj &$at$& dou.dk>
         Depends on whether you have a 'static' or 'dynamic' picture.
 In the static picture nuclei move infinitely slowly, i.e. along the IRC.
 Since there is no unique IRC from a second order saddle point, it
 may be unclear what reactant/product it connects. Furthermore, the
 energy of a second order saddle point can always be lowered along one of
 the imaginary modes, leading to a first order saddle point, a real TS.
 Again, unless there is symmetry in the system, there is no unique
 way of getting from a second order saddle point to a TS, i.e. there may be
 one, two or more real TSs, connecting the same reactant/product or
 different reactants/products.
 In this picture, second order saddle points have no chemical significance,
 as they always are higher in energy than the real TSs.
 In the dynamic picture nuclei are moving on the PES, with a kinetic
 energy depending on the temperature. In this case the reaction samples
 a larger fraction of the surface, and a representation of the reaction
 requires running many trajectories, and suitable averaging. In this case
 one can imagine a second order saddle point being sufficiently low
 in energy that it makes a significant contribution to the reaction, i.e.
 is traversed by a good fraction of the trajectories. In this case it is
 of chemical significance.
 In principle one could imagine two minima being connected dynamically
 by a second order saddle point, but without having a real TS connecting
 the two minima (there may be another path with real TSs going through
 one or more intermediates). In such a case the second order saddle point
 may be very important chemically, but I have not yet seen an explicit
 example of this.
         Frank
 ---------------------------------------------------------------
 ---------------------------------------------reply2------------
 from: Per-Ola Norrby, okpon &$at$& pop.dtu.dk
 >1) in some articles one may read something like that: "... we
 >found that reaction proceeds through the second order transition
 >state ... and gives desired products ..."
 >
 >while the other authors state: "... second order saddle points
 >are of no interest for chemical reactions ..."
 >
 >Where is the truth? May a chemical transformation proceed through
 >a second order transition state?
         The second is more true than the former.  All low-energy
 points on the potential energy surface will be visited, but the lower
 the energy is, the higher the population will be.  If there are any
 second order TS's, there MUST also be some first order TS's, and they
 must have a lower energy, therefore be more important.  You get very
 close to the truth if you simply ignore higher-order maxima.
 >2) Next,
 >
 >If there is a second order TS, does it mean that there are
 >also two independent first order TS's representing each of the
 >imaginary vectors of the former? Is the statement always correct?
         No.  If you verify that the second order maximum leads to the
 correct reactant and product, there must also be at least two
 flanking TSs (there could also be intermediates etc, you never know),
 but there is no good way to tell what the vector in the two flanking
 TSs will be.  If you have a very simple potential energy surface, you
 might try projecting the vector from reactant to product in the
 2D-space described by the two eigenvectors of the 2nd-order max.  The
 resulting vector could well be similar to BOTH the vectors of the
 TSs, whereas the orthogonal direction might connect the 2nd order max
 to the TSs, but you have no guarantee.  The high order max is not a
 good starting point for finding the 1st order TSs if you're using a
 Newton-Raphson-based method, but there are several other methods that
 might do it.  Frank Jensen wrote a review that includes some suitable
 methods, if I remember correctly it was in "Encyclopedia of
 Computational Chemistry".
         Per-Ola
 ---------------------------------------------------------------
 ---------------------------------------------reply3------------
 from: Trevor D. Power, tdp0006 &$at$& unt.edu
 Valentine,
         I actually ran into this head first when I was doing modeling of
 organolithium reactions about a year ago. Many theoreticians will tell you
 that a second-order transition state is a theoretical entity and cannot
 be called a transition state nor a molecule anyway.  If fact the only
 correct designation would probably be a second-order saddle point.  When
 one is looking for a transition state that describes a particular reaction
 and they instead find such a maximum, this usually means that the reaction
 they are trying to model is actually not concerted and may go through two
 transition states and 1 intermediate.  I would think that one needs to go
 to extreme measures to find a second-order saddle point; such as symmetry
 or geometry restrictions (a partial optimization --- which does not make
 a lot of sense).
         I would expect that those papers describing a reaction to go
 through second-order saddle points are in journals not set up to
 rigorously review computational manuscripts (like J. Org. Chem. for
 example).  You would do well to find papers on this subject in J. Phys.
 Chem or J. Chem. Phys.  Alan Isaacson, Miami University, wrote a few
 papers (none are reviews though) that began my literature search on this
 topic.
         I would agree with the "other authors" as stated below.  The
 principle of least motion is violated in a second-order saddle point.
 Hope my posting does not spark too many fires out there.
 Regards,
 Trevor D. Power
 ---------------------------------------------------------------
 ---------------------------------------------reply4------------
 from: Stefan Fau, fau &$at$& qtp.ufl.edu
 Hi Valentine,
 to 1)
 I think that 2nd order TSs are of little importance, chemically.
 If you have a 2nd order TS with one imaginary mode connecting
 your reactants and products, you can optimize along the other
 imaginary mode and get a 1st order TS that should still connect
 your reactants and products. (I don't know of any exceptions to
 that, but maybe there are a few.) Due to the optimization, the energy
 of this 1st order TS is lower, and therefore more reacting molecules
 should pass along here (assuming that the effects of nuclear motion
 are small enough).
 Classically, every point (geometry) on the potential energy
 hypersurface which is below the molecules energy is accessible
 by that molecule. So if you supply your molecule with lots of energy
 lots of geometries are possible. If you give just enough energy to
 keep
 a very slow reaction going virtually all product molecules will have
 passed the 1st order TS.
 to 2)
 A 2nd order TS is not generally connected to four 1st order TSs (two
 for each imaginary mode). I once had a 2nd order TS connected to
 two mirror-symmetric 1st order TSs and two minima. The first order
 TSs connected the same minima. I think that this scheme might be
 common with 2nd order TSs of higher symmetry.
 The example is published in
 S. Fau, G. Frenking, Theochem - J. Mol. Struct. 1995, 338, 117-130.
 I think the 2nd order TS was a C2v structure of B2CH4, the 1st order
 TS was non-planar Cs and the minimum was planar Cs.
 Stefan
 ---------------------------------------------------------------
 ---------------------------------------------reply5------------
 from: David E. Woon, woon &$at$& hecla.molres.org
 I agree with the other respondents about the importance (or lack
 thereof) of second-order saddle points, with one small exception
 that was not mentioned. If an nth order saddle point is defined
 as a stationary structure (all first derivatives zero) with n
 negative eigenvalues of the second derivative matrix, then a
 second order saddle point is still a true TS if the eigenvalues
 are degenerate. This occasionally occurs due to symmetry.
 Dave Woon
 ---------------------------------------------------------------
 ---------------------------------------------reply6------------
 from: George Vacek, vacek &$at$& schrodinger.com
 >         Many theoreticians will tell you
 > that a second-order transition state is a theoretical entity and cannot
 > be called a transition state nor a molecule anyway.  If fact the only
 > correct designation would probably be a second-order saddle point.  When
 > one is looking for a transition state that describes a particular reaction
 > and they instead find such a maximum, this usually means that the reaction
 > they are trying to model is actually not concerted and may go through two
 > transition states and 1 intermediate.  I would think that one needs to go
 > to extreme measures to find a second-order saddle point;
 A well stated response Trevor.  A couple papers where theoreticians concede that
 experimentalists may have seen evidence of a second order saddle point are:
 ``The Anomalous Behavior of the Zeeman Anticrossing Spectra of \~A
 $^1A_u$ Acetylene: Theoretical Considerations,'' G. Vacek,
 C. D. Sherrill, Y. Yamaguchi and H. F. Schaefer, {\it J. Chem. Phys.}
 {\bf 104}, 1774-8 (1996).
 ``The \~A $^1A_u$ State and the $T_2$ Potential Surface of Acetylene:
 Implications for Triplet Perturbations in the Flourescence Spectra of
 the \~A State,'' C. D. Sherrill, G. Vacek, Y. Yamaguchi,
 H. F. Schaefer, J. F. Stanton and J. Gauss, {\it J. Chem. Phys.} {\bf
 104}, 8507-15 (1996).
 Regards,
 George Vacek
 ---------------------------------------------------------------
 .... that's all.
 ====================================================================
                                              ,         ,      ,   ,
 Valentine P. Ananikov                        |\\\\ ////|     /////|
 NMR Group                                    | \\\|/// |    ///// |
 ND Zelinsky Institute of Organic Chemistry   |  |~~~|  |   |~~~|  |
 Leninsky Prospect 47                         |  |===|  |   |===|  |
 Moscow  117913                               |  |   |  |   |   |  |
 Russia                                       |  | A |  |   | Z |  |
                                               \ |   | /    |   | /
 e-mail: val &$at$& cacr.ioc.ac.ru                     \|===|/     |===|/
 http://nmr.ioc.ac.ru/Staff/AnanikovVP/          '---'
 '---'
   Fax +7 (095)1355328   Phone +7 (095)9383536
 ====================================================================