Summary: Imaginary freq. and rate constant



 Dear CCL people,
 to those of you who responded to my question about rate constant
 calculation based on transition state vibrational properties, I am
 very grateful for sharing your knowledge!
 I would like to send my thanks to the following people for
 their valuable coments:
                             Ryan Bettens
                             Joe Durant
                             Frank Jensen
                             Istvan Mayer
                             Arvi Rauk
                             John Rupley
                             Robert Q. Topper
                             Dom Zichi
 My original question was:
      As it is expected, MOPAC calculates one imaginary vibrational
 frequency at TS. Is it correct to use this frequency as pre-exponent
 factor (frequency factor) in Arrhenius rate constant equation?
 The unanimous answer was: NO, IT IS NOT CORRECT!
 RATIONALE:
 -------------------------------------
 From: Arvi Rauk <rauk;at;acs.ucalgary.ca>
 The imaginary frequency only measures the curvature
 of the reaction coordinate at the transition state. It may
 be used in the Bell formula for a quantum mechanical tunnelling
 correction to the classical rate which would only be a funcion of the
 barrier height.
 -------------------------------------
 From: Frank Jensen <frj;at;dou.dk>
       The A*exp(-E/RT) expression is equivalent to the TS expression
 kT/h * exp(deltaS/R) * exp(-deltaH/RT). Delta S and H may be calculated
 from S and H of the reactant and TS. S and H for a given structure may
 be calculated by means of stat. mech. from properties of the individual
 molecules, i.e. moments of inertia and vibrational frequencies. The
 important feature here is that there are 3N-6 vibrational contributions
 to the reactant, but only 3N-7 for the TS, i.e. the imaginary frequency
 is NOT involved. In TS theory it is assumed that the motion along
 the reaction coordinate (the imaginary frequency at the TS) is treated
 classically. Only if e.g. tunneling is considered does the imaginary
 frequency enter, for example by means of a Bell correction etc.
       So, to answer your question, no the imag. freq. has nothing
 to do with A in the A*exp(-E/RT) expression. A is essentially related to
 deltaS for the reaction.
 ---------------------------------------
 From: Joe Durant <jdurant;at;mephisto.ca.sandia.gov>
 No, the negative frequency associated with the reaction coordinate is
 not used in the calculation of the sum of states for the transition
 state!  I suggest a careful reading of Robinson and Holbrook
 "Unimolecular Reactions", or the kinetics texts by Johnston or Benson.
 I quote from Robinson and Holbrook, pg 12 "Q(double dagger) is the
 partition function for all the degrees of freedom of the activated
 complex except the reaction coordinate."  The translational degree of
 freedom gives rise to the kT/h term in the rate constant expression.
 The negative frequency is related to the barrier shape, and can be
 used in tunneling calculations.
      A closely related point is that the negative frequency does not
 contribute to the zero point energy of the transition state.
 ------------------------------------------
 From: John Rupley <local%rupley.UUCP;at;cs.arizona.edu>
 Transition State Theory assumes in its model free flight at
 the barrier top, i.e., a flat barrier top (corresponding to a zero
 frequency!).  The frequency factor is the frequency of attempt at
 barrier crossing, and can be taken as the frequency defining the shape
 of the reactant _well_ in the direction of the reaction coordinate.
 This frequency is built into the deltaS# part of the deltaG# of the
 Arrhenius expression.
      The imaginary frequency and the curvature of the potential
 surface at the barrier top also enter into the Kramers and related
 classical models, for reaction in a dense medium.
 -----------------------------------------------
 From: Dom Zichi <zichi;at;newt.nexagen.com>
       The correct frequency to be used in the transition state theory
       rate expression, kTST, is that of the reactant well, not the
       frequency describing the curvature at the barrier top.  This
       is easily seen from the following.  A simple TST rate assumes
       that every trajectory at the transition barrier q# which has a
       forward flux toward products will always end up as product.
       This leads to the following expression for kTST,
             /    /              .       .
       kTST= | dq | dp e(-H/kBT) q theta(q) delta(q#) / Qrct
             /    /
       where theta is a step function which equals one for positive
       flux toward products and 0 for negative flux, delta(q#) is a
       delta function of position along the reaction coordinate q at
       the barrier top and Qrct is the partition function for the
       reactant well,
             /    /
       Qrct= | dq | dp e(-Hr/kBT)
             /    /
                   2 2      2
       For Hr = m w q /2 + p /2 valid to q#, i.e.
       H(q)=  Hr(q)  q<q#
              Hp(q)  q>q#
       one can easily obtain kTST = (w/2*pi) exp(-H(q#)/kBT). The
 frequency term counts the number of times a forward reaction is
 attempted while the exponential term counts the number of
 trajectories that make it to the barrier top relative to the reactant
 well.
 -----------------------------------------------
 From: Istvan Mayer <mayer;at;cric.chemres.hu>
 The existence of an "imaginary frequency" at the TS is a reformulation
 of the fact, that the enrgy has a maximum (and not a minimum) along a
 given (normal) coordinate. Although it is expressed in frequency units,
 it is really a measure of the curvature of that point only, and has
 nothing to do directly with any reaction rate. The
 "quasi-thermodynamic"
 formulation of Eyring's "absolute reaction rate" or "transitiuon
 state"
 theory was a great-scale misunderstanding, which caused very much harm
 to science. 20 years ago I wrote: "The activated complexes are not true
 chemical species, do not fulfil the conditions defining a canonic
 ensemble, and therefore cannot be considered as independent subjects of
 the canonic distribution." (J. Chem. Phys. 60, 2564, 1974). This also
 means that there is no meaning to attribute any thermodynamic quantities
 to the TS.
      This does not mean that there can be no modern variants of
 transition state theory, which are free of the deffects of the old one.
 -----------------------------------------------
 From: Ryan Bettens <BETTENS;at;MPS.OHIO-STATE.EDU>
 Assuming that the rate constant, k, is given by, k = A exp{-E_a/(k_b T)}
 for all T. E_a is the barrier height and k_b is Boltzmann's constant. From
 RRKM theory A is given by,
 A = (alpha/h) f_{inf} G^*(E - E_a)/N(E - E_a).
 Where,
 alpha        is the reaction path degeneracy,
 h            is planks constant,
 f_{inf}      high-pressure centrifugal correction factor,
 E            the internal excitation energy,
 G^*(E - E_a) total number of states between 0 and (E - E_a) for the
              transition state (hereafter TS),
 N(E - E_a)   the number of states per unit energy range at (E - E_a) for
              the reactant.
 To get a nice interpretation for what A actually is we (a) ignore alpha,
 (b) ignore rotation (this cannot be done if the reactant is linear and the
 TS is non-linear and visa versa), (c) assume an harmonic oscillator and the
 approximation of Whitteen and Rabinovitch (1). A then becomes,
 A = {(E - E_a + a^* (E_z)^*)/(E - E_a + a E_z)}^(s - 1)
     x {Prod(i = 1 to s) nu_i}/{Prod(i = 1 to s - 1) (nu_i)^*}.
 Where,
 s   is the number of degrees of vibrational freedom of the reactant,
 a   is an empirical correction factor from ref. (1),
 E_z is the zero-point energy
 ^*  refer to the TS with one less degree of vibrational freedom.
 Approximating the first term, which is taken to the power of (s - 1), as
 unity, A finally becomes the ratio of the product of harmonic frequencies
 of the reactant to the product of the harmonic frequencies of the TS. Note
 that there is (s - 1) frequencies in the TS (the "missing" degree of
 freedom is the mode responsible for the chemical change, i.e., the mode
 with the negative force constant) compared with the reactant's s
 frequencies. Furthermore, if we make the assumption that the TS's
 frequencies are not much different from the reactant's frequencies, except
 for the "missing" frequency, we find,
 A = nu_i
 where nu_i is the mode in the reactant which when energized is responsible
 for the chemical change. Thus for a C-C single bond dissociating
 A ~ 900 cm^{-1} = 2.7 x 10^{13} s^{-1}. All of the above expressions come
 from Forst (2), which I suggest you take a look at if you really want a
 good understanding of the approximations made etc.
      Ref.'s
 (1)  G. Z. Whitten and B. S. Rabinovitch, J. Chem. Phys., V38 (1963) 2466.
 (2)  W. Forst, Theory of Unimolecular Reactions, 1973, Academic Press,
      New York.
 ---------------------------------------
 From: Robert Topper <topper;at;cooper.edu>
 May I suggest that you have a look at some of the recent papers by
 Donald Truhlar (Minnesota) and his group! They have developed two
 codes for the calculation of reaction rate constants. One of these,
 called MORATE, interfaces with MOPAC to carry out transition-state
 theory calculations for polyatomic systems. At its highest level of
 theory, MORATE includes corrections for multidimensional tunneling
 and zero-point motion. However, this would entail getting information
 about the potential energy all along the reaction coordinate and not
 just at the transition state.  However, MORATE can also carry out
 calculations using only the transition-state properties.
 With the experience you have gained using MOPAC you could probably
 use MORATE with ease. It is extremely well-documented and is
 available from QCPE and the CPC library...  and the methods are also
 described in a number of articles. They have been extensively
 benchmarked (where possible) against "exact" quantum scattering
 theory treatments and against experiment.  Overall, the methods work
 quite well for absolute rate constants and do a terrific job of
 getting at kinetic isotope effects.  Truhlar's group is constantly
 developing the technology and improving the user-friendliness of the
 code... and they have been doing so for a number of years now. It is
 written almost entirely in machine-portable FORTRAN and is available
 for a number of platforms.
 That said, you must realize that the absolute, accurate calculation
 of reaction rate constants can occasionally be frought with
 difficulties.  For example; transition-state theory makes some
 assumptions about the nature of reactive motion. In particular, TST
 is only "rigorously" correct when no back-reaction is possible (i.e.,
 when a molecule goes from reactants to products it never re-forms
 into reactants.  See "Chemical Kinetics and Dynamics" by Steinfeld,
 Francisco and Hase for a nice introductory discussion of this).  This
 is sort of a tough situation to achieve experimentally. However, in
 my own weird, brief experience in studying these effects I have never
 seen a case where the correction was larger than an order of
 magnitude...which is considered to be pretty good for rate constants.
 And that order of magnitude was in pretty extreme cases. There are
 other possibilities... the presence of "dynamical bottlenecks" to
 reaction can slow things down (this is a nonlinear effect).  Overall
 though, TST can be a very accurate and useful approximation.  One
 case in which you are pretty much guaranteed that TST will work is
 when the time scale of reaction is much much faster than the time
 scale of back-reaction... this can happen when the products' phase
 space is very much larger than that of the transition state.  Also,
 any reaction that involves dissociation is a good candidate.
 -----------------------------------
 THANK YOU ALL AGAIN!
 _______________________________
 Dr. Ivar Martin
 Department of Bioorganic Chemistry           tel: +372 2 526510
 Institute of Chemistry                       fax: +372 2 536371
 Akadeemia tee 15 EE0026                   e-mail: ivarm;at;boc.ic.ee
 Tallinn, ESTONIA